← Q3 journey review packs

Decisions & Approvals Log v2 (post-council)

1. Reconciliation: v1 to v2

Open by band: Band 1 = 16, Band 2 = 27, Band 3 = 42, Band 4 = 12.

Open by priority: P1 = 37, P2 = 49, P3 = 11.

Closures per journey council: J1 closed 3, J2 closed 2, J3 closed 4, J4 closed 3, J5 closed 4, J6 closed 4, J7 closed 3, J8 closed 0 (its pack is mid-review with the team; proposals only), J9 closed 4, J10 closed 1. Every delta id was verified against the v1 log: no missing ids, no double-closures, no item both changed and closed.

2. Conflicts: what dissolved, what still stands

The v1 log carried five items where two records or two packs disagreed. The council round dissolved the three pack-vs-pack conflicts; the two record-vs-record conflicts (where two written records disagree about whether a ruling exists) still need only a one-line confirm from Sheran each.

Dissolved in the council round (the packs now agree):

• DL-002 (M39_E post-purchase survey, J1 vs J6): dissolved. The J6 council withdrew its REWORK IN PLACE position and adopted the log's synthesis (J1's IN ADDITION shape: M39_E stays at purchase +1 day as pure feedback; J6's T+24h slot email is its own asset). No artefact still carries the conflicting shape. One Gate-1 confirmation covers both packs.
• DL-006 (REC-30 anniversary slot interim home, J7 vs J9): dissolved at working-assumption level. Both packs adopted the same interim posture: the slot runs under Journey 9's wave governance until the D2 churn score lands, then transfers to J7. Sheran's ruling on the conflict itself stays open, but no pack now carries a counter-position. J9 closed its duplicate item DL-111 as subsumed here, so one ruling surface remains.
• DL-007 (REC-06 decommission vehicle and midweek-warm, J7 vs J9): dissolved. The J7 council withdrew both contested claims: the midweek-warm re-home (which read against Sheran's 14 Jul STOP-both ruling) and the decommission-vehicle claim. The vehicle sits under Journey 9's REC-06 re-scope; J7 consumes the retirement schedule. Sheran confirms the final home.

Still standing (record-vs-record; each needs one line from Sheran):

• DL-003 (M22_P/M24_P vs ACT-07): the 6 Jul ruling vs the baseline's reopening. Both packs build to the ruling; the confirm closes it.
• DL-047 (travel-day yield order and Veezu call): tobe-composition's RULED labels (5-7 Jul) vs the baseline's open status. The J2 council verified the labels exist and rebased the pack on them; the confirm-and-close remains Sheran's.
• DL-076 (M21_E rework shape): the register's retire verdict vs the Gate-1 approval's rework shape. The pack treats the approval as the later, superseding verdict; Sheran signs off.

Checked and aligned (no conflict found): DL-005 (only the J2 council touched it; both packs already carry the paired J4c + J5x ruling), DL-013 (J1 and J5 both extended its launch-blocking scope; the changes compound, they do not compete), DL-014 (untouched; J7, J9 and J10 already agree), DL-015 (only J6 touched it; both packs carry the same owner/renderer line), DL-060 (only J4 touched it; its statement matches J3's fold proposal and both ride one ruling with DL-003), DL-089 (J1 and J6 converge: J1 adopted the boundary and closed its stale DL-039; J6 asks the same ruling to confirm the no-send deletion).

3. Open decisions by band

Band 1: portfolio-shaping rulings and hard external deadlines (16 items)

Rule these first. Each either shapes more than one journey's design or sits against a hard date (the Summer Leisure Wave is in-market late July; J8 ships 31 Jul). Most are grouped so one sitting covers several: the recovery ruling (DL-003 + DL-004 + DL-060, with DL-065 riding), the J1/J6 seam (DL-002 + DL-089, which also confirms the DL-039 deletion), and the wave-1 package (DL-125 + DL-108 + DL-109 + DL-006 + DL-007 + DL-008 + DL-009 + DL-010).

Band 2: ship and build gates (27 items)

Each of these gates a specific launch or build: a canvas cannot go live, or a build cannot start, until it is answered. Owners are named; several are Eng or Analytics answers rather than Sheran rulings.

Band 3: design-shaping (42 items)

The design proceeds on a stated working assumption; the ruling confirms or redirects it. Nothing here blocks a build from starting, but each shapes what gets built.

Band 4: later and phase-2 (12 items)

Later and phase-2: hygiene rulings, phase-2 scoping and items whose trigger has not arrived yet.

4. Closed by the council (28 items)

Each closure adopts a working assumption under the escalation discipline in COUNCIL-STANDARD.md: where no formal authority is genuinely required, the council decides, labels and moves on. The items stay in the log as the record of what was closed and why; any of them can be reopened at Gate-1 if Sheran disagrees with the adopted assumption. Grouped by the journey council that closed them.

Closed by the Journey 1 (first win to second transaction) council

DL-039: Journey 6 dependency named; interim D+30 exit = NO-SEND

• Status: closed: working assumption adopted.
• Resolution: council adopted working assumption: superseded by DL-089. Journey 6 is authored (20 Jul); the two-part boundary governs the J1/J6 seam, and the interim no-send rule survives only as the fallback if the DL-089 ratification slips (never to legacy lapse timers; batch exclusions stand). Cross-note: the J6 council independently routes the deletion of the interim D+30 no-send exit through Sheran's DL-089 ruling (Gate-1 agenda item 2); the two positions agree, and DL-089 is the single remaining decision surface for the seam.
• Why no formal authority was needed: The item's premise (Journey 6 unauthored) is stale and its surviving decision, the boundary itself, is already logged as DL-089; keeping both invites two verdicts on one seam.

DL-040: Travel-completion floor: departure + 1 day proxy until trip_completed

• Status: closed: working assumption adopted.
• Resolution: council adopted working assumption: interim travel-completion proxy = scheduled departure timestamp + 1 day, swapped to the real trip_completed event when it lands; B3 education touches stay exempt (ruled). Eng's confirm of departure-timestamp reliability at canvas evaluation time moves to the journey's validation plan.
• Why no formal authority was needed: No formal authority is genuinely required: the floor itself is Sheran-ruled (7 Jul), the proxy is the only buildable reading of it, the failure mode (a habit nudge landing slightly early on rare long/delayed journeys) is benign, and the remaining uncertainty is a verification task, not a decision.

DL-043: T21_P / T14_P trigger exclusivity (one push per win)

• Status: closed: working assumption adopted.
• Resolution: council adopted working assumption: treat T21_P/T14_P exclusivity as unverified and keep the dedupe on the shared win event in the build spec regardless (the STACK ban and win_moment_canvas_active key already require it), so either audit answer is safe. The exclusivity check itself stays tracked as a Lisa Braze verification task on the pack's Process tab.
• Why no formal authority was needed: An audit task with a design that is safe under both outcomes does not require formal authority; the log is for decisions, not checks.

Closed by the Journey 2 (pre-travel and travel day) council

DL-058: Delete GRW-16's gated T-2h not-opened recovery outright

• Status: closed: working assumption adopted.
• Resolution: council adopted working assumption: GRW-16's gated T-2h not-opened recovery is DELETED from the TO-BE design. It was drafted gated and never approved to build, it collides with GRW-17's T-3h push and M33_P to the minute inside the ratified 3-push morning cap, and its job (ticket opened before travel) is fully covered by the T-3h wayfinding push plus the persistent trip state. Gate-1 acceptance of the pack confirms; the deflection question (card OQ1) is answered by baselining the state, not by keeping the touch.
• Why no formal authority was needed: Striking a never-approved draft touch requires no formal authority (escalation discipline); the risks field of the item itself records none material. Keeping a permanently dormant touch in the matrix forces every future author to re-litigate it.

DL-059: Standardise Journey 2 volume on 710,616/yr confirmed everywhere

• Status: closed: working assumption adopted.
• Resolution: council adopted working assumption: 710,616/yr confirmed (auction 526,025 / BUN 184,591, Analytics tracker 8 Jul) is the single volume figure everywhere in the pack; the matrix rows' proxies (~210,000/yr, ~1,000/day, ~800/day) are retired from the pack's face. The residue is execution, not decision: Airtable matrix-row updates proceed on Sheran's go, and Analytics' per-beat daily derivation (AN-6 follow-through, email-leg volume at true scale) moves to the journey's validation plan.
• Why no formal authority was needed: AN-6 already ruled the audience is 3.4x the original sizing and the confirmed figure exists; standardising on it is the direct application of the full-set consistency rule, and the remaining work is tasks with named owners, not a choice requiring formal authority.

Closed by the Journey 3 (funnel drop-off recovery) council

DL-063: ACT-10 disposition: the shared Route-Watch engine, not a fourth dead-end moment

• Status: closed: working assumption adopted.
• Resolution: council adopted working assumption: ACT-10 is authored as the shared Route-Watch capture-and-alert engine (capture control + watch registration + window-open/coverage-live alert + the hand-off rule) consumed by ACT-25 and ACT-26, never a fourth standalone dead-end moment. The J3 pack carries ACT-10's card content under this framing and the journey still signs 8 masters; the crosswalk note keeps the card count honest.
• Why no formal authority was needed: Escalation discipline: the item's own risk line says none material, this is inventory hygiene. It applies the ruled moment-model rule (group by job; one mechanic serving two contexts = one owner, many treatments), and the cross-journey review independently adopts the same framing as common-capability build order 6. No formal authority is needed to not build a duplicate-alert bug.

DL-064: Interim power-user suppression flag for dwell recovery (legacy GBP 500 flag)

• Status: closed: working assumption adopted.
• Resolution: council adopted working assumption: dwell/hesitation recovery ships with power users suppressed on the legacy GBP 500 segmentation flag, the flag's known caveat carried in every read, re-cut on the AN-4/RFM power definition when it lands. Dwell recovery never ships with no power suppression. Break conditions: AN-4 landing (re-cut) or complaint-rate movement among flagged users (revisit the interim flag).
• Why no formal authority was needed: Guide ruling 11 already rules the substance (power users suppressed from hesitation recovery: their dwell is comparison, not confusion). The only open judgement is which interim flag, and an imperfect suppression beats chasing the best customers with confusion copy. That is an operating call within an existing ruling, not a new authority; the AN-4 dependency stays with Analytics as validation.

DL-066: Session-level suppression rule replaces the pairwise clauses

• Status: closed: working assumption adopted.
• Resolution: council adopted working assumption: one funnel nudge per session, deepest step reached wins, implemented once in canvas/SDK config, replaces the five pairwise suppression clauses the J3 cards carry. ACT-08 owns prove-on-landing, ACT-22 owns catch-on-hesitation, and ACT-13's user-initiated consent ask stays outside the rule per its signed row. The Product/Eng confirmation that session-scoped state is available to the in-app layer stays open as a validation on the Process tab, and per-step CVR is monitored by whether an upstream nudge was suppressed.
• Why no formal authority was needed: Escalation discipline: the rule produces the same customer outcome as the pairwise web with one place to build, test and audit, and it settles the ACT-08 vs ACT-22 price de-dupe both cards flag as unresolved. The stated risk (a deep-step nudge suppressing an upstream one the user needed) is a monitoring question, not a trade-off needing executive judgement; the fold-ACT-22-into-ACT-08 alternative stays deliberately unpicked until the first dwell read.

DL-068: ACT-05 BUN-variant investment level: ship-light-and-monitor

• Status: closed: working assumption adopted.
• Resolution: council adopted working assumption: the ACT-05 BUN confirm-abandon variant ships light as copy/config with guardrail monitoring only (flow completion, refund complaints), no dedicated holdout; analyst time goes to the ACT-02 headline test. Off-Avanti copy never names First Class or Standard Premium. Analytics can argue for a holdout before build (that veto is validation, not approval); the absence of a causal read on this arm is accepted knowingly.
• Why no formal authority was needed: Proportionate effort: the primary cohort is 2,728 users at GBP 0.62 net LTV against GBP 22.75 for auction upgrades, so the measurement cost exceeds the plausible value of a clean read. The item's own framing ('unless Analytics argues otherwise') already locates the remaining check with Analytics, which is validation; keeping it in the approvals log adds a decision where only an answer is needed.

Closed by the Journey 4 (loss recovery) council

DL-071: Enumerate which status pushes downgrade to critical-only when the LA is active

• Status: closed: working assumption adopted.
• Resolution: council adopted working assumption: where the Live Activity is active, the outbid-status and time-left pushes downgrade to critical-only (the LA carries both states); the knocked-out push (action-required) and the L0 close notification (transactional) stay at full delivery. The list is on the pack's logic tab; Gate-1 sign-off on the pack ratifies it.
• Why no formal authority was needed: This is a design enumeration inside an already-ratified cap (Sheran, 12 Jul), not a new authority question; the escalation discipline says decide, label and move on, and the build needed a concrete list.

DL-077: Park Android Live Updates as Phase 3 (after the iOS LA read)

• Status: closed: working assumption adopted.
• Resolution: council adopted working assumption: Android stays push-only for this release; a Phase-3 line for Android Live Updates opens only after (a) the iOS LA +4-week read and (b) platform availability, and no moment may ever depend on LA on either platform.
• Why no formal authority was needed: Pure sequencing with no material customer risk (Android users keep the unchanged push journey) and no investment committed; formal authority is not genuinely required until Phase-3 planning.

DL-078: Label the Journey 4 volume (159,053 losing bidders/yr) confirmed in the hub (closes roster decision f)

• Status: closed: working assumption adopted.
• Resolution: council adopted working assumption: the 159,053 losing-bidders/yr figure is treated as confirmed, sourced to the Analytics sizing tracker (Ivi, 8 Jul, 'all figures confirmed unless noted', where it replaced the assumed ~217k). The hub-row edit and roster decision (f) closure remain a housekeeping action for Sheran with Ivi cc'd.
• Why no formal authority was needed: Label housekeeping on a figure Analytics has already stood behind is not an executive judgement; keeping a P3 open against the programme's biggest recovery pool understated the journey in every review.

Closed by the Journey 5 (never-transacted activation) council

DL-083: Confirm Braze render-time targeting of bid-history + transacted-type, or brief the sync as launch-blocking

• Status: closed: working assumption adopted.
• Resolution: council adopted working assumption: the bid-state + transacted-type attribute sync is treated as required and is briefed to the Engineering Requests board as launch-blocking; T2/T3 stay blocked-not-buildable until Eng confirms render-time targeting (per ruling 11). The session-start-refresh alternative is rejected: stale-state exposure is the exact AS-IS defect being fixed.
• Why no formal authority was needed: Escalation discipline: this is validation (an Eng answer within the sprint), not formal authority. The action is already on the pack's Process tab and the ER board path; keeping it in the approvals log adds a decision where only an answer is needed.

DL-084: M63 operationalise-first staging before any REPLACE ships

• Status: closed: working assumption adopted.
• Resolution: council adopted working assumption: the M63 operationalise-first staging stands; nothing is switched off until Analytics declares a winner or folds the M63 control into the onramp holdout, and the eventual REPLACE clears at the normal launch gate (process step 11, Sheran clears launch).
• Why no formal authority was needed: The conservative default needs no separate approval: the risk the item guards against (switching off an attributed GBP 63k/yr surface without an incremental read) only arises if the staging is violated, and launch clearance already sits with Sheran.

DL-086: Split ACT-12's variants by audience owner (J5 never-transacted; J10 returning)

• Status: closed: working assumption adopted.
• Resolution: council adopted working assumption: ACT-12 variants split by audience owner; Journey 5 signs only the never-transacted variants (consent card + comprehension rotation) now, and returning-transactor personalisation waits for Journey 10's scope. The card's own conditions hold (Content Card audit, no incremental claim without a holdout, GBP 5 incentive cost sign-off).
• Why no formal authority was needed: Reversible and pre-empts nothing: Journey 10 is DEFINE, both packs already converge on the split, and holding J5's reach layer for an undefined Product scope costs coverage for no risk reduction.

DL-087: Resolve the 23 percent new-to-first-transaction baseline-vs-target conflict

• Status: closed: working assumption adopted.
• Resolution: council adopted working assumption: 23% is the TARGET (per the portfolio row); the interim baseline is the DN 30-day new-user purchase rate, 16.8% (confirmed as stated). Analytics still publishes the canonical baseline + target pair and definition, folded into the DL-009 reconciled sizing table; the pack quotes only that once published.
• Why no formal authority was needed: The two sources cannot both be right, and DN's 16.8-20.2% purchase-rate read makes a 23% baseline implausible as the same cut. A labelled working assumption unblocks holdout sizing while the validation proceeds; the remaining work is Analytics publication (validation), not executive judgement.

Closed by the Journey 6 (post-trip and habit) council

DL-094: Restate the J6-to-J8 hand-off; retire the legacy lapsing-recovery wording

• Status: closed: working assumption adopted.
• Resolution: council adopted working assumption: the restated J6-to-J8 hand-off stands; quiet past ~1.5x personal median (fixed 60d interim until D3/D2 land) flips journey_state to Journey 8's slip window; the legacy 'hand off to lapsing-recovery (61-90d)' wording is retired.
• Why no formal authority was needed: Pure alignment with standing ruling 4 (one lapse-owner per customer; journeys replace legacy sends); no formal authority needed. CRM wires journey_state at build.

DL-098: Propagate the persona correction into the J6 rows (copy tone only)

• Status: closed: working assumption adopted.
• Resolution: council adopted working assumption: Ivi's label-x-confidence persona model (25.2/53.6/19.6/1.7, confirmed 3 Jul) propagated into the J6 rows; persona = copy tone only, never cadence or targeting; the ~44%-Unclassified caveat retired; GRW-09's Commuter/Leisure timing question answered NO on the evidence.
• Why no formal authority was needed: Applies an existing confirmed correction; the remaining work is the D7 Braze sync of the persona label, an action not a decision.

DL-099: GRW-29 review-ask frequency policy (per trip + per store version)

• Status: closed: working assumption adopted.
• Resolution: council adopted working assumption: two-level review-ask cap (once per trip, once per App Store version) with journey-written state D9; frequent travellers rotate to the private survey once the store rating-prompt budget is spent.
• Why no formal authority was needed: No material risk; Product's confirmation of the native prompt vs deep-link sheet mechanics is a validation task at the design round, not a formal approval.

DL-100: Confirm the canonical M0-to-M1 figure and its label wording

• Status: closed: working assumption adopted.
• Resolution: council adopted working assumption: carry the brain figure (20.15% to 7.47%) as canonical, labelled 'cohort purchase/bid retention, month 0 to month 1 since sign-up' wherever cited; the Jun-2026 Tableau export variance (20.22/7.43) is noted where cited.
• Why no formal authority was needed: Hygiene, not authority; Ivi's confirmation of the canonical pull moves to the evidence and validation plan. J6's own KPI stays trip-anchored (DL-097).

Closed by the Journey 7 (win-back) council

DL-103: J7 double-read holdout design: within-band + REC-T01 propensity-vs-batch

• Status: closed: working assumption adopted.
• Resolution: council adopted working assumption: the double-read (within-band holdouts for journey lift + REC-T01 propensity-vs-batch for targeting value) is the only design compliant with the standing Gate-1 measurement ruling of 9 Jul (separate journey lift from model targeting where a model triggers the journey); Analytics owns sizing, holdout percentages and MDE per band, with leading indicators pre-agreed for underpowered bands.
• Why no formal authority was needed: The design follows an existing Sheran ruling, so no new approval is genuinely required; the open work is Analytics sizing, which is already an action on the pack's Process tab, not a decision.

DL-106: usual_route staleness fallback for deep-churned customers

• Status: closed: working assumption adopted.
• Resolution: council adopted working assumption: if the newest journey feeding usual_route is older than 12 months the route line drops (dropped-line pattern) and copy falls back to the strongest-live-route frame; Analytics implements the staleness test with CRM sign-off on the definition.
• Why no formal authority was needed: A definition-level detail with a designed fallback either way (fallback copy is mandatory for every merge field); no executive judgement is required, and a stale route asserted as 'your usual' has no defensible alternative.

DL-107: Two-touch hard cap per episode for REC-03 (supersedes the master's 2-3)

• Status: closed: working assumption adopted.
• Resolution: council adopted working assumption: the two-touch hard cap per episode stands for REC-03 (T2 only on UNSEEN, channel switch, Fri-Sun window kept); the supersession of the master's '2-3 touches' remains flagged in the pack as a change to the master's wording.
• Why no formal authority was needed: Direct propagation of the Sheran-ruled J8 episode shape to a colder audience; a third touch was never evidenced and the legacy canvas's 0.53% unsub is the warning against more pressure. Nothing remains that needs a ruling; the pack's Process tab no longer asks for the Gate-1 nod.

Closed by the Journey 9 (reactivation wave engine) council

DL-110: Add transaction-frequency band to the wave grid as the priority axis

• Status: closed: working assumption adopted.
• Resolution: council adopted working assumption: transaction-frequency band (one-time / multi / power) is the audience-B priority axis in the wave grid, with the persona overlay restated as message relevance, not propensity; the pack IS the Gate-1 review, so signing the pack ratifies the grid.
• Why no formal authority was needed: Design-level audience ordering on confirmed evidence (29.8% vs 4.3%, the strongest discriminator); no separate formal authority is needed beyond Gate-1 sign-off of the pack itself. Fallback if the D11 trait cannot be built for wave 1: persona x LTV as carded. The AN-4 power-definition dependency stays on the Analytics brief.

DL-111: REC-30 anniversary slot under wave governance; revisit when D2 lands

• Status: closed: subsumed by DL-006 (one ruling surface).
• Resolution: closed as subsumed by DL-006, which records the identical recommendation plus the J7 conflict on the interim home; the ruling happens there.
• Why no formal authority was needed: One home per fact: two open log items asking Sheran the same anniversary question invite divergent rulings.

DL-112: Suppression-as-treatment sunset send + deliverability gate on the 820k tier

• Status: closed: working assumption adopted.
• Resolution: council adopted working assumption (the conservative posture): sunset send at a ~12m working threshold pending the Analytics definition, then removal from all marketing audiences; the churned 820k tier stays capped at a partial test gated on the Cooling holdout read AND wired deliverability guardrails (complaint, bounce, blocklist monitoring).
• Why no formal authority was needed: Adopting the most protective option requires no risk acceptance; the threshold definition is already an Analytics deliverable on the pack's brief, and any LOOSER posture (touching the 820k early, skipping the sunset) would go back to Sheran as a new decision.

DL-113: Wave-learnings write-back loop to the Test & Learn roadmap

• Status: closed: working assumption adopted.
• Resolution: council adopted working assumption: per wave, Analytics (Ivi) logs the holdout read (+1wk leading indicators, end-of-wave holdout, cross-wave roll-up) and CRM (Lisa) logs the contact/deliverability guardrail read to the Test & Learn roadmap; the wave-1 entry feeds the September commuter wave design; no wave 2+ is scheduled without wave 1's logged read; Sheran reviews at the wave gate.
• Why no formal authority was needed: Operating cadence, not a formal-authority decision: it restates the already-decided campaign coupling rule with named owners; escalation discipline says decide, label, move on.

Closed by the Journey 10 (lifecycle-state home framework) council

DL-119: Confirm Journey 10's V2MOM objective label (O3 vs O4)

• Status: closed: working assumption adopted.
• Resolution: council adopted working assumption: O3 (habit loop) primary, O4 (personalised re-entry) secondary; the candidate metric (return-session to transaction rate) is a habit metric. Roster and portfolio rows to update at their next edit; Sheran can veto at the scoping session.
• Why no formal authority was needed: Escalation discipline: a labelling decision, not formal authority. The hub walk already carried O3 as the working assumption and the pack's own recommendation matched; keeping it open added load to the scoping agenda without a genuine trade-off to rule.

5. Full detail: open items (band order)

Every open item carries the full v2 field set from COUNCIL-STANDARD.md. Where a council changed an item in the round, the change notes record who changed what and why.

---

Band 1: portfolio-shaping rulings and hard external deadlines

DL-005: Ratify J4c + J5x as one paired restructure (Journey 1 and Journey 2)

• Journey(s): cross-journey, J1, J2
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: Sheran
• Recommended decision date: 2026-07-22

Decision required: Ratify, as ONE paired decision, the two restructures the roadmap still flags SHERAN TO RATIFY: J4c (Journey 1 = one transition, first or comeback transaction to second transaction, with four entry views: close-win, early/auto-win, IW/BUN purchase, win-after-return; one canvas family B1/B2/B3, one exit set) and J5x (Journey 2 = the pre-travel and travel-day beats extracted from the same legacy walks as their own trip-anchored journey).

Context and supporting evidence: The J2 pack now frames the ask as the paired ruling everywhere it appears (masthead, brief status and document-status lines, Process step 0, footer, strategy.md): ratify J4c + J5x together, because ratifying one without the other leaves the legacy win walks half-dismembered. Both packs now carry the same headline ask.

Recommended option: Ratify both together. Frame: Journey 1 = 'one transition, four ways in'; Journey 2 = the trip-anchored window that lets the disruption yield, the push cap and the state surface be designed once. The power-user entry (seg5) is deliberately not a fifth Journey 1 view (see the power-user roster item).

Rationale: The consolidation is real, not cosmetic: shared contracts (single-ask capture, win_moment_canvas_active dedupe, STACK ban, T31-is-the-receipt) bind all four views; the extraction fixes the orphaned-comm problem (M33_P) that happened precisely because no one owned the window.

Alternatives and trade-offs:

• Keep four walks (duplicates every contract four times)
• Ratify J4c alone (orphans the pre-travel beats)
• Keep pre-travel beats inside each win walk (status quo; re-creates the orphaned-comm problem)

Dependencies:

• Close-window payload contract and travel-completion floor (the two ruled seams)
• gate1-sequence.json journeys 1 and 2; roster clash 3

Risks:

• If ratification stalls, Journey 1's Gate-1 READY status is misleading (cards signed, journey not) and the J2 pack cannot go to Gate-1 review as a journey

Consequence of delay: Journey 1's Gate-1 READY status stays misleading (cards signed, journey not ratified) and the J2 pack cannot go to Gate-1 review as a journey.

Working assumption used in the journey pack(s): Both packs now frame the ask as the paired ruling everywhere it appears (masthead, brief status lines, Process step 0, strategy.md): ratify J4c + J5x together, because ratifying one without the other leaves the legacy win walks half-dismembered. Nothing ships before ratification.

Changed in the council round:

• J2 council (20 Jul 2026) changed context. Why: Both packs concur and the log already merged them; the pack's face had drifted behind the merge.

Merge note (v1): Merged: J1-D01 + J2 item 1. Both recommend ratifying jointly; no conflict.

Provenance: agent-derived DRAFT (both packs concur). Source ids: J1-D01 (J1/decisions.json); J2-#1 (J2/decisions.json).

DL-004: Recovery owner of record: J3 owns pre-bid abandon, J4 (ACT-07 D+7) owns post-close loss (J3, J4 and J5 agree)

• Journey(s): cross-journey, J3, J4, J5
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: Sheran (ratify); CRM authors the J3 spine
• Recommended decision date: 2026-07-22

Decision required: Ratify the recovery owner of record for the 13+ moments that reference a shared lost-bid / no-bid / abandoned-search recovery owner nobody had authored: the boundary is the live bid. Journey 3 authors and owns the pre-bid drop-off recovery spine (search, price-view, bid-add, dead-end, no-bid); Journey 4's ACT-07 D+7 non-return journey is the post-close lost-bid owner of record. Journey 5 registers its feeds (T2/T3 non-converters, BUN-box no-action) as inputs only and never sends into the pool.

Context and supporting evidence: Three packs reached the same split independently. J3: one episode per user, deepest-funnel-step-wins precedence, two-touch cap, T2 only on unseen T1, closes on bid/transaction/Route-Watch-set/opt-out/day-7 hand-back to J5; journey_state value = J3 episode. J4: ACT-07's D+7 journey is already specified as the convergence point (one enrolment, one send, convergence dedupe), which is the recovery-owner job in all but name. J5: Journey 3 authors the pool; J5 contributes source flags only. This is the single most-referenced unauthored dependency in the programme.

Recommended option: Ratify the split as stated and register both owners in the journey_state arbitration and the consumed-suppressions list, so every referencing moment points at a named owner. Reuses the Sheran-ruled J8 episode shape and the one-owner-per-customer ruling rather than inventing new machinery.

Rationale: Unowned, every upstream journey invents its own recovery and the stack pressure the minus 11.8 percent holdout read warned about returns; a single owner is also a measurement necessity (clean reads).

Alternatives and trade-offs:

• Author the recovery per moment (status quo in the cards; guarantees stacking and contaminated reads)
• A standalone cross-journey recovery mini-pack (duplicates ACT-07's spine)

Dependencies:

• journey_state exists (net-new: Analytics specs, Eng syncs)
• Batch audiences read journey_state (see the journey_state exclusion item)

Risks:

• If journey_state slips, the J3 spine cannot launch and Phase 1 ships in-session surfaces only

Consequence of delay: The 13+ moments that reference a shared recovery owner keep pointing at nobody; the J3 spine and ACT-07 build specs cannot register their suppressions, and the clash register regrows the rows this ruling exists to remove.

Working assumption used in the journey pack(s): Both packs are authored to the split (J3 owns the pre-bid drop-off spine; J4/ACT-07 owns post-close loss; J5 registers its feeds), and DL-065's 7-day window rides this same ruling.

Merge note (v1): Merged: J3 item 1 + J4-D04 + J5-D-07. The three journeys' recommendations AGREE on the split; no conflict.

Provenance: agent-derived DRAFT (three packs concur). Source ids: J3-#1 (J3/decisions.json); J4-D04 (J4/decisions.json); J5-D-07 (J5/decisions.json).

DL-003: M22_P/M24_P vs ACT-07: ruled 6 Jul, reopened in error by the baseline (confirm and close)

• Journey(s): cross-journey, J3, J4
• Priority and urgency: P1
• Status: open, confirm-and-close (stale-ruling conflict)
• Recommended decision-maker: Sheran (one-line confirmation); CRM wires the suppression
• Recommended decision date: 2026-07-22

Decision required: Confirm the 6 Jul ruling stands and close the reopened clash: M22_P/M24_P own PRE-close abandonment and are suppressed the moment auction_closed fires for that auction; ACT-07 owns post-close; no overlap window. Carry the implementation forward: add M22_P/M24_P to the consumed-suppressions list and wire the auction_closed suppression before ACT-07 goes live.

Context and supporting evidence: The baseline (compiled 20 Jul from the 5 Jul clash register) carried this as an open P1 clash. The J4 pack flagged the staleness: tobe-composition.json seg3 (6-7 Jul lock-down) records 'M22 PRE-CLOSE BOUNDARY (RULED, Sheran 6 Jul)', verified in the file. The item in this log is therefore the conflict resolution (two records disagreeing on whether a ruling exists), not the original boundary decision. Stakes if genuinely unruled: the abandonment pushes carry GBP 4.99M plus GBP 457k lifetime attributed (confirmed as attributed, not incremental).

Recommended option: Sheran confirms in one line that the 6 Jul ruling stands (or corrects the tobe-composition label if it does not). On confirmation this item closes and only the implementation actions remain (consumed-suppressions list, auction_closed suppression wiring, both already specified). The J3 pack's separate bid-placed split proposal is a NEW boundary decision and is logged separately (see the J3 section).

Rationale: A ruled item sitting open invites relitigating a decision Sheran already made; two sources of truth on the estate's highest-attributed clash is the exact drift the log exists to stop.

Alternatives and trade-offs:

• Re-confirm the full ruling from scratch if the lock-down provenance is doubted

Risks:

• Two sources of truth on the same clash until confirmed; the wiring gates ACT-07 go-live

Consequence of delay: Two sources of truth persist on the same clash, and the auction_closed suppression wiring that gates ACT-07 go-live stays unwired.

Working assumption used in the journey pack(s): Both packs build to the 6 Jul ruling (M22_P/M24_P own pre-close abandonment and are suppressed at auction_closed; ACT-07 owns post-close; no overlap window). The item stays open only for Sheran's one-line confirm-and-close.

Council alignment: STILL STANDS as a record-vs-record conflict (not pack-vs-pack): the 6 Jul ruling and the baseline's reopening cannot both be right. Both packs build to the ruling; only Sheran's one-line confirm-and-close resolves which record wins.

Provenance: source-ruled (Sheran 6 Jul, tobe-composition seg3); the baseline reopened it in error. Source ids: DL-014 (decisions-baseline.json); J4-D06 (J4/decisions.json).

DL-060: M22_P/M24_P audit + bid-placed split: no-bid slice folds into the J3 spine

• Journey(s): J3, J4
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: Sheran (rule), Lisa (audit), CRM (replace plan)
• Recommended decision date: 2026-07-22

Decision required: Rule the M22_P/M24_P boundary: audit their entry criteria, then split ownership at bid-placed. The no-bid slice folds into the J3 spine (replace, not stack); the bidder slice belongs to J4/ACT-07.

Context and supporting evidence: The J4 pack now states its side on the brief and logic tabs: the bidder slice belongs to J4/ACT-07 either way, the boundary at bid_placed is agreed by all records, and the no-bid slice's fate (fold vs keep-with-boundary) is presented as riding in one Sheran ruling with the DL-003 confirm, per cross-journey review item 5.

Recommended option: Lisa audits who M22_P/M24_P actually chase; rule bid-placed as the boundary; J3's spine replaces the no-bid slice per ruling 4 (journeys replace legacy sends). No J3 recovery ships while M22_P still fires on the same cohort.

Rationale: The attributed £96k/wk makes silent retirement politically hard, but the figure is attributed-not-incremental and the stacking risk is a customer-facing three-push pile-up.

Alternatives and trade-offs:

• Keep M22_P as-is and add J3 on top (stacking, ruled against)
• Retire M22_P outright before the spine is live (drops the only recovery that exists today)

Dependencies:

• Extends DL-003 (open, P1) with the J3-specific split
• M22_P entry-criteria audit: Lisa

Risks:

• Attributed revenue optics: replacing a '£96k/wk' push needs the attributed-not-incremental framing stated up front

Consequence of delay: The highest-attributed canvas in the estate (GBP 4.99M plus 457k attributed, not incremental) keeps firing on a cohort the new J3 spine also targets; replacing it without the audit and framing invites an optics fight at review.

Working assumption used in the journey pack(s): The J4 pack states its side on the brief and logic tabs: the boundary at bid_placed is agreed by all records; the bidder slice belongs to J4/ACT-07 either way; the no-bid slice's fate (fold vs keep-with-boundary) rides in one Sheran ruling with the DL-003 confirm, per cross-journey review item 5. No J3 recovery ships while M22_P still fires on the same cohort.

Changed in the council round:

• J4 council (20 Jul 2026) changed context. Why: The highest-attributed canvas in the estate (GBP 4.99M + 457k attributed, not incremental) was sitting under three records in three states; one decision surface replaces them.

Provenance: agent-derived DRAFT. Source ids: J3-#2 (J3/decisions.json).

DL-114: Rule what Journey 10 IS: the lifecycle-state home framework

• Journey(s): J10
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: Sheran + Product (scoping session)
• Recommended decision date: 2026-07-24

Decision required: Rule what Journey 10 IS: the lifecycle-state home framework (a shared capability with one owned moment) vs a peer journey bundling three moments.

Context and supporting evidence: Every other journey lands customers on Home and assumes it behaves intelligently (J8's tap-landing table proposes 'the personalised Home screen' for every touch, still unruled: DL-029). The cross-journey review sequences the framework as shared-capability build order 4, downstream of journey_state (NOW), the app_open + lifecycle-state event (NOW ER) and the content-card instrumentation fix (DL-013); the pack now carries this and the urgency chain (DL-029 walkthrough recommended 22 Jul, scoping by 24 Jul, J8 ships 31 Jul).

Recommended option: Rule J10 as the lifecycle-state home framework: one state model, one slot map (S0 trip-state hero / S1 recognition card / S2 opt-in banner / S3 tiles rail), one governance rule (state decides which journey's treatment renders; slots never stack). J10 owns the framework plus ONE moment (the ACT-12 returning treatment); GRW-09 and REC-21 render through it but stay owned elsewhere.

Rationale: Prevents nine per-journey Home hacks; matches the moment-model rule (different jobs sharing a screen are separate moments); the 9 Jul ruling already constrains the slot to at most one eligibility-rotated card, which only a state-resolver can enforce across J5/J7/J10 treatments.

Alternatives and trade-offs:

• Keep J10 as a peer journey owning all three moments (double-ownership with J6/J7; per-journey hacks persist)
• Hand the whole thing to Product as a surface project with no CRM journey (loses the treatment governance and the comms boundary rules)

Dependencies:

• The Sheran + Product scoping session (the DEFINE gate)

Risks:

• Unruled, every other pack's landing table stays 'to confirm' and DL-029-style decisions multiply per journey

Consequence of delay: Unruled, every other pack's landing table stays 'to confirm', DL-029-style destination decisions multiply per journey, and the scoping session has no ruling to work from.

Working assumption used in the journey pack(s): The pack is authored as the framework (one state model, one slot map S0-S3, one governance rule: state decides which journey's treatment renders, slots never stack); J10 owns the framework plus ONE moment (ACT-12 returning treatment); nothing builds before the DEFINE gate.

Changed in the council round:

• J10 council (20 Jul 2026) changed context. Why: The build-order dependency and the dates existed in CROSS-JOURNEY-REVIEW.md and the log but were not in the pack; the scoping session needs both to sequence correctly.

Provenance: agent-derived DRAFT. Source ids: D-J10-01 (J10/decisions.json).

DL-125: Wave-1 readiness gate and the slip-vs-ship rule (four conditions; wave 1 slips rather than ships degraded)

• Journey(s): J9, cross-journey
• Priority and urgency: P1 (P1, before the wave-1 freeze (late Jul))
• Status: open
• Recommended decision-maker: Sheran (gate); Lisa + Analytics run the checklist per wave
• Recommended decision date: 2026-07-24

Decision required: Ratify the wave-1 readiness gate and the slip-vs-ship rule: wave 1 freezes only when (1) BRZ-83 has landed and the 181/365d leisure windows are reconciled with BRZ-84, (2) the journey_state exclusion is live in every wave audience (DL-012), (3) the canonical baseline, holdout size and power calculation are published (DL-009), and (4) Gate-2 creative is approved under the claim and incentive rulings (DL-109/DL-108). If any is not true by the late-July window, wave 1 SLIPS rather than ships degraded.

Context and supporting evidence: Journey 9 is the only journey with a hard external calendar (Summer Leisure Wave in-market late Jul; BRZ-83 ~w/c 20 Jul, gate1-sequence confirmed). The estate's lesson is confirmed: untargeted outbound destroyed value (-11.8% power-user holdout; LAPSING50D -10.2%; 0.24% vs 1.5% legacy KPI). An untiered send is not reproducible (BRZ-84 unreconciled), an unexcluded send double-messages J7/J8 episodes, and an unpowered read at a ~1% response ceiling is unfalsifiable either way. The single permitted de-scope is already designed into the pack: if only the D8 comparison-anchor feed is missing, wave 1 ships Frame A (occasion) with the anchor line dropped.

Recommended option: Adopt the four-condition gate as the freeze checklist; Sheran holds slip-vs-ship authority; the campaign calendar moves before the engine's standards do.

Rationale: The default failure mode under a hard date is ship-degraded, which recreates the exact estate the engine replaces and burns the engine's one clean funding read (the wave-1 holdout is the go/no-go for the whole calendar).

Alternatives and trade-offs:

• Ship on the assumed dials and patch after (holds the campaign date but risks a contaminated, unpowered first read and an opt-out spike on the 61.6% push reach asset)
• Frame-A-only de-scope extended to other missing conditions (only the D8 feed degrades gracefully; the other three failures corrupt targeting or measurement, not copy)

Customer and business implications: Customers: no double-messaging of people already in triggered journeys; no untiered blast to the dormant pool. Business: a possibly later wave 1 in exchange for a readable funding decision on the ~£0.7M (assumed) reactivate+defend prize.

Dependencies:

• BRZ-83 (Eng/Ivi, ~w/c 20 Jul)
• BRZ-84 reconciliation (DL-010)
• journey_state batch read (DL-012)
• INITIATIVE SIZING row 22 + power calc (DL-009)
• Gate-2 creative under DL-108/DL-109. Risk: a slip compresses the August amplification window; the campaign's social/OOH legs are not gated by this rule and can hold the calendar.

Consequence of delay: Unruled, the hard date decides by default and wave 1 ships whatever is ready, degraded; the engine's first read is unusable and waves 2-4 inherit assumptions instead of evidence.

Working assumption used in the journey pack(s): The gate is in the pack (Campaign brief tab + strategy.md section 3) as agent-proposed DRAFT; the pack designs to all four conditions being met.

Provenance: agent-derived DRAFT (council review). Source ids: J9/decisions-delta.json (new, council round 20 Jul 2026).

DL-108: Wave-1 incentive posture: no monetary incentive in CRM; strike or ratify the master's GBP 5 line

• Journey(s): J9
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: Sheran
• Recommended decision date: 2026-07-22

Decision required: Rule the incentive posture for wave 1 and the status of the ACT-31 master's 'Instant Win / Beginner's Luck / £5 first-win' offer line.

Context and supporting evidence: The ruled lifecycle posture is no-incentive at non-RECOVER touches (incentives are RECOVER's lever; the £5 referral is gated post-positive-trip; DEF-20 ships no promo at any touch). The campaign's £5-off is a download/onboarding acquisition offer carried on social + OOH only (campaign ruling: CRM carries no download CTA), and its redemption cap at boosted scale is an open action (Joe/Lisa). Yet the ACT-31 master's crm_tobe line lists a '£5 first-win' offer, and the legacy win-back estate used a £5 incentive (M40-M42).

Recommended option: Wave 1 runs NO monetary incentive in CRM: audience A leads on Beginner's Luck as a de-risk mechanic and population stat, audience B on the prior-win re-anchor plus a substantiated comparison anchor. Any £5 first-win or discount cell is a Sheran-gated phase-2 holdout test inside the wave engine. Strike or explicitly ratify the master's £5 line so the moment record and the posture cannot drift.

Rationale: The barrier evidence is value confidence and comprehension, not price-of-entry; a discount answers the wrong barrier, trains discounting on a ~1% ceiling pool, and contradicts the ruled posture. Flagged honestly rather than silently resolved because the master record says otherwise.

Alternatives and trade-offs:

• Run a £5 first-win cell inside wave 1 with its own holdout (fastest read on incentive lift, but contaminates the engine's clean first read and pre-empts the posture ruling)
• Keep the £5 line for the churned 820k tier only (deepest-dormancy argument, but that tier is deferred anyway)

Dependencies:

• £5-off redemption cap confirmation (Joe/Lisa) before ANY CRM surface references the offer
• Wave-1 Gate-2 creative

Risks:

• Unruled, creative teams inherit a contradiction between the master record and the posture; a discount ships by default

Consequence of delay: Creative teams inherit a contradiction between the master record and the posture, and a discount ships by default in wave 1 (in-market late Jul).

Working assumption used in the journey pack(s): Wave 1 runs NO monetary incentive in CRM: audience A leads on Beginner's Luck as a de-risk mechanic and population stat, audience B on the prior-win re-anchor plus a substantiated comparison anchor; any GBP 5 or discount cell is a Sheran-gated phase-2 holdout test inside the wave engine.

Provenance: agent-derived DRAFT (applies the ruled no-incentive posture). Source ids: J9-D02 (J9/decisions.json).

DL-109: Wave-1 claim standard: substantiated anchor leads CRM; campaign claim stays market-level

• Journey(s): J9
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: Sheran (with the campaign/brand side; Caroline when back)
• Recommended decision date: 2026-07-22

Decision required: Rule which claim standard leads wave-1 CRM copy: the lifecycle substantiated-anchor standard or the campaign's 'Save up to 60% on First Class' constant claim.

Context and supporting evidence: Every J9 card carries the copy guardrail 'never a generic save-up-to; always a substantiated dynamic comparison anchor'. The campaign's constant market-facing claim IS 'Save up to 60%', and its substantiation vs walk-up First fares is an open campaign action (ASA). The campaign is wave 1 of this journey, so the two standards meet in one email.

Recommended option: CRM wave sends lead on the occasion/value-treat frame with a per-route substantiated anchor (the lifecycle standard); the 'up to 60%' claim appears only as a secondary market-level line, only after the substantiation lands, and never framed as a personal saving. Social and OOH carry the campaign claim; CRM carries the anchor.

Rationale: The lifecycle guardrail exists because generic value claims demonstrably underperform on this pool and carry ASA risk; the campaign claim is legitimate at market level once substantiated. Splitting by channel keeps both true.

Alternatives and trade-offs:

• Campaign claim leads CRM too (one voice across channels, but violates the card guardrail and is unsubstantiated today)
• Hold ALL 60% usage from CRM permanently (cleanest, but wastes a substantiated market claim once it lands)

Dependencies:

• 'Up to 60%' substantiation vs Brain fares + Seatfrog upgrade prices (open campaign action)
• Comparison-anchor fares feed (Eng)

Risks:

• Unruled, wave-1 creative is drafted against two contradictory standards days before in-market

Consequence of delay: Wave-1 creative is drafted against two contradictory claim standards days before in-market.

Working assumption used in the journey pack(s): CRM wave sends lead on the occasion/value-treat frame with a per-route substantiated anchor; the 'up to 60 percent' claim appears only as a secondary market-level line, only after substantiation lands, never framed as a personal saving; social and OOH carry the campaign claim.

Provenance: agent-derived DRAFT. Source ids: J9-D04 (J9/decisions.json).

DL-006: ACT-31 / REC-30 / REC-06 shared-moment split across J5, J7 and J9, with one open conflict on the REC-30 anniversary trigger

• Journey(s): cross-journey, J5, J7, J9
• Priority and urgency: P1
• Status: open, needs Sheran ruling on the conflict itself (anniversary interim home)
• Recommended decision-maker: Sheran (rule); CRM + Data implement
• Recommended decision date: 2026-07-22

Decision required: Ratify the ownership split for the three moments shared across Journeys 5, 7 and 9, and the wave arbitration that goes with it, before the Summer Leisure Wave (wave 1, in-market late Jul).

Context and supporting evidence: Three packs converged on the same boundary rule. AGREED: triggers keyed to the individual's timeline (day-60 hand-off, purchase anniversary, propensity gate, app re-open) belong to Journey 7; calendar waves belong to Journey 9; Journey 5 supplies the never-transacted tier definitions and first-win creative frame and authors no second engine; ACT-31 executes wholly under Journey 9's wave engine; hard arbitration: a customer inside a J7 or J8 episode is suppressed from every wave, wave membership never suppresses a triggered journey, and every wave audience reads journey_state before wave 1. CONFLICT (REC-30's per-user anniversary trigger, interim home): J7 (DL-J7-04) says the anniversary is a customer-clock trigger so it stays with J7 now (one moment, two treatments); J9 (J9-D01 with J9-D08) says the anniversary slot runs under J9 wave governance (anniversary/seasonal mutual exclusion per cycle) until the D2 churn score lands, then J7's score-triggered win-back replaces the heuristic.

Recommended option: Synthesis (agent-derived DRAFT): adopt the agreed boundary rule and the J9 interim posture for the anniversary slot. Reason: J7 cannot run triggered sends until the D2 score lands (model brief due end-Q2, not landed), so J7-now is ownership on paper with no live machinery, while the wave engine ships this month and can govern the slot's caps and mutual exclusion immediately. Hand the slot to J7 when D2 lands (both packs agree on that end state). The conflict is recorded and needs Sheran's ruling either way.

Rationale: Unruled, three journeys can legitimately send to overlapping dormant pools in one wave window and the wave-1 holdout read is contaminated from day one. Triggered-outranks-scheduled is already decided; this makes it executable for waves.

Alternatives and trade-offs:

• J7 position: anniversary trigger stays J7 from day one (cleaner ownership on the boundary rule, but no live J7 machinery until D2)
• Duplicate ownership with dedupe-at-send only (fragile: batch audiences do not read journey_state today)

Dependencies:

• journey_state readable by wave/batch audiences (Eng/CRM)
• BRZ-83 persona/tier sync (~w/c 20 Jul)
• D2 churn-score family (for the eventual hand-over)

Risks:

• Unruled before wave 1, double-messaging users already in triggered flows burns the one clean read; anniversary and a seasonal wave can collide in one month (cap: wave yields to the anniversary beat, 90d spacing per the map guardrail)

Consequence of delay: Unruled before wave 1 (in-market late Jul), double-messaging users already in triggered flows burns the wave's one clean read, and an anniversary beat and a seasonal wave can collide in one month.

Working assumption used in the journey pack(s): Both packs now carry the same interim posture (the inter-pack conflict is dissolved at working-assumption level): the REC-30 anniversary slot executes under Journey 9's wave governance with per-cycle anniversary/seasonal mutual exclusion until the D2 churn score lands, then the slot comes home to J7, which keeps the trigger's definition and end-state throughout. Sheran's ruling on the conflict itself stays open; the packs no longer disagree on the interim.

Council alignment: DISSOLVED at working-assumption level: J7 and J9 both adopted the log's synthesis (wave governance until D2 lands, then the slot transfers to J7), and each pack now states the other's position explicitly. Sheran's ruling on the conflict itself stays open, but the packs no longer disagree. J9 also closed its duplicate item DL-111 as subsumed here, so one ruling surface remains.

Changed in the council round:

• J7 council (20 Jul 2026) changed working assumption used in the journey pack. Why: Escalation discipline: J7 has no live machinery before D2 (model brief due end-Q2, not landed) while the wave engine ships this month, and both packs already agree on the end state. This removes the cross-pack contradiction named in CROSS-JOURNEY-REVIEW item 3. Sheran's ruling on the conflict itself stays open.
• J9 council (20 Jul 2026) changed working assumption used in the journey pack. Why: The consolidated log and cross-journey review (both 20 Jul) supersede the pack's authoring-time premise; the Gate-1 circle must review the real decision state, not a stale one. Sheran still rules the conflict itself.

Provenance: agent-derived DRAFT (three packs; split agreed, anniversary home conflicted). Source ids: J5-D-04 (J5/decisions.json); DL-J7-04 (J7/decisions.json); J9-D01 (J9/decisions.json).

DL-007: REC-06 disposition after the 14 Jul STOP ruling: decommission-and-replace, with two points needing a ruling

• Journey(s): cross-journey, J7, J9
• Priority and urgency: P1
• Status: open, needs Sheran ruling on the conflict itself (vehicle + midweek-warm)
• Recommended decision-maker: Sheran (confirm); CRM executes; agent re-issues the card
• Recommended decision date: 2026-07-24

Decision required: Settle REC-06's scope under the 14 Jul ruling (Sheran, pilot v42: STOP both M44 midweek-warm and M45 weekly-reactivation at launch; the reactivation job is fully superseded by the triggered layer; any future weekly pulse is a DIFFERENT job, a new healthy-actives GROW moment; the stopped +134% canvas is evaluated inside RECOVER): who carries the decommission moment, and whether midweek-warm has any surviving form.

Context and supporting evidence: Both packs accept the decommission-and-replace direction and both specify sequenced retirement of the 80+ M44P/M45P estate (not a flip), with holdouts during wind-down (J7) and a re-issued REC-06 card before Gate-1 sign-off so no reviewer approves the stale consolidation (J9). CONFLICT: (a) J7 (DL-J7-05) proposes re-homing midweek-warm outside J7 as a DEFEND-adjacent send (owner TBC), which reads against the STOP-both ruling: the ruling allows only a NEW carded GROW moment, never a re-home of the stopped send. (b) Both packs claim the decommission vehicle: J7 scopes REC-06 in J7 as the decommission/governance vehicle; J9 carries REC-06 as its decommission-and-replace moment.

Recommended option: Synthesis (agent-derived DRAFT): the 14 Jul ruling governs: no re-home of midweek-warm; any weekly pulse returns only as a newly carded healthy-actives GROW moment with a holdout. Vehicle: Journey 9 carries the REC-06 decommission moment (its wave engine provides the replacement coverage and its card re-issue is already specified); Journey 7 consumes the retirement schedule for its REC-03 arbitration. Attributed-revenue optics (GBP 44k attributed on M45P, attributed not incremental) are handled by the wind-down holdouts.

Rationale: Provenance honesty: a Sheran ruling outranks agent-drafted cards and pack preferences; building the card as written resurrects a ruled-dead broadcast and burns push consent (the 61.6 percent reach asset, confirmed).

Alternatives and trade-offs:

• J7 as the decommission vehicle (its REC-03 canvas absorbs the weekly-reactivation job; workable, but the replacement coverage story lives with J9's waves)
• Keep midweek-warm re-homed per J7 (reads against the 14 Jul STOP-both ruling; would need Sheran to explicitly relax it)

Dependencies:

• M44/M45 decommission sequencing (Eng/CRM)
• RECOVER evaluation of the stopped +134% canvas (Analytics)
• REC-06 card re-issue before Gate-1 sign-off

Risks:

• Unactioned, the Gate-1 circle reviews and signs a recommendation Sheran already overruled; retiring 50+ sends moves attributed revenue (state the attributed-not-incremental framing up front)

Consequence of delay: The Gate-1 circle reviews and signs a recommendation Sheran already overruled, and the 50+ legacy sends keep firing alongside the new triggered layer with no retirement schedule.

Working assumption used in the journey pack(s): J7 withdraws both: no re-home of midweek-warm (the 14 Jul STOP-both ruling governs; any future weekly pulse returns only as a newly carded healthy-actives GROW moment with a holdout), and the decommission vehicle sits under Journey 9's REC-06 re-scope; J7 consumes the retirement schedule so REC-03 never stacks on a legacy send. Pack and strategy amended.

Council alignment: DISSOLVED: the J7 council withdrew both contested claims (the midweek-warm re-home, which read against the 14 Jul STOP-both ruling, and the decommission-vehicle claim); the vehicle sits under Journey 9's REC-06 re-scope and J7 consumes the retirement schedule. Sheran's confirmation of the vehicle's final home stays open, but no pack now carries a counter-position.

Changed in the council round:

• J7 council (20 Jul 2026) changed working assumption used in the journey pack. Why: Provenance honesty: a Sheran ruling outranks an agent-drafted pack preference, and the earlier re-home proposal read against the STOP-both ruling. One decommission home also closes the two-owners defect named in CROSS-JOURNEY-REVIEW item 4. Sheran's ruling on the vehicle's final home stays open.

Provenance: source-ruled (14 Jul STOP, CONTEXT.md pilot v42) + agent-derived DRAFT (vehicle). Source ids: DL-J7-05 (J7/decisions.json); J9-D03 (J9/decisions.json).

DL-008: Contact Strategy Backbone dials: formal sign-off in one pass, before wave 1 (cap class C explicitly included)

• Journey(s): cross-journey, J9
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: Sheran (ratify); Lisa + Analytics (per-wave check)
• Recommended decision date: 2026-07-24

Decision required: Formally sign off the Contact Strategy Backbone: L0-L3 layers, cap values (weekly ceiling, Class-B 3/day, Class-C ~3/wk, flood-guard 4/day), quiet-hour tiers, S1=48h, and auction duration. Cap class C (scheduled programmes, '~3/week, no two promo journeys same week', still assumed pending ratify) must be ratified BEFORE the Summer wave 1 ships.

Context and supporting evidence: Six backbone dials were recommended 7 Jul pending ratify; F1/F2/F3 were ruled 6 Jul (F2 settled session=30min) but the formal sign-off list remains open per the 15 Jul status banner. Sheran corrected 15 Jul that the cap threshold is a rule to decide, not an Analytics-owned task. The J9 pack adds urgency: wave 1 coincides with the campaign's weekly CRM staycation drop (campaign decision 11 puts the drop inside existing caps; Lisa + Analytics own the check) plus a possible second knock per wave.

Recommended option: Ratify the six recommended dials as drafted in one pass (Class-B 3/day, Class-C 3/wk, flood-guard 4/day, session 30min already ruled, S1 T-48h, S2 close+2h) and, before wave 1, run the worst-case-user arithmetic (weekly drop + wave T1 + one T2) against class C; if it does not fit, the wave T2 yields (the drop is the campaign's committed cadence).

Rationale: Every journey pack restates policy only as journey-scoped application and links the backbone; unsigned dials propagate uncertainty into every pack. The cap is the only thing standing between 'wave engine' and 'the old blast estate with better branding'.

Alternatives and trade-offs:

• Sign dials journey by journey as each pack reaches Gate-1 (slower, risks divergence)
• Ship wave 1 on the assumed dial and ratify after (risks exactly the over-contact the engine exists to end)

Dependencies:

• Per-wave contact check vs the live Test & Learn roadmap (Lisa + Analytics)

Risks:

• Journeys building against unratified caps need rework if the numbers move; over-contact in week 1 of the campaign window; opt-out spike on the 61.6 percent push reach asset (confirmed)

Consequence of delay: Journeys building against unratified caps need rework if the numbers move; week 1 of the campaign window risks over-contact and an opt-out spike on the 61.6 percent push-reach asset (confirmed).

Working assumption used in the journey pack(s): Packs design to the drafted dials (Class-B 3/day, Class-C 3/wk, flood-guard 4/day, session 30min ruled, S1 T-48h, S2 close+2h); nothing ships on them before ratification and the worst-case-user arithmetic runs against class C.

Merge note (v1): Merged: baseline DL-008 + J9-D10 (class-C ratification pulled forward by wave 1). Compatible; the J9 ask raises priority to P1.

Provenance: agent-derived DRAFT. Source ids: DL-022 (decisions-baseline.json); J9-D10 (J9/decisions.json).

DL-009: One reconciled dormant-audience and baseline table (never-transacted pool, reactivation pool, organic baseline)

• Journey(s): cross-journey, J5, J7, J9
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: Analytics (Ivi)
• Recommended decision date: 2026-07-24

Decision required: Analytics (Ivi) publishes ONE reconciled audience-and-baseline table (as part of INITIATIVE SIZING row 22) covering: (a) the marketable never-transacted pool: retire the 820k / ~850k / 857.8k drift; tiers with per-tier contactability and the ~12m engagement filter; (b) the reactivation pool set: ~600k dormant (roster) vs 597,561 true reactivation universe (workbook, confirmed on sheet) vs ~317k contactable reactivation pool (corrected figures) vs 1,592,082 Churned (Contactability); settle 29.8 percent vs 32 percent Lapsed-PU historical reactivation with the source named; (c) the canonical organic-reactivation baseline (0.84 percent Snowflake vs 0.24 percent legacy KPI vs the ~1 percent response-ceiling band) plus the pool definition for the wave-1 read.

Context and supporting evidence: Three packs independently hit the same wall: contradictory pool and baseline figures with no canonical home. The campaign's ~GBP 0.7M reactivate+defend prize assumes 1 percent of the ~317k contactable pool at GBP 219.94 per reactivation (GBP 219.94 confirmed; the prize stays assumed). The workbook itself flags that Audience Pool and Movers & Sliders churned counts differ by an order of magnitude (TBC: Analytics).

Recommended option: One table, one owner, one home; every pack and deck cites only that. Until published, all mentions keep the label their source carries and are never blended. The GBP 0.7M line stays 'assumed' in every deck until wave 1's holdout converts it to a measured read. Value reactivations at GBP 219.94 (confirmed), never GBP 22.75.

Rationale: One home per fact: wave sizing, holdout design, the win-back business case and the thesis narrative all hang off these numbers; a wave read against the wrong baseline is unfalsifiable.

Alternatives and trade-offs:

• Carry all figures with labels indefinitely (invites a Gate-1 challenge on arithmetic)
• Use 0.84 percent by default (most recent, but not ratified as the standard)

Dependencies:

• INITIATIVE SIZING row 22 (Analytics)
• Workbook population-base clarification (flagged in out-reactivation.md)

Risks:

• A business case built on the wrong denominator over- or under-states the programme by 2-5x; wave 1 declared a success or failure against a baseline nobody agreed

Consequence of delay: A business case built on the wrong denominator over- or under-states the programme by 2-5x, and wave 1 is declared a success or failure against a baseline nobody agreed.

Working assumption used in the journey pack(s): Until the table publishes, every mention keeps the label its source carries and figures are never blended; the GBP 0.7M line stays assumed in every deck; reactivations are valued at GBP 219.94 (confirmed), never GBP 22.75.

Merge note (v1): Merged: J5-D-03 + DL-J7-11 + J9-D07. Same owner, same deliverable frame; no conflict.

Provenance: agent-derived DRAFT (three packs concur). Source ids: J5-D-03 (J5/decisions.json); DL-J7-11 (J7/decisions.json); J9-D07 (J9/decisions.json).

DL-010: Lapse threshold family: kill the persona entry thresholds, reconcile the leisure 181/365d overlay with BRZ-84 in one decision

• Journey(s): cross-journey, J7, J9
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: Analytics (Ivi) reconciles; Sheran ratifies
• Recommended decision date: 2026-07-24

Decision required: Reconcile the whole lapse-threshold family in ONE decision before wave 1: (a) kill the REC-03 persona entry thresholds (Commuter >30d, Mixed >60d, Leisure >120d): contradicted by the point-of-lapse evidence (business users are 1 to 4 weeks SLOWER than leisure at every transition, confirmed; empirical Lapsed ~26wk) and superseded by the ruled day-60 window; uniform day-60 entry via journey_state, the D2 score sets intensity (HIGH/MID/LOW), persona sets copy register only; (b) complete the BRZ-84 reconciliation of the leisure 181/365d overlay (a persona OVERRIDE of the canonical 0-60/61-90/91-240/240+ cut-offs) before wave 1 sends; (c) rule now that the September commuter wave gets NO tighter commuter window without new Point-of-Lapse evidence: canonical bands are the default for every non-leisure wave.

Context and supporting evidence: The J7 pack asked for the threshold family to be reconciled as one decision, not five scattered ones; the J9 pack asked for the same reconciliation from the wave side. The one empirical timing dataset (Point-of-Lapse workbook) supports the leisure direction (80 percent of next orders in by ~182d) but contradicts every commuter-faster assumption.

Recommended option: Analytics reconciles placeholders vs empirical vs the J9 overlay in one pass; Sheran ratifies; single definition, no drift. Persona-differentiated recency thresholds in the assumed direction do not ship.

Rationale: Keeping the placeholder thresholds would gate entry on a disproven assumption; minting overrides per wave without evidence recreates the placeholder mess the workbook just corrected; two live lapse definitions make wave audiences non-reproducible and the holdout read unauditable.

Alternatives and trade-offs:

• Keep persona thresholds until Analytics validates (ships a disproven assumption)
• Per-wave bespoke windows (maximum relevance, guaranteed definitional drift)

Dependencies:

• BRZ-84
• D2 band spec
• Segmentation ratification session
• Point of Lapse extract (out-reactivation.md section 3.1)

Risks:

• Leisure users at day 60 may be mid-cycle, not lapsed (mitigated: intensity, not entry, is score-set; LOW band gets no triggered touch)

Consequence of delay: Persona entry thresholds contradicted by the point-of-lapse evidence can ship by default, and wave 1 targets on an unreconciled 181/365d overlay, making its read non-reproducible.

Working assumption used in the journey pack(s): Uniform day-60 entry via journey_state; the D2 score sets intensity, not entry; persona-differentiated recency thresholds in the assumed direction do not ship.

Merge note (v1): Merged: DL-J7-02 + J9-D05. Same reconciliation from two sides; no conflict.

Provenance: agent-derived DRAFT (two packs concur); the day-60 window itself is source-ruled (20 Jul). Source ids: DL-J7-02 (J7/decisions.json); J9-D05 (J9/decisions.json).

DL-002: M39_E post-purchase survey vs the habit arc: J1 and J6 recommend different dispositions

• Journey(s): cross-journey, J1, J6
• Priority and urgency: P1
• Status: open for Gate-1 confirmation (one ruling covering J1 and J6); the inter-pack conflict is dissolved: the J6 council withdrew the REWORK IN PLACE position and adopted the log's synthesis (J1's IN ADDITION shape) as the working assumption, so no artefact still carries the conflicting shape
• Recommended decision-maker: Sheran
• Recommended decision date: 2026-07-22

Decision required: Rule the disposition of M39_E (POST_UPGRADE_SURVEY_EMAIL, ACTIVE at purchase +1 day, GBP 165,839.80 lifetime attributed, confirmed as attributed not incremental): IN ADDITION alongside the GRW-01 habit arc, or REWORK IN PLACE as Journey 6's post-trip T+24h email asset.

Context and supporting evidence: J1 pack now states the contested verdict explicitly (brief existing-asset table, TO-BE map open questions, strategy.md section 5): Journey 6 recommends rework in place as the post-trip T+24h slot; one Sheran ruling covers both packs. IN ADDITION is held as J1's working assumption because it respects the A2 purchase-anchor ruling and touches no revenue-attributed asset before Gate-1.

Recommended option: Synthesis (agent-derived DRAFT): adopt the J1 position for Q3 (IN ADDITION with the channel-separation rule) because it respects the A2 purchase-anchor ruling, touches no active revenue-attributed asset before Gate-1, and preserves the pre/post baseline. Revisit the J6 rework when the J6 spine enters build: if Sheran prefers one email at T+24h, the A2 anchor must be explicitly re-ruled at the same time. Either way, J1 B3's proposed D+1 test arm for surveyed users stays blocked (both packs agree on the contingency).

Rationale: The two positions cannot both build: one keeps M39_E at purchase +1d untouched, the other moves and rewrites it. The J1 route is the lower-risk interim; the conflict itself needs a ruling before either journey's suppression list can finalise.

Alternatives and trade-offs:

• J6 position: rework in place as the T+24h post-trip email (one email, no stack; requires re-ruling the A2 purchase anchor)
• Retire M39_E and build net-new (both packs reject: loses the baseline)

Dependencies:

• M39_E trigger-coverage check (Lisa + Analytics)
• J1/J6 boundary item (grw01_canvas_active key)

Risks:

• Unruled, the GRW-01 habit canvas suppression list and the J6 spine cannot both finalise; two emails can land D+1/D+2 for email-only users under the J1 shape (flood-guard absorbs)

Consequence of delay: Unruled, the GRW-01 habit-canvas suppression list and the J6 spine cannot both finalise; the J6 T+24h email asset stays unbuildable, and two emails can land D+1/D+2 for email-only users under the J1 shape (flood-guard absorbs).

Working assumption used in the journey pack(s): M39_E stays at purchase +1d as pure feedback; the J6 T+24h slot email is the journey's OWN trip-anchored master asset; if M39_E lands the same calendar day the slot email shifts to T+48h; M39_E counts to the weekly cap, not J6's episode cap; new guardrail: M39_E survey completes must not fall. Rework-in-place returns only at build with the A2 anchor explicitly re-ruled.

Council alignment: DISSOLVED in the catch-up round: the J6 council withdrew the REWORK IN PLACE position and adopted the log's synthesis (J1's IN ADDITION shape) as its working assumption, realigning the J6 pack (brief, map, matrix, T&L, walkthrough email mock). No artefact still carries the conflicting shape. One Gate-1 confirmation covers J1 and J6 together (Gate-1 agenda item 4).

Changed in the council round:

• J1 council (20 Jul 2026) changed context. Why: The register asked for the M39_E verdict to be explicit on the GRW-01 rows; presenting it as settled would have misled the Gate-1 review group.
• J6 council (20 Jul 2026) changed status. Why: J6 was the only artefact still carrying the conflicting shape; the council realigned the pack (brief, map, matrix, T&L, walkthrough email mock) to the DL-002 synthesis, which the clash register's own recommendation also supports. Removes the sharpest inter-pack contradiction and protects an active attributed asset pre-Gate-1.
• J6 council (20 Jul 2026) changed working_assumption. Why: Council-adopted working assumption per the assessment; Sheran confirms DL-002 once, for J1 and J6 together (Gate-1 agenda item 4).

Provenance: agent-derived DRAFT (both positions); the conflict is recorded, not resolved. Source ids: DL-013 (decisions-baseline.json); J1-D03 (J1/decisions.json); JD6-06 (J6/decisions.json).

DL-089: Journey 1 / Journey 6 boundary: two-part split (canvas days vs trip-anchored jobs)

• Journey(s): J6, J1
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: Sheran (boundary) + Lisa (state key)
• Recommended decision date: 2026-07-24

Decision required: Ratify the Journey 1 / Journey 6 boundary: J1 owns the win-to-2nd-transaction forward asks for its 30 canvas days (grw01_canvas_active); J6 owns trip-anchored reflection and review from trip 1, and the full spine including the habit ladder from the 2nd transaction onward.

Context and supporting evidence: The J1 pack now carries DL-089's two-part boundary as its working assumption (TO-BE map rules, D+30 map node, strategy.md sections 3d and 5): J1 owns forward-looking asks for its 30 canvas days via grw01_canvas_active; J6 owns trip-anchored reflection and review from trip 1 with forward asks suppressed while the canvas is active; no-send is retained only as the fallback if ratification slips. Both packs now carry the same contract.

Recommended option: Adopt the two-part boundary; J6's forward-looking asks (tracker default, GRW-09 nudge) suppress while grw01_canvas_active; J1's B-canvas D+30 non-convert exit hands into GRW-09 as the GRW-01 draft already states. Rule together with DL-039: the cross-journey review recommends deleting J1's stale interim D+30 no-send exit in the same ruling, since J6 now exists to receive the hand-off.

Rationale: Jobs, not counts, define the split: J1's job is habit ignition from the win event; J6's per-trip jobs (value proof, review) have no J1 equivalent and losing them for first-trip customers wastes the peak-trust window.

Alternatives and trade-offs:

• Hard boundary at the 2nd transaction (first-trip customers get no reflection or review ask at all)
• J6 owns everything post-trip including first-winners (duplicates J1's canvas asks)

Dependencies:

• grw01_canvas_active journey_state key confirmed (CRM/Lisa, proposed in the GRW-01 draft)

Risks:

• Without the key, a first-time winner gets J1 D+3 and J6 T+24h forward asks in the same week

Consequence of delay: Without the boundary key, a first-time winner gets J1's D+3 and J6's T+24h forward asks in the same week; and while the ruling waits, the stale no-send exit survives only as a fallback record.

Working assumption used in the journey pack(s): Both packs carry the same contract: J1 owns forward-looking asks for its 30 canvas days via grw01_canvas_active; J6 owns trip-anchored reflection and review from trip 1 with forward asks suppressed while the canvas is active; the interim no-send rule survives only as the fallback if ratification slips.

Council alignment: AGREEMENT, converging from two sides: J1 adopted the two-part boundary as its working assumption and closed its stale DL-039 as superseded; J6 asks that Sheran's DL-089 ruling also confirm the deletion of J1's interim D+30 no-send exit (Gate-1 agenda item 2). One ruling covers the boundary and the deletion.

Changed in the council round:

• J1 council (20 Jul 2026) changed context. Why: Cross-journey review item 2: the two proposals were compatible but not identical, and J1's premise no longer held.
• J6 council (20 Jul 2026) changed recommendation. Why: The assessment aligns the boundary wording to the consolidated log and puts DL-089 and DL-039 to Sheran as one ruling (Gate-1 agenda item 2).

Provenance: agent-derived DRAFT. Source ids: JD6-02 (J6/decisions.json).

DL-029: J8 landing destinations, and one default-destination contract for every pack (extends to the J10 framework)

• Journey(s): J8, cross-journey, J10
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: Sheran (walkthrough) with Product input
• Recommended decision date: 2026-07-22

Decision required: Two linked rulings at the team walkthrough: (a) rule the landing destination for each J8 touch (where every push, email and banner tap lands in-app); (b) rule that 'the personalised Home screen' referenced by every journey's landing tables IS the Journey 10 lifecycle-state home framework, so landing-destination decisions close against one definition instead of nine walkthrough debates.

Context and supporting evidence: The J8 feedback round is closed bar this: landing destinations were deferred to a team walkthrough; every touch has a copy direction and a channel rule but not a ruled tap destination. The J10 pack (D-J10-09) generalises: J8's tap-landing table proposes the personalised Home screen for Touch 1 push, Touch 1 email and Touch 2 push, and eight more packs will produce the same rows. The destination surface does not exist yet; J10 defines it.

Recommended option: Hold the walkthrough and rule the lifecycle-state Home as the default deep-link destination for journey touches (exceptions: context-specific targets like a tracked auction, per test design). J8 does NOT wait on the framework build: it ships 31 Jul landing on today's Home (or the per-touch surfaces the walkthrough picks) with a follow-up migration once the framework exists.

Rationale: Destinations shape both eng scope (deeplinks) and the customer journey; one ruled default gives deep-link QA a single contract and stops DL-029-style decisions multiplying per journey.

Alternatives and trade-offs:

• Default every touch to the app home screen (loses the mission context)
• Rule destinations journey-by-journey at each walkthrough (slowest, and drifts)

Dependencies:

• Team walkthrough scheduled
• J10 framework scoping session (same session or adjacent)

Risks:

• Unruled destinations block the Braze build spec even after Gate-1 lock; if the framework slips, J8 must not wait on it

Consequence of delay: Unruled destinations block the Braze build spec even after Gate-1 lock; every journey's landing table stays 'to confirm' and destination debates multiply per pack. J8 ships 31 Jul against this (walkthrough recommended 22 Jul, J10 scoping by 24 Jul).

Working assumption used in the journey pack(s): J8 does not wait on the framework: it ships landing on today's Home (or the per-touch surfaces the walkthrough picks) with a follow-up migration once the framework exists; the lifecycle-state Home is the recommended default deep-link destination for journey touches.

Merge note (v1): Merged: baseline DL-029 + D-J10-09 (which explicitly extends it). Compatible; J10 adds the general contract.

Provenance: TBC (per-touch destinations need the walkthrough) + agent-derived DRAFT (default contract). Source ids: DL-001 (decisions-baseline.json); D-J10-09 (J10/decisions.json).

---

Band 2: ship and build gates

DL-031: Analytics delivers the J8 numbers and definition checks (EOP 21 Jul)

• Journey(s): J8
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: Ivi (Analytics)
• Recommended decision date: 2026-07-21

Decision required: Analytics delivers the J8 numbers and definition checks due end of play 21 Jul.

Context and supporting evidence: Eight asks: the six above plus two trailing non-blocking validations the council surfaced: (7) validate the ~20% slip-save vs ~0.5% churned benchmark against our own cohorts (it is labelled assumed in the T&L but reads as fact in three other pack homes), and (8) confirm the consent-state totals 47,836 / 18,477 / 11,327 as official sizing (currently CI-schema source with purchase-recency as the activity proxy; the caveat was lost from the matrix caption and should be restored, proposed edit 3 in J8/COUNCIL-ASSESSMENT.md).

Recommended option: Hold Gate-1 lock until the D1 thresholds land (the one launch-gating item); let D3/D4-class numbers trail as non-blocking fallbacks per the ruled priority order (Sheran-ruled pattern, 20 Jul).

Rationale: The pack's Analytics brief panel already marks only D1 as launch-blocking; blocking the lock on the rest would over-gate.

Alternatives and trade-offs:

• Lock Gate-1 on assumed thresholds and re-cut after (breaks the honest-figures rule)

Dependencies:

• Ivi's capacity; Slack ask already composed

Risks:

• Analytics submitted no J8 feedback in the round; late numbers slip the lock

Consequence of delay: Late numbers slip the Gate-1 lock; the D1 per-persona trigger thresholds are the single launch-gating item.

Working assumption used in the journey pack(s): The lock holds until D1 lands; D3/D4-class numbers trail as non-blocking fallbacks per the ruled priority order; the two council-added validations (slip-save benchmark, consent-state totals) trail as non-blocking within the Ivi bundle.

Changed in the council round:

• J8 council (20 Jul 2026) changed context and supporting evidence. Why: Both figures are load-bearing in the Gate-1 read; neither blocks the lock, but both need an owner and a due date, and they belong with the existing Ivi bundle rather than as new items.

Provenance: agent-derived DRAFT. Source ids: DL-004 (decisions-baseline.json).

DL-001: Close-event contract: instant_win / auction_won / auction_closed (Q5, extended)

• Journey(s): cross-journey, J1, J4, J5
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: Eng/PMM
• Recommended decision date: 2026-07-22

Decision required: Eng/PMM confirm the full auction-termination event contract in one audit: (a) does instant_win emit auction_won (council question Q5: gates the GRW-01/ACT-24 entry triggers, B3's entry event and the shared win_moment_canvas_active dedupe key); (b) does every termination path (instant win, buy-out, pre-close resolution, all five close reasons) emit auction_closed with readable close-reason metadata (gates Journey 4's outcome fork, the L0 close notification and the Live Activity end-state).

Context and supporting evidence: The baseline carried Q5 as a hard gate (opendecisions + clash register, win-moment cluster). The J4 pack found the same contract ambiguity under its entry fork and asked for one folded Eng check rather than a second ticket. The J1 pack restated Q5 as one of Journey 1's two build prerequisites. A termination path that skips or mislabels the event double-fires the loss journey or sends 'you lost' to a winner.

Recommended option: One Eng/PMM audit covering both questions, chased to an answer at the Gate-1 review itself (hours, not weeks). If instant_win does not emit auction_won: add instant_win to the entry-trigger set and the shared dedupe key on the Braze side rather than remapping events upstream; B3 falls back to the ACT-24-emitted entry event already sketched in the rows.

Rationale: One Eng check answers three journeys' entry gates; splitting it risks two half-answers. Trigger-set addition is a Braze-side change; remapping events risks breaking the L0 receipt layer.

Alternatives and trade-offs:

• Eng remaps instant_win to emit auction_won at source
• Separate ticket per journey (rejected by the J4 pack: two half-answers)

Risks:

• Unconfirmed contract means the win canvases can double-fire or miss IW winners entirely; the loss fork can misfire on never-ran auctions (falsehood risk on outcome copy)

Consequence of delay: Win canvases can double-fire or miss instant-win winners, the loss fork can misfire on never-ran auctions (falsehood risk on outcome copy), and three journeys' entry triggers (GRW-01, ACT-24, the J4 outcome fork) stall at trigger wiring.

Working assumption used in the journey pack(s): If instant_win does not emit auction_won, the fix is Braze-side: add instant_win to the entry-trigger set and the shared win_moment_canvas_active dedupe key; B3 falls back to the ACT-24-emitted entry event already sketched in the rows. Builds spec to that contingency; no win canvas goes live until the audit answers.

Merge note (v1): Merged: baseline DL-001 + J4-D05 (extension to the full close-event contract) + J1-D12 part (a) (restated build hard-gate). All three agree; no conflict.

Provenance: agent-derived DRAFT. Source ids: DL-007 (decisions-baseline.json); J4-D05 (J4/decisions.json); J1-D12 (J1/decisions.json).

DL-030: J8 Live Activity consent position + webhook start shape

• Journey(s): J8
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: Sheran
• Recommended decision date: 2026-07-22

Decision required: Rule the J8 Live Activity consent position and approve the webhook start shape.

Context and supporting evidence: iOS Live Activities render with notifications off (device-confirmed 14 Jul), so the LA branch is a reach experiment to push-declined iOS customers. The 15 Jul reshaped proposal has the journey canvas keep eligibility, holdout and caps, and fire the LA start via a webhook step (to Braze's API if it holds start tokens, else to our backend, which works today). Sheran's consent ruling and a yes to the webhook shape are both outstanding.

Recommended option: Agent-derived DRAFT: approve the webhook shape and run the LA branch as the iOS-only reach cell already specified in the pack (iOS 17.2+ filter), treating LA's own per-app setting as the consent surface.

Rationale: The pack, contact-rules row and T&L cell are already built to this shape; the platform facts (iOS-only reach, Android Live Updates need notification permission) are confirmed and applied.

Alternatives and trade-offs:

• Hold the LA cell out of the J8 release entirely and ship ladder-only

Dependencies:

• DL-033 token-holder answer decides the webhook target, not the shape

Risks:

• Consent ambiguity on a new surface class; ruling it explicitly protects the PECR posture

Consequence of delay: The LA cell's consent posture stays ambiguous on a new surface class and the cell cannot enter build; core J8 locks without it, holding the LA cell.

Working assumption used in the journey pack(s): Webhook start shape as drafted; the LA branch runs as the iOS-only reach cell (iOS 17.2+ filter) with LA's own per-app setting treated as the consent surface, pending the ruling.

Provenance: agent-derived DRAFT. Source ids: DL-002 (decisions-baseline.json).

DL-034: T31 consent-forked CTA: PECR class ruling (ship-blocking)

• Journey(s): J1
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: Lisa + compliance
• Recommended decision date: 2026-07-22

Decision required: Get the PECR (Privacy and Electronic Communications Regulations) class ruling on T31's consent-forked CTA block: the reworked win receipt carries a consent-forked CTA whose lead job is the second purchase; a transactional-classed send doing marketing work is a PECR exposure, explicitly ship-blocking.

Context and supporting evidence: The clash register marks the win cluster RECONCILED on timing, but that reconciliation does not cover this classification ruling. The J1 pack restates it as one of Journey 1's two build prerequisites: a READY journey with two unanswered hard gates is only ready to SIGN, not to build. The GRW-16 email-leg ruling shares the same consent-fork logic and precedent.

Recommended option: Retain the consent fork so unconsented recipients get a purely transactional receipt (no marketing CTA), and have compliance confirm that shape keeps T31's transactional classification before ship. Book the compliance session now; chase to an answer at the Gate-1 review.

Rationale: The fork already exists in the design; the ruling needed is confirmation, not redesign, unless compliance disagrees.

Alternatives and trade-offs:

• Strip the CTA block entirely from T31 and move the second-purchase job wholly into the consented habit canvas

Dependencies:

• GRW-16 email-leg ruling (shares the same consent-fork logic)

Risks:

• PECR breach exposure; shipping without the ruling is prohibited by the gate

Consequence of delay: T31 cannot ship (the gate prohibits it); the win-receipt rework stalls, and the consent-fork precedent GRW-16 depends on (DL-045) stalls with it.

Working assumption used in the journey pack(s): The consent fork stands as designed: unconsented recipients get a purely transactional receipt with no marketing CTA; compliance confirms that shape keeps T31's transactional classification before ship.

Merge note (v1): Merged: baseline DL-034 + J1-D12 part (b) (restated build hard-gate). No conflict.

Provenance: agent-derived DRAFT. Source ids: DL-008 (decisions-baseline.json); J1-D12 (J1/decisions.json).

DL-035: ACT-24 C1: delete the BUN-card re-prompt; the ruled capture ladder governs unmodified (baseline, J1 and J5 agree)

• Journey(s): J1, J5
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: Sheran
• Recommended decision date: 2026-07-22

Decision required: Resolve the ACT-24 C1 contradiction: the BUN card's 're-prompt once on next session' line vs the ruled capture ladder (Q8, Sheran 6 Jul: 30-day decliner cooldown per surface, max 2 capture asks per 30 days, shared decline counter). DELETE the re-prompt so C1 obeys the ruled ladder.

Context and supporting evidence: Flagged as a hard gate (blocks the opt-in ladder build). Three sources reached the same answer: the baseline (delete), J1-D02 (delete; provenance argument: the ladder is Sheran-ruled, the re-prompt is agent-drafted; cost of deletion is small because a declining BUN buyer stays eligible for one sanctioned ask via a different surface, and the D+15 soft re-ask inside B3 is ladder-respecting), and J5-D-08 (resolve in favour of the cooldown as part of extending the ladder ladder-wide).

Recommended option: DELETE the re-prompt. If Sheran prefers to keep it, it must be recorded as an explicit dated per-surface exception on the row AND the C1 decline must still count in the shared ladder.

Rationale: Over-suppressing is benign, over-asking is the harmful failure (the recorded suppression principle); an exception forks the ladder state machine per surface for marginal capture gain and makes the decline counter ambiguous to build.

Alternatives and trade-offs:

• Record an explicit dated exception (workable but adds a per-surface state fork)
• Shorten the global cooldown (re-litigates a ruling without new evidence)

Dependencies:

• Blocks the opt-in ladder build until ruled

Risks:

• Slightly lower BUN capture in the 30d window; mitigated by the D+15 soft re-ask inside B3

Consequence of delay: The opt-in ladder build stays blocked; C1 either ships against the ruled ladder or not at all.

Working assumption used in the journey pack(s): The re-prompt is deleted so C1 obeys the ruled 30-day ladder; any keep requires an explicit dated per-surface exception AND the C1 decline still counting in the shared ladder.

Merge note (v1): Merged: baseline DL-035 + J1-D02. J5-D-08 (kept separate: the ladder-wide extension) also endorses DELETE. No conflict.

Provenance: agent-derived DRAFT (three sources concur; the ladder itself is Sheran-ruled 6 Jul). Source ids: DL-009 (decisions-baseline.json); J1-D02 (J1/decisions.json).

DL-045: GRW-16 email leg for unconsented users: suppress or shift

• Journey(s): J2
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: Lisa
• Recommended decision date: 2026-07-22

Decision required: Rule the GRW-16 email leg for unconsented users: suppress or shift.

Context and supporting evidence: The J2 pack now carries the working assumption in the build spec (decisioning matrix GRW-16 row, ladder, CRM build actions): not-email-opted users are suppressed on the email leg, never re-classed as transactional; dual-consented users KEEP the proven email leg alongside the new push (the proven family is never switched off on an assumption, and the one interaction design reads the overlap). The item stays open for Lisa's ruling under the DL-034 precedent.

Recommended option: Agent-derived DRAFT: suppress the email leg for non-opted users and let the push / in-app confirmation state carry the reassurance job, rather than re-classing marketing email as transactional.

Rationale: Reassurance is delivered by the channel ladder anyway, and re-classing would recreate the T31 PECR exposure on the proven send.

Alternatives and trade-offs:

• Shift the send to a transactional classification with reassurance-only content (needs the same compliance ruling as DL-034)

Dependencies:

• DL-034 compliance ruling sets the precedent

Risks:

• Wrong call either loses proven reach or creates PECR exposure

Consequence of delay: The consent fork on the day-before send cannot be built; the estate's only proven-incremental send risks either a PECR exposure or lost proven reach.

Working assumption used in the journey pack(s): Not-email-opted users are suppressed on the email leg, never re-classed as transactional; dual-consented users KEEP the proven email leg alongside the new push (the proven family is never switched off on an assumption), and the one interaction design reads the overlap. Carried in the J2 build spec (decisioning matrix GRW-16 row, ladder, CRM build actions); open for Lisa's ruling under the DL-034 precedent.

Changed in the council round:

• J2 council (20 Jul 2026) changed context. Why: The build needs a stated default to spec against; the assumption follows the channel ladder and avoids recreating the T31 PECR exposure on the estate's only proven-incremental send.

Provenance: agent-derived DRAFT. Source ids: DL-010 (decisions-baseline.json).

DL-069: Pin the returned_within_7d definition; re-state the return rate and re-size

• Journey(s): J4
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: Ivi (Analytics) + Data/Eng, Sheran ratifies the definition
• Recommended decision date: 2026-07-22

Decision required: Pin the operational definition of 'return' for the D+7 gate and re-state the organic ≤7d return rate on that pinned definition before the holdout is sized.

Context and supporting evidence: The signed ACT-07 walk justifies the 7-day gate with 74–79% organic return (RW definition: follow-up activity). Analytics' sizing (⑤) reads 37.1% first-time / 41.5% repeat / 57.8% power on their event definition, and sizes the eligible cohort at 102,832/yr, roughly 2.5x what the 74–79% frame implies. The clash is definitional, not a contradiction of the gate: organic returners re-monetise at near-winner levels (£105.42 vs £115.62, confirmed) either way.

Recommended option: Agent-proposed DRAFT: Data/Eng pin returned_within_7d to the ruled event set (auction-flow re-engagement: search, trip-add, bid, BUN; never bare app_open); Analytics re-states the return rate and re-sizes the D+7 audience and holdout on it. The gate's timing stands; only the sizing moves.

Rationale: The send volume, holdout power and expected £ all move materially with the definition; building on the unpinned number risks a 2.5x surprise at launch.

Alternatives and trade-offs:

• Ship on the RW definition and reconcile after launch (risks an unpowered or mis-sized holdout)

Dependencies:

• Eng event-set confirmation
• Analytics re-statement

Risks:

• Ship-blocking for the D+7 send: the journey's headline test cannot be sized without it

Consequence of delay: Ship-blocking for the D+7 send: the journey's headline test cannot be sized and the holdout waits.

Working assumption used in the journey pack(s): returned_within_7d pins to the ruled event set (auction-flow re-engagement: search, trip-add, bid, BUN; never bare app_open); the gate's timing stands, only the sizing moves once Analytics re-states the rate.

Provenance: agent-derived DRAFT. Source ids: J4-D02 (J4/decisions.json).

DL-011: Land BRZ-83: RFM power-user + persona model synced into Braze

• Journey(s): cross-journey
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: Ivi + Ben
• Recommended decision date: 2026-07-24

Decision required: Land BRZ-83: the RFM power-user + persona model synced into Braze.

Context and supporting evidence: Four or more moments' suppression and eligibility depend on this segment (GRW-01 A3, ACT-24 A3, ACT-07 power-suppress, GRW-16 persona registers). The recorded stance: sign-as-design is fine on the interim RW 6+/multi-product fallback, but no build until the sync lands, and it also gates the incremental read.

Recommended option: Proceed on the recorded stance: interim RW fallback for design sign-off, build gated on the sync, and the first RFM holdout must re-confirm the minus 11.8 percent over-send evidence.

Rationale: This is the already-recorded operating position across opendecisions, raid and governance; the open item is delivery and a dated commitment, not shape.

Alternatives and trade-offs:

• Build on the RW fallback permanently (loses the persona layer the journeys were designed around)

Dependencies:

• Ivi (model)
• Ben (eng sync)

Risks:

• Over-sending to power users is the one behaviour with confirmed negative holdout evidence

Consequence of delay: Design sign-off proceeds on the interim RW fallback only; every power-suppression and persona-tone dependency across the estate (suppression rules, wave grid, J6 rows) stays wired to a placeholder.

Working assumption used in the journey pack(s): Interim RW fallback for design sign-off; build gated on the sync; the first RFM holdout must re-confirm the minus 11.8 percent over-send evidence.

Provenance: source-ruled stance (recorded); open item is delivery + dated commitment. Source ids: DL-011 (decisions-baseline.json).

DL-012: journey_state exclusion mandated in every recurring batch audience

• Journey(s): cross-journey
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: CRM (Lisa) + Data
• Recommended decision date: 2026-07-24

Decision required: Mandate the journey_state exclusion segment in every recurring batch audience (M44_P, M45_P, M50-M53 lapse canvas) before GRW-01 and ACT-07 go live.

Context and supporting evidence: GRW-01 and ACT-07 carry journey_state dedupe (the canvas owns the customer exclusively), but the as-is batch sends do not read journey_state today, so the arbitration rule is unenforced at the batch layer. The register records this as the implementation ruling required.

Recommended option: Adopt as a launch precondition: no GRW-01/ACT-07 go-live until every recurring batch audience carries the journey_state exclusion; mirrors the J8-round one-lapse-owner ruling where broadcasts stand down for owned episodes.

Rationale: Without it the 30-canvas-day ownership rule is fiction and the weekly cap leaks.

Alternatives and trade-offs:

• Retire the batch sends outright (bigger call, revenue attached, not required for the boundary)

Dependencies:

• journey_state trait live in Braze (eng board rec96xM0KAyIUTBdK, NOW)

Risks:

• Batch and journey sends stacking in week one of go-live

Consequence of delay: Batch and journey sends stack in week one of go-live, violating the one-lapse-owner ruling in production.

Working assumption used in the journey pack(s): Adopted as a launch precondition: no GRW-01/ACT-07 go-live until every recurring batch audience carries the journey_state exclusion.

Provenance: agent-derived DRAFT. Source ids: DL-015 (decisions-baseline.json).

DL-013: Content-card impression/click instrumentation fix = launch-blocking

• Journey(s): cross-journey, J10, J8, J1, J5
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: Sheran (rule), Eng (fix), Analytics (verify)
• Recommended decision date: 2026-07-24

Decision required: Make the content-card impression/click instrumentation fix LAUNCH-BLOCKING for any home-personalisation test.

Context and supporting evidence: Also launch-blocking for Journey 1's not-armed content-card arc: paired with DL-036 (the placement build) as one build. The J1 pack now carries the pairing in its build gates, channel rules, Analytics build table (row 6) and Eng actions; Analytics verifies fixed events land in the warehouse before the J1 holdout starts.

Recommended option: Gate the recognition pilot on the fix: log it as the journey's only genuinely launch-blocking Engineering Requirement; Analytics verifies fixed events land in the warehouse before the holdout starts.

Rationale: An unreadable test is worse than no test: it burns the audience, contaminates the calendar, and produces an unfalsifiable claim on a journey class already at session-vanity risk (+108% sessions / -12% purchases precedent, confirmed).

Alternatives and trade-offs:

• Launch and read on downstream transactions only (loses the funnel diagnostics; cannot separate render failure from copy failure)

Dependencies:

• Eng sprint capacity (Engineering Requirements board; Eng briefed, not approver)

Risks:

• If the fix slips, the whole journey slips: say so at scoping rather than shipping blind

Consequence of delay: If the fix slips, every card-read journey slips with it: J1's not-armed arc, J5's T3/T4 reads and J10's recognition pilot are unreadable, and shipping unmeasured burns the majority cohort (about 56 percent of IW first-winners).

Working assumption used in the journey pack(s): The fix is logged as the only genuinely launch-blocking Engineering Requirement for home personalisation. J1 pairs it with the DL-036 placement as one launch-blocking build (Analytics verifies fixed events land in the warehouse before the J1 holdout starts); J5 carries it as launch-blocking for its T3/T4 reads and notes no ACT-12 CTR or opt-in CVR baseline can exist until it lands; J10's recognition pilot is gated on it.

Changed in the council round:

• J1 council (20 Jul 2026) changed context. Why: Card tracking registers ZERO today (confirmed); shipping the placement unmeasured burns the majority cohort (~56% of IW first-winners) and makes the journey's primary KPI claim unreadable for that arc.
• J5 council (20 Jul 2026) changed journeys. Why: J5's T3 onramp card and T4 home-card variants (ACT-12) read through card impressions/clicks, which register zero today (confirmed). The J5 pack now carries the fix as launch-blocking for its T3/T4 reads (Analytics brief build row 5) and its sizing ask notes that no ACT-12 CTR or opt-in CVR baseline can exist until the fix lands.

Provenance: agent-derived DRAFT. Source ids: D-J10-05 (J10/decisions.json).

DL-032: Record per-team verdicts on the J8 pack and declare Gate-1 lock

• Journey(s): J8
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: Sheran (gate) with per-team leads
• Recommended decision date: 2026-07-24

Decision required: Record per-team verdicts on the updated J8 pack and declare Gate-1 lock.

Context and supporting evidence: All 8 feedback records are answered (5 Applied, 1 Declined, 2 In discussion pending Analytics numbers). The ruled process is per-team sign-off after the changes round; verdicts are recorded in the feedback table's Overall verdict field. Two portal uploads (pack + hub zips) must land first so teams review the current pack.

Recommended option: Collect verdicts off the Monday PM session and the re-uploaded pack; lock once DL-029 (landing destinations) and DL-031 (D1 thresholds) close.

Rationale: Locking before the two P1 opens close would make the lock provisional and force a re-round.

Alternatives and trade-offs:

• Conditional lock with named carve-outs (mirrors the Q3 strategy conditional approve)

Dependencies:

• DL-029
• DL-030
• DL-031
• Portal re-uploads of pack + hub zips

Risks:

• Only Marketing/CRM fed back so far; Design, Product, Eng and Analytics verdicts are cold starts

Consequence of delay: The pack sits reviewed-but-unlocked and the 31 Jul ship date compresses against build QA.

Working assumption used in the journey pack(s): Verdicts collect off the Monday PM session and the re-uploaded pack; lock declares once DL-029 (landing destinations) and DL-031 (D1 thresholds) close; the LA cell holds separately on DL-030/DL-033.

Provenance: agent-derived DRAFT. Source ids: DL-005 (decisions-baseline.json).

DL-036: Not-armed content-card placement = launch-blocking for Journey 1

• Journey(s): J1
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: PMM/Product + Lisa (Sheran to sequence)
• Recommended decision date: 2026-07-24

Decision required: Treat the not-armed content-card placement as a LAUNCH-BLOCKING build for this journey, not a phase-2 nicety.

Context and supporting evidence: 56% of IW first-winners are not push-opted at the win (confirmed). If capture misses, the not-armed arc's ONLY habit-window channel is the content-card placement approved 7 Jul (NEXT, via dynamic tiles) but not built. Until it ships, the majority cohort's habit window is silence, the exact AS-IS defect the journey exists to fix.

Recommended option: Sequence the placement build with the canvas build (PMM/Product + Lisa); if it cannot land for launch, launch with an explicit 'armed-only phase 1' label and report reach honestly (the pack must not imply the not-armed arc is live).

Rationale: Reach honesty: a journey whose headline metric covers all first-winners cannot silently exclude the majority consent state.

Alternatives and trade-offs:

• Email-only fallback for not-armed users (invalid: not-armed includes non-consented; PECR); do nothing (rejected).

Dependencies:

• PMM/Product placement build; Lisa Braze wiring; dynamic-tiles surface decision (shared with Journey 10); content-card impression/click instrumentation fix (DL-013, Eng): the arc is unmeasurable without it, so the fix ships with the placement as one launch-blocking pair.

Risks:

• Build-capacity contention with Journey 2's trip-state surface.

Consequence of delay: The not-armed arc (the majority cohort, about 56 percent of IW first-winners) has no reach surface at launch; shipped unmeasured it burns the cohort (paired with DL-013).

Working assumption used in the journey pack(s): The placement is sequenced with the canvas build as launch-blocking; if it cannot land for launch, the journey launches with an explicit 'armed-only phase 1' label and reach is reported honestly.

Changed in the council round:

• J1 council (20 Jul 2026) changed dependencies. Why: A reach fix without its measurement fix repeats the estate's unreadable-send defect on day one.

Provenance: agent-derived DRAFT. Source ids: J1-D06 (J1/decisions.json).

DL-046: Journey 2 headline test (GRW-16 reframe A/B) + one interaction design

• Journey(s): J2
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: Sheran (headline pick) + Ivi (design)
• Recommended decision date: 2026-07-24

Decision required: Confirm the journey's ONE headline test (GRW-16 reassurance-reframe A/B) and commission the single interaction design across the three cells that can hit one trip (GRW-16 A/B, GRW-18 holdout, phase-2 GRW-17 holdout).

Context and supporting evidence: Ruling 8 allows one headline test per journey release. J2 carries a Sheran-ruled GRW-18 test cell (9 Jul override) PLUS the GRW-16 test-then-replace A/B protecting the estate's only proven-incremental family. An early winner sits in both cells on the same trip; uncoordinated randomisation contaminates both reads, and the GRW-16 family's existing no-email holdout adds a third constraint.

Recommended option: Headline = the GRW-16 reframe A/B with non-inferiority guard (it defends the +16.9%/+16.3% family, the highest-stakes read in the journey). GRW-18's holdout stands as ruled but is explicitly NOT the headline: its read gates only the mid-window beat. Analytics owns one interaction design: independent randomisation, stratified reads, no overlap with the existing no-email holdout on the same trip.

Rationale: The proven family is the one asset in the estate that can be destroyed by a badly-designed test; everything else in the journey is greenfield and can wait for clean reads.

Alternatives and trade-offs:

• Make GRW-18 the headline (smaller audience, lower stakes, wastes the release's one headline on the tail)
• Run all cells uncoordinated (contaminated reads)

Dependencies:

• DL-054 (rollback floor X + M31 offset)
• GRW-18 sizing (next item)

Risks:

• Wiring tests without the interaction design breaks the holdout-commit ruling; a contaminated GRW-16 read could roll back the proven send on noise

Consequence of delay: Wiring tests without the one interaction design breaks the holdout-commit ruling; a contaminated GRW-16 read could roll back the proven send on noise.

Working assumption used in the journey pack(s): Headline = the GRW-16 reassurance-reframe A/B with non-inferiority guard; GRW-18's holdout stands but is explicitly not the headline; Analytics owns one interaction design (independent randomisation, stratified reads, no overlap with the existing no-email holdout).

Provenance: agent-derived DRAFT. Source ids: J2-#2 (J2/decisions.json).

DL-061: Ratify the J3 phase split: no-new-event moments now, dwell-gated briefed BLOCKED

• Journey(s): J3
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: Sheran (ratify), Sophie/Eng (sprint slot)
• Recommended decision date: 2026-07-24

Decision required: Ratify the phase split: Phase 1 ships the no-new-event moments now; the dwell/abandon-gated moments are briefed BLOCKED, not buildable, on the J3 Eng event package.

Context and supporting evidence: Half the journey's triggers do not exist (confirmed taxonomy audit): no dwell events, no confirm_abandon, no lead_state, no queryable push consent, no dead-end reason split. Gate-1 ruling 11 requires blocked-on-data moments briefed as blocked.

Recommended option: Phase 1 = ACT-08 static reword (operator split already ruled), ACT-25 all-copy redirect, ACT-26 permission-split card on the existing C1 mechanic, ACT-13 copy branch (signed), consent write-back. Blocked = ACT-02, ACT-22, ACT-05 in-moment arm, the spine's trigger feed, the reason-split copy.

Rationale: Ships real value this quarter without pretending the trigger layer exists; keeps the Eng ask honest and prioritised.

Alternatives and trade-offs:

• Hold the whole journey for the event package (loses the cheap copy wins)
• Proxy dwell with session-time heuristics (unvalidated, risks firing on readers not hesitators)

Dependencies:

• Eng event package logged as Engineering Requirements
• Analytics dwell-threshold definitions precede the dwell event builds

Risks:

• Phase 1 alone will not move step-recovery CVR much; expectation-set that the headline test waits for the dwell events

Consequence of delay: Buildable no-new-event moments wait on events that do not exist, or dwell-gated moments get built against missing events.

Working assumption used in the journey pack(s): Phase 1 = the no-new-event moments (ACT-08 static reword, ACT-25 redirect, ACT-26 permission-split card, ACT-13 copy branch, consent write-back); the dwell/abandon-gated moments are briefed BLOCKED, not buildable, on the J3 Eng event package.

Provenance: agent-derived DRAFT. Source ids: J3-#3 (J3/decisions.json).

DL-079: Journey 5 audience boundary: lifetime_transactions = 0; 31.8k onramp dual-listed with J1 B3

• Journey(s): J5, J1
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: Sheran (with Journey 1 pack author)
• Recommended decision date: 2026-07-24

Decision required: Ratify the exact audience boundary: Journey 5 owns lifetime_transactions = 0 and exits on the first successful transaction event; the 31.8k transacted-never-bid onramp cohort (ACT-19 Rows 5 and 7) is dual-listed with Journey 1's GRW-01 B3 under the shared win_moment_canvas_active key.

Context and supporting evidence: The journey's headline confirmed volume (31.8k onramp, analytics-sizing ⑲) describes users who have ALREADY transacted (BUN/tickets, never bid), so it is not the never-transacted pool. GRW-01 B3 (practice-bid re-onboarding) already claims the same cohort post-purchase, and ACT-24's exit feeds it.

Recommended option: Boundary = first successful transaction event (auction_won first / instant_win / auto_win_on_open / BUN confirm), not first bid and not journey completion. Rows 5/7 treatments stay authored under ACT-19 (moment home) but run as the in-app surface of Journey 1's B3 canvas; one coverage, one journey_state owner, no second build.

Rationale: Prevents double-coverage of one cohort by two journeys and keeps the pack's volume claims honest: the truly never-transacted addressable volume is in-session flow plus the wave tiers, not the 31.8k.

Alternatives and trade-offs:

• Keep the 31.8k inside J5 and carve the cohort out of GRW-01 B3
• Move Rows 5/7 wholesale to the Journey 1 pack

Dependencies:

• win_moment_canvas_active key confirmation (Lisa/CRM)
• instant_win emits auction_won audit (Eng/PMM, in the consolidated log)

Risks:

• Unruled, the same user can sit in two journeys' practice arcs and receive duplicate onramp nurture

Consequence of delay: The same user can sit in two journeys' practice arcs and receive duplicate onramp nurture.

Working assumption used in the journey pack(s): Boundary = first successful transaction event (auction_won first / instant_win / auto_win_on_open / BUN confirm); the 31.8k Rows 5/7 treatments stay authored under ACT-19 but run as the in-app surface of Journey 1's B3 canvas: one coverage, one journey_state owner, no second build.

Provenance: agent-derived DRAFT. Source ids: J5-D-01 (J5/decisions.json).

DL-080: ACT-15: create the Airtable master row and Gate-1 card, scoped as governance (closes roster decision c)

• Journey(s): J5
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: Sheran + CRM (Airtable write); Product co-approve
• Recommended decision date: 2026-07-24

Decision required: Close roster decision (c): create the ACT-15 (native push opt-in prompt) Airtable master row (canonical MoT code) and Gate-1 card; until then the Journey 5 pack signs 4 of 5 moments.

Context and supporting evidence: ACT-15 exists only as a named master in mots.json: no code row in crosswalk.json / mots_id_update.json, no card in cards_v4; gate1-sequence nonetheless lists it in the Journey 5 gate. Left uncarded, the journey's Gate-1 status is permanently BUILD and the capture-ladder decline counter has an undefined rung.

Recommended option: Write the master row + card scoped as GOVERNANCE of the native OS prompt (value-led pre-prompt after a first value moment, never cold launch, declines honoured). Build priority P2 within the pack: the real capture gap is the marketing-subscription layer (AN-3: OS 93 percent vs marketing 68 percent, confirmed), which ACT-13/24/12 own at peak trust. The row-and-card creation itself is P1 because it gates Journey 5's Gate-1 completeness.

Rationale: The pack cannot complete Gate-1 sign-off with a code-less moment; carding it as governance avoids inflating a Product-owned system prompt into a CRM build.

Alternatives and trade-offs:

• Drop ACT-15 from the Journey 5 roster and fold its rule into the capture-ladder contract
• Card it as P1 alongside the onramp

Dependencies:

• Airtable MoTs master write (fldKdzFompu8iMkyr)
• Product co-approval (native surface)

Risks:

• Left uncarded, Gate-1 stays BUILD and the decline counter has an undefined rung

Consequence of delay: Gate-1 stays BUILD (the Journey 5 pack signs 4 of 5 moments) and the capture ladder's decline counter has an undefined rung.

Working assumption used in the journey pack(s): The master row and card are drafted scoped as GOVERNANCE of the native OS prompt (value-led pre-prompt after a first value moment, never cold launch, declines honoured); build priority P2 within the pack, the row-and-card creation itself P1.

Merge note (v1): Merged: roster decision (c) + J5-D-02 (which answers it). No conflict.

Provenance: agent-derived DRAFT. Source ids: ROSTER-c (ROSTER.md section 4 (c)); J5-D-02 (J5/decisions.json).

DL-081: One celebration per win event: Beginner's Luck as a copy branch, not a third modal (mechanic strike flagged)

• Journey(s): J5, J1
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: Sheran; Product co-approve (in-app surface)
• Recommended decision date: 2026-07-24

Decision required: One celebration per win event: extend the ACT-19 Row 9 handoff logic to Row 8, making Beginner's Luck the never-won COPY BRANCH of the win-moment owner's surface (ACT-24 for instant_win / auto_win_on_open / BUN; GRW-01 A1 for close-window wins), not a third modal.

Context and supporting evidence: As drafted, ACT-19 Row 8 (dedicated first-win modal on auction_won) and GRW-01 A1 (win-moment celebration on auction_won) can both fire for a never-won close-window winner: two celebrations and two opt-in asks at one event, the same defect the ACT-24 adversarial pass fixed for IW.

Recommended option: ACT-19 contributes the Beginner's-Luck eligibility flags + copy branch (repeatable framing, next-trip seed, no discount) to the owning surface. This is a MECHANIC-OWNERSHIP change vs the Gate-1 card's 'NEW dedicated modal' intent, flagged explicitly for Sheran to strike.

Rationale: Preserves the Gate-1 KEEP (every eligible first win celebrated) while enforcing the STACK ban and the single-ask contract at the highest-trust moment in the lifecycle.

Alternatives and trade-offs:

• Keep Row 8 as a dedicated modal and suppress GRW-01 A1's celebration for never-won winners (two builds, one suppressed)

Dependencies:

• ~70% guarantee-fires rule + eligibility (Analytics, Ivi/Ben)
• GRW-01 / Journey 1 pack mirror edit

Risks:

• Double celebration + double ask at the doubt-to-loyalty switch; opt-in completion depressed

Consequence of delay: Double celebration plus double ask at the doubt-to-loyalty switch; opt-in completion depressed.

Working assumption used in the journey pack(s): Beginner's Luck runs as the never-won COPY BRANCH of the win-moment owner's surface (ACT-24 or GRW-01 A1), not a third modal; the mechanic-ownership change vs the card's dedicated-modal intent is flagged explicitly for Sheran to strike.

Provenance: agent-derived DRAFT. Source ids: J5-D-05 (J5/decisions.json).

DL-082: Beginner's-Luck substantiation rule + named compliance owner for claims

• Journey(s): J5, J9
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: Sheran to name (Marketing/legal)
• Recommended decision date: 2026-07-24

Decision required: Ratify the Beginner's-Luck substantiation rule and name a compliance owner for win-rate and price-comparison claims.

Context and supporting evidence: 68.6% (confirmed) is the first-auction win rate among users who complete a first auction: a self-selected denominator. Presented as a personal prediction to the cold pool it is unsubstantiated; 'guaranteed' is false for the ~30% of wins the guarantee mechanic does not fire for; fixed-threshold naming is banned while the 50% dynamic-threshold test runs. No compliance owner exists today (ASA/CAP exposure).

Recommended option: Standing rule for every ACT-19/ACT-31/ACT-12 asset: transparent population stat only ('7 in 10 first-time bidders win their first auction'), never 'you'll probably win', 'guaranteed' only on the verified guarantee path, comparison anchors substantiated at fire-time (never 'cheaper', never 'save up to'). Sheran names the claims sign-off owner (Marketing/legal). The same owner's remit also covers the consent-incentive framing (the £5 opt-in lead on ACT-12) and the PECR classification of marketing-layer capture copy, inheriting the Journey 1 T31 precedent (DL-034; unbundling review DL-038).

Rationale: The onramp mechanic depends on this stat; one unsubstantiated headline puts the whole activation narrative at regulatory risk.

Alternatives and trade-offs:

• Route all claims through Gate-2 creative review without a named owner (status quo)

Dependencies:

• Guarantee-fires eligibility confirmation (Analytics)

Risks:

• ASA challenge; forced creative withdrawal mid-quarter

Consequence of delay: Claims ship without a sign-off owner: ASA challenge risk and forced creative withdrawal mid-quarter, with wave-1 creative (J9) inheriting the same gap.

Working assumption used in the journey pack(s): The substantiation rule is applied across every ACT-19/ACT-31/ACT-12 asset as drafted (population stat only, 'guaranteed' only on the verified path, anchors substantiated at fire-time); the named owner's remit also covers the GBP 5 consent-incentive framing and the PECR class of capture copy per the T31 precedent (DL-034/DL-038). Open for Sheran to name the owner.

Changed in the council round:

• J5 council (20 Jul 2026) changed recommendation. Why: Legal/Privacy/Risk lens: the win-rate claim, the price anchors, the incentivised consent ask and the PECR class of capture copy are one claims-governance job; naming one owner for the first two while leaving the second two unowned recreates the gap the item exists to close.

Provenance: agent-derived DRAFT. Source ids: J5-D-06 (J5/decisions.json).

DL-088: Ratify the Journey 6 moment set and single post-trip spine

• Journey(s): J6
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: Sheran
• Recommended decision date: 2026-07-24

Decision required: Ratify the Journey 6 moment set (GRW-04/09/28/29) and the single post-trip spine that carries them (T+90min value-proof push, ONE arbitrated T+24h email, one cadence nudge, one episode per trip).

Context and supporting evidence: All four moments are unauthored Batch-2 moments; the cards, read independently, would fire four-plus sends into the same 48h window alongside M39_E and Journey 1's canvas. This is a design-stage pack: the spine is the pack.

Recommended option: Keep all four moments; ratify the spine and the one-email-at-T+24h rule as the journey's shape.

Rationale: Post-trip is peak trust (81.4% positive, confirmed) but the window is a single slot in the customer's attention; four independent walks would burn it. The spine preserves each moment's job while enforcing one-job-per-message.

Alternatives and trade-offs:

• Author the four moments as independent walks with per-pair dedupe rules (more sends, more arbitration debt)
• Cut GRW-28/29 from Q3 and ship reflection + re-book only

Dependencies:

• JD6-04 positive_trip signal
• JD6-05 slot order
• JD6-06 M39_E disposition

Risks:

• Unratified, the four cards ship as-written and stack sends in the post-trip window

Consequence of delay: Unratified, the four cards ship as-written and stack sends in the post-trip window.

Working assumption used in the journey pack(s): The pack is authored to the single spine: T+90min value-proof push, ONE arbitrated T+24h email, one cadence nudge, one episode per trip.

Provenance: agent-derived DRAFT. Source ids: JD6-01 (J6/decisions.json).

DL-090: Re-gate GRW-28 to positive trip AND 2nd/3rd transaction

• Journey(s): J6
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: Sheran (confirm the re-gate matches the rule's intent)
• Recommended decision date: 2026-07-24

Decision required: Re-gate GRW-28 from '2nd_win OR post-positive-trip' to 'positive trip AND 2nd/3rd transaction'.

Context and supporting evidence: The card and the mots trigger use OR, which lets a single-win customer get the £5 referral ask off one good trip. Sheran's advocacy sequencing rule (standing, non-negotiable) requires the AND.

Recommended option: AND gate; cut the card's single-win arm from release 1. Its no-incentive 'soft share' variant may return as a phase-2 test only on an explicit Sheran relaxation.

Rationale: Rule compliance plus the evidence: trust off one trip is fresh but shallow (the card's own words); referred-in customers are worth ~6x organic (confirmed, mention-me CLV), so protecting the ask's timing protects the channel.

Alternatives and trade-offs:

• Keep the OR and treat the single-win arm as the softer copy variant (violates the standing rule)

Dependencies:

• D4 transaction_count gate available at fire time

Risks:

• Volume: the AND gate shrinks an already small pool; guardrail total referral volume vs the recapture (Analytics)

Consequence of delay: The single-win referral arm ships against the advocacy sequencing rule.

Working assumption used in the journey pack(s): The AND gate ('positive trip AND 2nd/3rd transaction') is adopted in the pack; the single-win arm is cut from release 1; the no-incentive soft-share variant returns only as a phase-2 test on an explicit Sheran relaxation.

Provenance: source-ruled (standing advocacy sequencing rule); open work is the re-gate. Source ids: JD6-03 (J6/decisions.json).

DL-091: Sign the operational definition of positive_trip (launch-blocking)

• Journey(s): J6
• Priority and urgency: P1
• Status: open
• Recommended decision-maker: Analytics (Ivi) definition + Eng/Data build
• Recommended decision date: 2026-07-24

Decision required: Sign the operational definition of positive_trip and confirm its inputs are queryable at fire time.

Context and supporting evidence: Every J6 leg gates on it and no such signal exists. Proposed DRAFT: trip_completed AND no disruption flag AND no declassification AND no open Delay Repay / refund / dispute on the trip; phase 2 adds negative-survey suppression.

Recommended option: Adopt the DRAFT definition as v1; Eng exposes the flags at fire time; Ivi signs the definition; treat as launch-blocking for the whole journey.

Rationale: Disruption is joint-#1 CX pain (64.2% neg, 12,359 tickets, confirmed) and Delay Repay is 80.4% negative; celebrating or asking after a disrupted trip converts goodwill to insult.

Alternatives and trade-offs:

• Ship on trip_completed alone with disruption suppression only (leaves refund/dispute holes)
• Wait for survey-based sentiment (delays the whole journey behind the P0 link fix)

Dependencies:

• Eng flag exposure
• Zendesk/Delay-Repay linkage to trips

Risks:

• A false-positive 'positive' trip is the journey's worst failure mode; err toward suppression

Consequence of delay: Launch-blocking for the whole journey: no J6 send can safely fire, and a false-positive 'positive' trip is the journey's worst failure mode.

Working assumption used in the journey pack(s): The DRAFT definition is treated as v1, erring toward suppression; Eng exposes the flags at fire time; Ivi signs the definition; the journey's launch waits on it.

Provenance: agent-derived DRAFT. Source ids: JD6-04 (J6/decisions.json).

DL-121: Interim entry anchor for Journey 6: schedule-derived arrival stands in for the missing trip_completed event

• Journey(s): J6
• Priority and urgency: P2
• Status: adopted as working assumption (council); Eng validates; no Gate-1 ruling sought, Sheran informed
• Recommended decision-maker: Eng/Data feasibility + Analytics (Ivi) definition; Sheran informed
• Recommended decision date: 2026-07-20

Decision required: Adopt the interim entry anchor for the journey: a schedule-derived arrival timestamp (scheduled arrival from timetable data) stands in for the trip_completed event, which does not exist (dbt R4 gap, confirmed).

Context and supporting evidence: The whole spine enters on trip_completed and the pack's first draft presented it as available. J1's interim proxy (scheduled departure + 1d, J1-D08) cannot time a T+90min reflection push. GRW-16/17 already anchor to scheduled times, so the arrival-side derivation is the same anchor family. Surfaced by the council as data-dependency pin D11.

Recommended option: Working assumption adopted: schedule-derived arrival + 90min times the reflection push; Eng confirms feasibility at the validation step. If it cannot be derived, the reflection push waits for the real event and the journey launches with the mirror, the T+24h slot and the nudge on the J1-D08 proxy. Swap to the real trip_completed event when it lands; brief it as one build with J1's travel-completion floor.

Rationale: Without an anchor the journey is unbuildable; without stating the gap the pack overstates its readiness. The schedule-derived anchor is the cheapest credible interim and reuses an existing anchor family (recommendation-framing rule).

Alternatives and trade-offs:

• Wait for the real event (delays the whole journey behind an unscheduled eng build)
• Use J1's departure+1d proxy for everything (loses the T+90min moment entirely)

Customer and business implications: Affects every J6 entrant. Keeps the pack's readiness claim honest and gives Eng a single briefed build shared with J1's travel-completion floor.

Dependencies:

• Eng feasibility of the arrival-time derivation
• Delay data caveat: a schedule-derived anchor mistimes delayed trips, which the positive_trip gate (D1) largely suppresses anyway

Risks:

• A late-running clean trip gets the push slightly early; benign for a value proof, and D1 suppresses the disrupted cases

Consequence of delay: No entry anchor means no J6 send can be timed and the journey stays unbuildable; if the feasibility check slips, the T+90min reflection push is at risk of waiting behind an unscheduled eng build for the real event.

Working assumption used in the journey pack(s): Schedule-derived arrival + 90min times the reflection push until the real trip_completed event lands; fallback launches without the T+90min push (mirror, T+24h slot and nudge on the J1-D08 proxy).

Provenance: agent-derived DRAFT (council). Source ids: JD6-16 (J6/decisions.json); J6/decisions-delta.json (new, council round 20 Jul 2026).

DL-033: Braze answer on Live Activity push-to-start token holder

• Journey(s): J8
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Lisa (with the Braze AM)
• Recommended decision date: 2026-07-24

Decision required: Get the Braze answer on who holds the Live Activity push-to-start tokens (and close the SDK 8.2.0+ / iOS 17.2+ contract check).

Context and supporting evidence: Open question from CRM (15 Jul): does Braze hold our registerPushToStart tokens, or is has_ios_live_activities_start_token a segmentation-only flag while our backend keeps the real token? If the latter, Braze cannot send the LA start at all yet. The contract check with the Braze AM was running as of 15 Jul.

Recommended option: Agent-derived DRAFT: if Braze does not hold start tokens, point the webhook at our own backend (confirmed working today) and revisit the Braze-native route later.

Rationale: The backend route removes the dependency without changing the canvas-owned eligibility/holdout/caps shape.

Alternatives and trade-offs:

• Wait for Braze-native support before shipping the LA cell

Dependencies:

• Braze AM response
• Eng confirmation of backend token store

Risks:

• Answer arriving after Gate-1 lock leaves the LA cell's build path unresolved

Consequence of delay: An answer arriving after Gate-1 lock leaves the LA cell's build path unresolved.

Working assumption used in the journey pack(s): If Braze does not hold start tokens, the webhook points at our own backend (confirmed working today); the Braze-native route is revisited later.

Provenance: agent-derived DRAFT. Source ids: DL-003 (decisions-baseline.json).

DL-101: Journey 7 entry contract: day-60 journey_state hand-off + metered stock backfill

• Journey(s): J7
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Sheran (ratify) with Analytics + CRM
• Recommended decision date: 2026-07-31

Decision required: Ratify the Journey 7 entry contract: live entry = Journey 8's day-60 journey_state hand-off only; stock entry = one-time propensity-ranked metered backfill.

Context and supporting evidence: Sheran ruled the window model 20 Jul (J8 owns 0-60d, J7 owns 60d+, journey_state hands over at day 60 with a VIP-priority flag to REC-03). The existing dormant stock (~600k, TBC: Analytics) never passed through J8 and needs a separate one-time door.

Recommended option: REC-03 entry = the journey_state transition event, never an independent day-count (one clock, no gap, no overlap). Backfill the stock once via persona-keyed lapsed_threshold_crossed, released in propensity-ranked waves, highest band first; steady state is hand-off-only.

Rationale: Two independent clocks (J8's episode clock and a J7 day-count) would drift and double-message; a single journey_state write is the only gapless, overlap-free contract. Metering the backfill protects deliverability and lets the propensity model prove itself on the warmest bands first.

Alternatives and trade-offs:

• J7 recomputes lapse independently from recency (risks gap/overlap with J8)
• No backfill: only new hand-offs enter (strands the entire existing pool for 60+ days)

Dependencies:

• D2 churn-score family landing
• journey_state event consumable as a Braze trigger (with J8 build)

Risks:

• If the hand-off event is missed or delayed, users fall into an unowned void; needs a daily reconciliation check

Consequence of delay: If the hand-off event is missed or delayed, lapsed users fall into an unowned void; without the metered backfill rule, the stock gets blasted in one wave.

Working assumption used in the journey pack(s): REC-03 entry = the journey_state transition event only, never an independent day-count (one clock, no gap, no overlap); the stock backfills once via propensity-ranked metered waves, highest band first; steady state is hand-off-only; a daily reconciliation check covers the missed-event risk.

Provenance: source-ruled (window ruled 20 Jul) + agent-derived DRAFT (backfill design). Source ids: DL-J7-01 (J7/decisions.json).

DL-122: Email consent-age floor for win-back's deep tail (24-month working floor; Legal read commissioned)

• Journey(s): J7
• Priority and urgency: P2 (P2; needed before the backfill's deep-tail waves are designed, not before the day-60 hand-off sends (those customers are 60d dormant, inside any plausible floor).)
• Status: open
• Recommended decision-maker: Sheran commissions Legal; Legal rules the floor; Analytics sizes the split
• Recommended decision date: 2026-08-07

Decision required: Rule the email consent-age floor for win-back's deep tail (PECR / consent-decay posture), and commission the Legal read behind it.

Context and supporting evidence: The one-time backfill reaches customers dormant for years (296.3k churned >240d, confirmed cards_v4; 1,592,082 Churned, confirmed Contactability). Email marketing to them rests on consent or the PECR soft opt-in taken at purchase (PECR = the UK Privacy and Electronic Communications Regulations), and the quality of both decays with dormancy. No Legal read exists anywhere in the repo; the pack previously required only 'consented on at least one channel' with no age test. Push permission is a live, device-level, revocable signal, so it is the lower-risk door on the deep tail, and it is also the confirmed strongest reactivation channel (1.12%).

Recommended option: Adopt a 24-month floor as the working assumption: no outbound email to customers with no engagement and no consent refresh in the last 24 months; push carries the deep tail where held; opt-out honoured immediately; the REC-14 hard end-state and the suppression-list deliverable stand as the programme's consent hygiene. Sheran commissions a Legal read on the floor before the backfill's deep tail is emailed; Analytics sizes how much of the pool the floor removes.

Rationale: Emailing a years-dormant list on decayed permission risks complaints and deliverability damage to the whole estate that no reactivation yield justifies, and a stated floor keeps the sendable-pool arithmetic honest.

Alternatives and trade-offs:

• No floor: email the full consented tail (maximises reach; carries the PECR and deliverability risk unpriced)
• Re-permission campaign first (cleanest legally; costs a quarter and spends the one first-touch on an admin ask)

Customer and business implications: Long-dormant customers are not contacted on stale permission (respect posture, matches the journey's CX thesis); the backfill's email reach shrinks by a sized, known amount; the estate's deliverability and sender reputation are protected for every other journey.

Dependencies:

• Legal read (owner to be named by Sheran); DL-009 pool reconciliation (the floor changes the sendable denominator); risk: the floor is guessed too tight or too loose until Legal rules, which is why it is labelled a working assumption.

Consequence of delay: Either the deep tail is emailed on unexamined permission (regulatory and deliverability exposure) or the backfill silently stalls at the tail with no stated reason.

Working assumption used in the journey pack(s): 24-month engagement/consent-refresh floor on outbound email; push carries the deep tail where held; stated in strategy.md section 3.1 and the pack's channel rules and decisioning matrix as DL-J7-13.

Provenance: agent-derived DRAFT (council). Source ids: J7/decisions-delta.json (new, council round 20 Jul 2026).

DL-120: Day-14 episode close for the loss-recovery journey (replaces the 72h timeout Journey 1 assumed)

• Journey(s): J4, J1
• Priority and urgency: P2 (P2; needed before the J4 canvas build starts (the exit logic is built into the canvas).)
• Status: open
• Recommended decision-maker: Sheran, with Lisa (CRM) confirming the canvas mechanics
• Recommended decision date: At the J4 Gate-1 review

Decision required: Ratify the day-14 episode close for the loss-recovery journey: the J4 episode (bid_placed to exhaust) closes at day 14 post-close, replacing the 72h post-close timeout Journey 1's architecture assumed, with lapse handling handed to the lifecycle owner and never a re-fired loss message.

Context and supporting evidence: The D+7 recovery send needs a 7-day non-return gate plus a post-send observation window before hand-back, so a 72h episode close is arithmetically impossible for this journey. The signed ACT-07 row specifies one send then drop to organic; the day-14 close is the earliest boundary that contains the gate, the send and the silent-exit check. Confirmed inputs: the 102,832/yr eligible cohort (Analytics sizing tracker) and the one-send-then-organic ruling in the signed row.

Recommended option: Day-14 close: first-time (never-transacted) losers exit to Journey 5's pool with the first-timer flags (first_loss, close_reason, route; DL-073); transacted losers exit to the lifecycle bands under journey_state.

Rationale: It is the minimum window that fits the ruled D+7 mechanics, and it is logged as its own decision because it moves a Journey-1 boundary rather than being J4-internal.

Alternatives and trade-offs:

• Keep the 72h timeout (breaks the D+7 gate: the episode would close before the journey's one send); day-21 or longer (delays the J5 activation hand-off for the 81%-quit cohort with no added signal).

Customer and business implications: Every losing bidder gets exactly one recovery window and then a clean hand-off; no loser sits in limbo between J4 and the lifecycle bands, and no lapse timer re-fires a loss message.

Dependencies:

• journey_state (one owner per customer, DL-012 batch exclusions); DL-073 ratification for the J5 leg. Risk if unruled: J1 and J4 encode different episode boundaries and the hand-off double-fires or gaps.

Consequence of delay: The canvas cannot encode its exit; either the build waits or it ships with J1's 72h boundary and strands every D+7-eligible loser before the journey's one send fires.

Working assumption used in the journey pack(s): The pack encodes the day-14 close as the proposed boundary (flagged TBC Sheran/Lisa on the brief, logic and map tabs) and designs the J5/lifecycle split on it.

Provenance: agent-derived DRAFT (council). Source ids: J4/decisions-delta.json (new, council round 20 Jul 2026).

DL-123: Opt-in banner moment: give it an owner and a matrix row before J8 build QA

• Journey(s): J8, cross-journey
• Priority and urgency: P2 (P2 (does not gate the sendable states; gates the no-channel cell build))
• Status: open
• Recommended decision-maker: Sheran with CRM (Lisa)
• Recommended decision date: before J8 Braze build starts (post Gate-1 lock)

Decision required: Give the opt-in banner moment an owner and a matrix row before J8 build QA: the pack assigns it to 'the channel-infrastructure journey', which is not one of the canonical 10 and has no authored moment or matrix row.

Context and supporting evidence: The J8 no-channel arc (11,327 actives, 14.6% of the active band, confirmed 14 Jul with caveats) has the mid-Home opt-in banner as its ONLY message; the pack's map and matrix both flag 'matrix row to create' and 'owned by the channel-infrastructure journey'. The exposure ladder (two varied asks, rest, re-arm) is designed but unowned. Related but distinct: DL-080 covers ACT-15, J5's native push opt-in prompt.

Recommended option: Create the Airtable master row and matrix row now, owned by CRM as channel infrastructure (same governance scoping as DL-080's ACT-15), consumed by J8 and any journey with a no-channel cell; decide at the same sitting whether channel-infrastructure moments live outside the 10 by design (roster decision d pattern).

Rationale: J8's no-channel treatment is unbuildable and unmeasurable without an owning record; the ladder rules (session caps, re-arm triggers) need one home before three more packs reference them.

Alternatives and trade-offs:

• Fold it into J10's home framework (the banner renders in the mid-Home slot J10 governs): cleaner surface ownership but couples J8's launch to a DEFINE-stage journey.

Customer and business implications: Until owned, 14.6% of the slipping active band has no designed path back to contactability.

Dependencies:

• DL-116 (Home slot map arbitration) decides where the banner may render; DL-013 instrumentation applies if a card mechanic is ever swapped in.

Consequence of delay: The no-channel arc ships as designed-but-unowned; the banner's caps and re-arm rules drift per journey.

Working assumption used in the journey pack(s): Banner exists, is Braze IAM in the mid-Home slot, two varied asks then rest; ownership 'channel-infrastructure journey' named but not instantiated.

Provenance: agent-derived DRAFT (council). Source ids: J8/decisions-delta.json (new, council round 20 Jul 2026).

---

Band 3: design-shaping

DL-014: REC-21 (welcome-back) ownership: J7, J9 and J10 all recommend Journey 7 owns it (AGREEMENT; closes roster decision b)

• Journey(s): cross-journey, J7, J9, J10
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Sheran (roster ruling)
• Recommended decision date: 2026-07-24

Decision required: Assign REC-21 (welcome-back on app re-open; carded, currently in no journey) for Q3: Journey 7 owns the moment, its measurement and its suppression/dedupe against the win-back email+push twin; Journey 10's home framework renders it as one state of the S1 recognition slot; Journey 9 registers as a feeder only.

Context and supporting evidence: Three packs answered roster decision (b) independently and AGREE on J7 ownership. J7 (DL-J7-03): REC-21 is the receiving surface of every J7 send and J9 wave, and its conversion is part of the win-back double-read; contract: when Journey 10 scopes home re-entry, the recognised-state/home-personalisation job transfers, J7 keeps the 61d+ dormant-returner branch. J9 (J9-D11): the returner a wave produces is a reactivating lapsed/churned customer, which is J7's transition; leaving the landing beat unowned wastes the engine's output. J10 (D-J10-03): the one-lapse-owner rule makes welcome-back the win-back episode's landing beat; split owner (J7) from renderer (J10). The verified AS-IS defect (scheduled reactivation sends keep firing after the user has returned) can only be fixed by the journey that owns those sends.

Recommended option: Ratify J7 ownership with the owner/renderer contract and the Q3 transfer clause. An easy ratification: the three positions are complementary, not conflicting.

Rationale: A route-anchored send that lands on a generic cold home re-creates the returner drop (Searching 73.1 percent negative, confirmed); parking REC-21 behind an unscoped Journey 10 orphans a carded, buildable moment that J7's measurement needs.

Alternatives and trade-offs:

• Journey 10 owns it (waits on DEFINE; J7 sends and wave returns land cold all quarter)
• Leave unassigned (the AS-IS defect keeps its no-owner status)

Dependencies:

• app_open + lifecycle-state event (Eng, NOW Engineering Requirement)
• journey_state arbitration
• J10 framework item (for the renderer split)

Risks:

• REC-21 fires for organic returners too: its read must not be claimed as J7 send lift (holdout on the zero-return/laggard cohort per the card)

Consequence of delay: REC-21 stays packless with no owner of record, and its read risks being claimed as J7 send lift without the holdout contract on the zero-return/laggard cohort.

Working assumption used in the journey pack(s): All three packs are authored to the same contract (J7 owns, J10 renders it as one state of the S1 recognition slot, J9 feeds); an easy ratification since the three positions are complementary, not conflicting.

Council alignment: No conflict and untouched by the council round: J7, J9 and J10 already agree (J7 owns, J10 renders, J9 feeds); the item stays an easy ratification.

Merge note (v1): Merged: roster decision (b) + DL-J7-03 + J9-D11 + D-J10-03. All three journeys recommend J7-owns; differences (renderer split, transfer clause) are complementary. No conflict.

Provenance: agent-derived DRAFT (three packs concur). Source ids: ROSTER-b (ROSTER.md section 4 (b)); DL-J7-03 (J7/decisions.json); J9-D11 (J9/decisions.json); D-J10-03 (J10/decisions.json).

DL-016: Darwin real-time feed alignment session (adjacent eng decision)

• Journey(s): cross-journey, J2, J4
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Jon (session) / Sheran (marketing ask)
• Recommended decision date: 2026-07-24

Decision required: ADJACENT (engineering, not a journey decision): hold the Darwin real-time feed alignment session (Jon, Lewis, Dirk, Sheran) and settle roadmap home, first slice, resourcing and Braze scoping.

Context and supporting evidence: The Darwin real-time feed (RFC 008) is logged as eng requirement recVy2TBDl71xNFbx; Lewis says full scope is very large and should route through Product; Jon agrees it should be prioritised this quarter and is setting up the session. Use case 1 has flipped to service comms (acceptance options + Delay Repay pointer) converging with the travel-day Live Activity; the swap sell moves to non-disrupted contexts needing no feed.

Recommended option: Per the logged thread: route ownership through Product, phase a narrow Connected Content slice for a user's booked service ahead of full RFC 008, keep 30 Sep as the marketing need-by only, and require Braze in scoping.

Rationale: This is the position already recorded on the eng record after the Lewis/Jon rounds; the session ratifies rather than reopens it.

Alternatives and trade-offs:

• Commit full RFC 008 in-quarter (Lewis: would disrupt in-quarter work)

Dependencies:

• Jon scheduling the session
• Linear ticket still to be raised by eng

Risks:

• Without alignment the travel-day LA use case has no data path this quarter

Consequence of delay: Without the alignment session the travel-day Live Activity use case has no data path this quarter; GRW-17 phase 2 and the J4 LA cell inherit the gap.

Working assumption used in the journey pack(s): Journeys design to the blocked-dependency posture: launch scope assumes no real-time feed, with graceful copy degradation covering the gap (DL-049).

Provenance: source-ruled position (logged eng thread); session ratifies. Source ids: DL-006 (decisions-baseline.json).

DL-047: Travel-day yield order and Veezu call: ruled 5-7 Jul per tobe-composition, held open by the baseline (confirm and close)

• Journey(s): J2
• Priority and urgency: P2
• Status: open, confirm-and-close (stale-ruling conflict)
• Recommended decision-maker: Sheran (one-line confirmation)
• Recommended decision date: 2026-07-24

Decision required: Reconcile the ruling provenance on the same-day Class-B yield order (GRW-17 wayfinding > M33_P > M34_P > M37_P), the Veezu coexistence call and GRW-16's morning-cede: tobe-composition.json records the yield order and Veezu qualification as 'RULED Sheran 5 Jul + Q12 ratified 7 Jul'; the consolidated baseline (compiled from the 5 Jul clash register) carries the same ratification as open.

Context and supporting evidence: The J2 pack now builds on the ruled order as its working assumption and states the verified position on its face (TO-BE map rules, Reference contact rules, matrix GRW-17 row, strategy.md section 3): the RULED labels are verified in tobe-composition (yield order RULED Sheran 5 Jul, Q12 ratified 7 Jul; F3 clamp RULED Sheran 6 Jul); the decisions baseline, compiled from the earlier clash register, is the stale record. The item stays open only for Sheran's one-line confirm-and-close.

Recommended option: Sheran confirms in one line which record is right. If ruled: close this item and mark the drafts' TBCs resolved. If not ruled: strip the RULED labels from tobe-composition (rule-provenance honesty: nothing agent-drafted may read as signed). Note the separate J2 proposal to DELETE GRW-16's gated T-2h not-opened recovery outright, which goes further than the softer morning-cede and is decided on its own item.

Rationale: One home per fact; a ruling that exists in one file and not another will be relied on and disputed simultaneously; a builder treats an unruled yield order as signed, or re-litigates a signed one.

Alternatives and trade-offs:

• Leave both records (guarantees a build-time dispute)

Risks:

• Ordering wrong burns morning cap on commercial over wayfinding

Consequence of delay: Build-time re-litigation of a signed yield order; if the order is wrong, the morning cap burns on commercial over wayfinding.

Working assumption used in the journey pack(s): The J2 pack builds on the ruled order as its working assumption: the RULED labels are verified in tobe-composition (yield order RULED Sheran 5 Jul, Q12 ratified 7 Jul; F3 clamp RULED Sheran 6 Jul); the decisions baseline, compiled from the earlier clash register, is treated as the stale record. Open only for Sheran's one-line confirm-and-close.

Council alignment: STILL STANDS as a record-vs-record conflict (not pack-vs-pack): tobe-composition's RULED labels vs the baseline's open status. The J2 council verified the labels exist and rebased the pack on them; Sheran's one-line confirm-and-close remains, per the rule-provenance honesty rule.

Changed in the council round:

• J2 council (20 Jul 2026) changed context. Why: DL-047's own verification already established the labels exist; presenting a verified ruling as an unresolved either-way dispute would invite build-time re-litigation of a signed order. Rule-provenance honesty is preserved: the confirm-and-close remains Sheran's.

Provenance: source-ruled (per tobe-composition, Sheran 5-7 Jul); held open by the baseline compile. Source ids: DL-018 (decisions-baseline.json); J2-#8 (J2/decisions.json).

DL-015: GRW-09 ownership (J6 owns, J10 renders) plus the referral estate reconciliation

• Journey(s): cross-journey, J6, J10
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Sheran (boundary); Lisa (estate audit)
• Recommended decision date: 2026-07-31

Decision required: Two linked ratifications: (a) GRW-09 (re-booking nudge, in both J6's and J10's moment lists): Journey 6 owns the moment, the outbound legs, the timing clock and the measurement; Journey 10 renders its in-app leg as an S3 tile under a rendering contract (J6 sets content and timing; the framework arbitrates placement). (b) Referral estate: Lisa audits which referral assets still fire (mots names M16E/M17E/M18P as live fragmented asks; the GRW-01 audit shows M46/M47 STOPPED; refer-a-friend broadcasts still fire ad hoc) and consolidates under GRW-28 phase 1; broadcasts stand down for owned episodes.

Context and supporting evidence: Both packs propose the same GRW-09 split (J6's JD6-13 and J10's D-J10-04 agree: one owner, a rendering contract, no fork of the nudge's suppression and clock across two journeys). The referral consolidation is the GRW-28 card's own core recommendation and progresses the open 'referral merge' arbitration job (one of the two left open after the 14 Jul one-lapse-owner ratification).

Recommended option: Ratify the owner/renderer split and commission the estate audit now (cheap now, expensive at QA if zombie sends surface, the X67 pattern).

Rationale: Two owners of one nudge double-message; one owner with a rendering contract does not. One referral programme is the standing direction.

Alternatives and trade-offs:

• J10 owns the in-app leg outright (forks GRW-09's suppression and clock)
• Leave broadcasts running alongside (double-ask risk inside reward windows)

Dependencies:

• J10 framework item
• J6 pack (GRW-09 walk not yet authored)
• REC-21 / roster (b) context

Risks:

• Unruled, both packs spec the same nudge and the clash register gains a row

Consequence of delay: Unruled, both packs spec the same nudge and the clash register gains a row; zombie referral sends risk surfacing at QA (the X67 pattern).

Working assumption used in the journey pack(s): J6 proceeds on DL-015's merged shape: J6 owns the GRW-09 moment, outbound legs, timing clock and measurement; J10 renders the in-app leg under a rendering contract; both packs carry the same arbitration line. GRW-09/J10 wording in the J6 pack now matches this shape.

Council alignment: No conflict: only the J6 council touched this item in the round, adopting the log's merged shape (J6 owns, J10 renders) as its working assumption; both packs carry the same arbitration line. The ratification and Lisa's referral estate audit remain open.

Changed in the council round:

• J6 council (20 Jul 2026) changed working_assumption. Why: Council adopted the log's merged shape as the working assumption (assessment change 5 and the working-assumptions list); the ratification itself and Lisa's referral estate audit remain open as logged.

Merge note (v1): Merged: JD6-13 + D-J10-04. The GRW-09 recommendations agree; the referral audit rides with the J6 item it came from.

Provenance: agent-derived DRAFT (both packs concur). Source ids: JD6-13 (J6/decisions.json); D-J10-04 (J10/decisions.json).

DL-017: Real Snowflake sizing before any powered test starts

• Journey(s): cross-journey
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Ivi (Analytics)
• Recommended decision date: 2026-07-31

Decision required: Replace assumed volumes with real Snowflake sizing before any powered test starts.

Context and supporting evidence: Holdout percentages and runtimes for powered tests depend on real numbers; the sizing gate is with Analytics. Gate-1 ruling 1 permits leading indicators in the interim, so this gates powered reads rather than build.

Recommended option: Analytics delivers sizing per journey as each approaches its test window, following the journey-by-journey Analytics-brief pattern ruled 20 Jul; no powered claim before its sizing lands.

Rationale: Journey-by-journey delivery matches the ruled briefing pattern and avoids a single monolithic sizing exercise blocking everything.

Alternatives and trade-offs:

• One up-front sizing pass across all journeys

Dependencies:

• Snowflake access (in place)
• DL-018 holdout structure

Risks:

• Tests launched under-powered produce unusable reads

Consequence of delay: Tests launched under-powered produce unusable reads; each cohort's one shot at evidence this quarter is burned.

Working assumption used in the journey pack(s): No powered claim before its sizing lands; Analytics delivers sizing per journey as each approaches its test window, per the journey-by-journey Analytics-brief pattern ruled 20 Jul.

Provenance: agent-derived DRAFT. Source ids: DL-012 (decisions-baseline.json).

DL-018: Q3 holdout structure: global control vs canvas-level holdouts

• Journey(s): cross-journey
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Ivi (Analytics) with Sheran ratifying
• Recommended decision date: 2026-07-31

Decision required: Rule the Q3 holdout structure: one global 10% control vs canvas-level holdouts per moment.

Context and supporting evidence: Gate-1 ruling 1 committed to standing up the long-term holdout (Analytics sizes the percentage) with attributed-not-incremental labelling until live, but the structural question from opendecisions remains open. The J8 round added the constraint that methodology must separate journey lift from model targeting where a model triggers the journey.

Recommended option: Agent-derived DRAFT: layered structure: a small global always-out control for programme-level reads plus canvas-level holdouts on moments carrying £ claims, with the model-vs-journey split designed per the J8-round ruling; Analytics sizes both layers.

Rationale: A single global control cannot attribute per-moment lift; per-canvas alone cannot answer the programme question the thesis rests on.

Alternatives and trade-offs:

• Global 10% only
• Canvas-level only

Dependencies:

• DL-017 sizing
• Analytics capacity

Risks:

• Late structure decision delays every powered read in the quarter

Consequence of delay: A late structure decision delays every powered read in the quarter.

Working assumption used in the journey pack(s): Packs design to the layered structure (a small global always-out control for programme-level reads plus canvas-level holdouts on moments carrying GBP claims); Analytics sizes both layers before any test wires.

Provenance: agent-derived DRAFT. Source ids: DL-023 (decisions-baseline.json).

DL-020: Test & Learn roadmap owner and read cadence (interim)

• Journey(s): cross-journey
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Sheran (to assign)
• Recommended decision date: 2026-07-31

Decision required: Name the Test & Learn roadmap owner and read cadence while the Data Strategist role is unfilled.

Context and supporting evidence: Open from 7 Jul. The roadmap is accumulating gated phase-2 rows from every journey (J8 alone added warm-clickers and channel-order), and the one-headline-test-per-release discipline needs someone running the read schedule.

Recommended option: Agent-derived DRAFT: Sheran owns arbitration interim with Ivi running reads on a fortnightly cadence aligned to the eng sprint rhythm, handing over when the Data Strategist is hired.

Rationale: Reads need an owner before the first J8 headline test lands; the fortnightly sprint rhythm already exists to hang the cadence on.

Alternatives and trade-offs:

• Leave unowned until the hire (tests land with no read discipline)

Dependencies:

• Data Strategist hire timeline

Risks:

• Test results accumulate unread; phase-2 gates never open

Consequence of delay: Test results accumulate unread and phase-2 gates never open.

Working assumption used in the journey pack(s): Sheran owns arbitration interim with Ivi running reads on a fortnightly cadence aligned to the eng sprint rhythm, handing over when the Data Strategist is hired.

Provenance: agent-derived DRAFT. Source ids: DL-026 (decisions-baseline.json).

DL-021: Design scope: happy-path plus mandated suppression states

• Journey(s): cross-journey
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Sheran with Design lead
• Recommended decision date: 2026-07-31

Decision required: Rule the design scope: phase-1 happy-path only, or all states designed now.

Context and supporting evidence: Open Sheran question for Design from 7 Jul. Since then, governance suppression rules mandate specific non-happy states (disruption yield, refund/cancel suppression, S1 overlap behaviour) on multiple journeys, so a pure happy-path phase 1 would leave ruled behaviours undesigned.

Recommended option: Agent-derived DRAFT: happy path plus the states the suppression rules already mandate (disrupted, refunded/cancelled, S1 overlap); defer genuinely exotic states to phase 2.

Rationale: The mandated states are trust-critical (a green confirmed state over a cancelled train is the recorded worst-case failure); everything else can wait.

Alternatives and trade-offs:

• Strict happy-path phase 1
• Full state matrix now

Dependencies:

• Design capacity read

Consequence of delay: Design either over-builds exotic states or ships happy-path only with the suppression-mandated states undesigned.

Working assumption used in the journey pack(s): Happy path plus the states the suppression rules already mandate (disrupted, refunded/cancelled, S1 overlap); genuinely exotic states defer to phase 2.

Provenance: agent-derived DRAFT. Source ids: DL-027 (decisions-baseline.json).

DL-022: Transactional-notification reliability check with Eng

• Journey(s): cross-journey
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Eng
• Recommended decision date: 2026-07-31

Decision required: Run the transactional-notification reliability check with Eng.

Context and supporting evidence: Gate-1 ruling 7 (closed) reframed the fatigue ceiling: the majority signal is missing transactional comms, not too many, so reliability is prioritised over throttling. The Eng reliability check it mandated is the open follow-through.

Recommended option: Eng audits delivery of the L0 transactional layer (receipts, win/loss notifications) end to end and reports failure rates before new journey volume lands on the same pipes.

Rationale: Every new journey assumes the L0 layer works; the recorded evidence says that assumption is unsafe.

Risks:

• New journeys amplify an unreliable transactional layer, worsening the top trust complaint

Consequence of delay: New journey volume lands on unaudited transactional pipes and amplifies the top trust complaint.

Working assumption used in the journey pack(s): none adopted for the journeys themselves (they proceed on the existing transactional layer at known risk); the audit is commissioned as a check, not a gate.

Provenance: source-ruled follow-through (Gate-1 ruling 7). Source ids: DL-034 (decisions-baseline.json).

DL-037: AN-1: first habit touch at D+1 for IW/auction winners, channel-separated

• Journey(s): J1
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Sheran + Analytics (Ivi)
• Recommended decision date: 2026-07-31

Decision required: Adopt AN-1: first habit touch moves to D+1 for IW/auction winners (BUN keeps ~D+3), with a CHANNEL-SEPARATION contingency vs M39_E: D+1 habit touch is push-led for armed users; email-only users shift the habit email to D+2; the hard block narrows from 'no D+1 arm for surveyed users' to 'no same-channel same-day stacking'.

Context and supporting evidence: The hazard curve (ask ㉗, confirmed) shows 81.2% of IW winners' quick returns land by D+1, the ruled machinery (floor + ㉗ picks the day, D+1–D+7) has now picked, but the pick collides with M39_E's D+1 email slot and the earlier blanket block on D+1 arms for surveyed users.

Recommended option: Adopt with the channel-separation rule. Persona stays copy-tone only (curves identical, confirmed).

Rationale: The 2nd transaction converts fast or never; waiting to D+3 forfeits the window for 81% of quick returners. Channel separation preserves the M39_E IN ADDITION ruling without giving up the evidence-picked day.

Alternatives and trade-offs:

• Keep D+3 default (rejected: contradicts the ruled machinery's own evidence); re-time M39_E (see J1-D03 alternatives).

Dependencies:

• J1-D03; travel-completion floor (D+1 applies AFTER the floor, same-day travellers satisfy it naturally); flood-guard dial ratification.

Risks:

• Perceived pressure on day 1; mitigated by the one-CTA, no-offer copy rule and the floor.

Consequence of delay: The habit arc keeps the later D+3 start for winners, and the channel-separation contingency against M39_E stays unwired.

Working assumption used in the journey pack(s): AN-1 as drafted: first habit touch at D+1 for IW/auction winners (BUN keeps about D+3), push-led for armed users, email-only users shift to D+2; the hard block is no same-channel same-day stacking; persona stays copy-tone only.

Provenance: agent-derived DRAFT. Source ids: J1-D04 (J1/decisions.json).

DL-038: AN-3: re-order ACT-13's fork; pull the PECR unbundling review forward

• Journey(s): J1
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Sheran + Lisa
• Recommended decision date: 2026-07-31

Decision required: Adopt AN-3: re-order ACT-13's fork so OS-granted-but-not-subscribed is the PRIMARY branch (still Layer 1 transactional-alerts framing); the never-asked OS-prompt fork becomes the minority branch; pull the PECR unbundling review forward.

Context and supporting evidence: Analytics sizing (asks ⑰⑱, confirmed): ~93% of bidders already hold the OS permission; the real gap is the marketing-subscription layer (32%, ~79k/yr addressable). The row design as signed emphasises the OS-prompt fork, and most of the addressable audience sits behind the GATED phase-2 marketing-opt-in test.

Recommended option: Reframe the fork order now (copy/config, no mechanic swap, the slideup mechanic and the Q8 ladder stand unchanged); schedule the PECR unbundling design review so the phase-2 test is not the bottleneck on the sized audience.

Rationale: A consent moment aimed at a 7% path wastes the sanctioned peak-trust window; the mechanic is untouched so this is a wording-and-branch-order change per the as-is-mechanic rule.

Alternatives and trade-offs:

• Leave as signed and let phase 2 carry the reframe (rejected: phase 2 is gated on baseline + PECR sign-off with no date).

Dependencies:

• M29 baseline volume (Analytics); PECR review (Lisa + compliance); real-time consent write-back ER (Eng).

Risks:

• Subscription-layer asks edge closer to marketing consent territory, keep the transactional-alerts value framing and the explicit no-bundling rule.

Consequence of delay: The sized capture audience waits on a phase-2 bottleneck and the fork keeps chasing the minority branch.

Working assumption used in the journey pack(s): The fork is re-ordered now as copy/config (OS-granted-but-not-subscribed becomes the primary branch, Layer-1 transactional-alerts framing); the slideup mechanic and Q8 ladder stand unchanged; the PECR unbundling review is pulled forward.

Provenance: agent-derived DRAFT. Source ids: J1-D05 (J1/decisions.json).

DL-041: Journey 1 headline test = the win-moment one-job rework

• Journey(s): J1
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Sheran (Analytics designs)
• Recommended decision date: 2026-07-31

Decision required: The journey's ONE headline test = the win-moment one-job rework (T31 + in-app: realised-value-led, referral stripped, single NBA) vs the current 5-CTA send. Everything else (opt-in framing, C1 vs bare screen, D+1 confirmation, cumulative savings, ACT-13 marketing A/B) = phase-2 rows gated on the headline read.

Context and supporting evidence: Propagating ruling 8: one headline test per journey release. Four candidate tests exist in this journey's rows.

Recommended option: Win-moment rework as headline; guardrail floor = AS-IS open/click 71.3%/12.2% + opt-out/complaint + total referral volume (referral is recaptured at GRW-28).

Rationale: Largest audience, known guardrail, and it tests the thesis itself (win-to-2nd is the top £ lever in the re-weighted thesis). Voice guardrails are inherited from J8's frog-vs-brand headline; flagged, not re-run.

Alternatives and trade-offs:

• Opt-in capture as headline (rejected: blocked on the opt-in→Tableau P0 for measurement anyway).

Dependencies:

• Correlation token (P0) for the downstream 2nd-txn read; Analytics sizing (holdout %, MDE, runtime, set up front).

Risks:

• Referral-strip effect and copy effect are confounded unless both arms isolate them, the card's test design already splits them; keep that.

Consequence of delay: Five test ideas compete for the win moment, none reads clean, and the journey's one headline read slips past the quarter.

Working assumption used in the journey pack(s): Headline = the win-moment one-job rework vs the current 5-CTA send, guardrail floor at the AS-IS 71.3 percent open / 12.2 percent click plus opt-out/complaint and total referral volume; everything else runs phase-2 gated on the read.

Provenance: agent-derived DRAFT. Source ids: J1-D09 (J1/decisions.json).

DL-042: Standardise Journey 1 figures: 108,821 entry mix + one 2nd-txn KPI

• Journey(s): J1
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Analytics (Ivi)
• Recommended decision date: 2026-07-31

Decision required: Standardise the journey's figures: (a) entry population = 108,821 first purchases/yr (confirmed mix: IW 58,084 · BUN 30,307 · auto-win 16,312 · close-win 4,118), replacing the hub's '62k + 30.3k'; (b) ONE canonical 2nd-transaction KPI definition (portfolio's 5%→7.5% vs T&L's M1 repeat-bid 7.03%→8.8% are different definitions).

Context and supporting evidence: Full-set consistency work already flagged 'standardise shared figures' as a gap; this journey carries both drifts.

Recommended option: Analytics (Ivi) defines the canonical KPI (window, denominator, bid vs any transaction) before the headline test is sized; the hub roadmap row updates to the confirmed mix.

Rationale: One home per fact; a test sized on one definition and reported on another is unreadable.

Alternatives and trade-offs:

• Keep both with labels (rejected: invites drift in the packs that follow).

Dependencies:

• None, definitional.

Risks:

• None material.

Consequence of delay: The headline test is sized against the wrong KPI definition and the hub keeps the stale '62k + 30.3k' mix.

Working assumption used in the journey pack(s): The pack uses the 108,821/yr confirmed entry mix; the canonical 2nd-transaction KPI waits for Ivi's definition (window, denominator, bid vs any transaction) before the headline test is sized.

Provenance: agent-derived DRAFT. Source ids: J1-D10 (J1/decisions.json).

DL-048: GRW-18 power go/no-go: size the T-7-or-earlier slice first

• Journey(s): J2
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Ivi (sizing) + Sheran (go/no-go)
• Recommended decision date: 2026-07-31

Decision required: GRW-18 power go/no-go: size the T-7-or-earlier lead-time slice before promising the holdout read.

Context and supporting evidence: 85% of upgrades are bought within 24h of departure (confirmed, ask 28); only ~15% of wins land more than 24h early and the T-7+ slice is unsized. The Sheran override ships GRW-18 as a test with holdout, but a test on an unsized sliver risks an unreadable result presented as a read.

Recommended option: Analytics sizes the slice first. If underpowered for a Q3 significance read: pre-agree leading indicators (ruling 1 permits them) and consider widening the beat's window (T-5 to T-2) to grow the audience, rather than quietly running an underpowered holdout.

Rationale: Data-strategy standard: flag underpowered tests before they run; agree explicitly where leading indicators substitute.

Alternatives and trade-offs:

• Run the holdout regardless and accept a directional-only read (breaks the honest-reporting standard)
• Fold the expectation-set content into GRW-16's email for short-lead winners (changes the moment model; only if the slice is truly negligible)

Dependencies:

• Analytics lead-time distribution beyond D-3

Risks:

• An unreadable test burns the beat's one shot at evidence and stalls the mid-window design into Q4

Consequence of delay: An unreadable test burns the beat's one shot at evidence and stalls the mid-window design into Q4.

Working assumption used in the journey pack(s): Sizing precedes the promise: if the T-7-or-earlier slice is underpowered, pre-agreed leading indicators and a widened window (T-5 to T-2) replace a quietly underpowered holdout.

Provenance: agent-derived DRAFT. Source ids: J2-#3 (J2/decisions.json).

DL-049: Darwin-feed posture: blocked product-led; confirm the interim payload source

• Journey(s): J2
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Eng/Product (session) + Sheran (posture sign-off)
• Recommended decision date: 2026-07-31

Decision required: Adopt the Darwin-feed posture: real-time platform/coach and live disruption state are briefed as BLOCKED product-led dependencies; launch scope assumes neither; confirm the interim payload source.

Context and supporting evidence: Darwin real-time feed logged as recVy2TBDl71xNFbx (Sprint Backlog, XL); RFC 008 routes through Product; Jon agrees on this-quarter priority and is setting up the alignment session (Lewis, Jon, Dirk, Sheran); 30 Sep is the marketing need-by only, delivery TBC. GRW-17's whole job is 'right platform', so the interim source matters more than the feed's ETA.

Recommended option: Launch GRW-17 Phase 1 on the best-available timetable/Connected-Content payload with the correction rule as the stop-gap, and require Eng to name that interim source + its freshness before build. If no source can carry platform at T-3h reliably, the copy degrades gracefully ('check live platform, tap here') rather than hard-coding a number.

Rationale: Blocked-not-buildable is the standing rule (gate1-rulings 11); a wrong platform number is worse than a deep-link.

Alternatives and trade-offs:

• Hold GRW-17 for the Darwin feed (loses the quarter for the journey's core beat)

Dependencies:

• Jon/Dirk/Lewis alignment session
• Eng interim-source confirmation

Risks:

• A stale hard-coded platform number on the lock screen is the exact failure the moment exists to prevent

Consequence of delay: A stale hard-coded platform number on the lock screen is the exact failure the moment exists to prevent.

Working assumption used in the journey pack(s): Launch scope assumes no real-time feed: GRW-17 Phase 1 ships on the best-available timetable/Connected-Content payload with the correction rule as stop-gap, and copy degrades gracefully ('check live platform, tap here') if no source can carry platform at T-3h reliably.

Provenance: agent-derived DRAFT. Source ids: J2-#4 (J2/decisions.json).

DL-050: Seat-swap boundary: sell only in non-disrupted contexts (propagate the ruling)

• Journey(s): J2
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Sheran (spec ruling) + Product/Eng (gates)
• Recommended decision date: 2026-07-31

Decision required: Fix the seat-swap boundary in the journey design: swap SELL only in non-disrupted contexts (powered by eligibility + deeplink parts 1+2); disrupted contexts get service comms only (acceptance options + Delay Repay pointer).

Context and supporting evidence: The 20 Jul Darwin-thread ruling conceded Lewis's challenge: ticket acceptance already covers disrupted customers free, so a paid swap there charges for an entitlement. The travel-day LA converges with the service-comms play. The old GRW-17 TrainSwap ladder pre-dates this ruling.

Recommended option: Write the boundary into the journey spec: disruption = zero commercial, service content only; the swap sell is a separate non-disrupted play gated on parts 1+2 landing and the eligibility defect fix (DL-056). Retire the disruption-swap framing everywhere it survives in older rows.

Rationale: It is the direct application of a ruling already made; leaving old rows unedited re-creates the charge-for-an-entitlement exposure at build time.

Alternatives and trade-offs:

• None credible: the ruling is made; the open work is propagation

Dependencies:

• DL-056 (eligibility defect)
• parts 1+2 landing (Lewis, Wednesday earliest)

Risks:

• A builder working from the older matrix rows ships a paid swap into a disruption context

Consequence of delay: A builder working from the older matrix rows ships a paid swap into a disruption context.

Working assumption used in the journey pack(s): Disruption = zero commercial, service content only (acceptance options plus the Delay Repay pointer); the swap sell is a separate non-disrupted play gated on parts 1+2 landing and the DL-056 eligibility fix.

Provenance: source-ruled (20 Jul Darwin-thread ruling); open work is propagation. Source ids: J2-#5 (J2/decisions.json).

DL-051: GRW-TD modal single-job scope: savings reflection only, gated on substantiated GBP

• Journey(s): J2
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Sheran (scope) + Ivi (figure) + Product (flag)
• Recommended decision date: 2026-07-31

Decision required: Scope the GRW-TD modal single-job: savings reflection only; move the card's peak-trust opt-in line back to the post-win moment; gate the modal on substantiated saved-GBP and a reliable disrupted-trip flag.

Context and supporting evidence: Ruling 6 held the in-journey upsell and kept the savings-reflection modal. The Gate-1 card still bundles bank-the-win with an opt-in capture line, and its anchor figures (saved-GBP vs walk-up) are unconfirmed. On-Train is a net-negative host stage (67.6%/91.6% negative).

Recommended option: One job: 'you saved X GBP today' + this TOC's inclusions. No opt-in ask (consent capture lives at the post-win moment), no referral ask (advocacy sequencing rule), hard suppression on disrupted trips, and the modal does not ship a GBP figure until Analytics substantiates it.

Rationale: One job per message; never present an unverified number as fact; never let a celebration land on a sour trip.

Alternatives and trade-offs:

• Keep the opt-in line (two asks at the value moment, both weakened)
• Ship with a rounded unverified saving (breaks the figure-integrity rule)

Dependencies:

• D7 saved-GBP feed + floor (Analytics)
• disrupted-trip flag at fire-time (Product/Eng)

Risks:

• An unsubstantiated or wrong saving at the peak-trust moment is a self-inflicted Upgrade Value failure

Consequence of delay: An unsubstantiated or wrong saving lands at the peak-trust moment: a self-inflicted Upgrade Value failure.

Working assumption used in the journey pack(s): One job only (savings reflection plus this TOC's inclusions); no opt-in ask, no referral ask, hard suppression on disrupted trips; no GBP figure ships before Analytics substantiates it.

Provenance: agent-derived DRAFT. Source ids: J2-#6 (J2/decisions.json).

DL-052: Close the post-win slot arbitration: J2 owns S1; J1 rebook floors at travel+1d

• Journey(s): J2, J1
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Sheran
• Recommended decision date: 2026-07-31

Decision required: Close the 'post-win slot' arbitration (one of the 4 open arbitration jobs) by ratifying: the pre-travel state owns the slot per rule 6 L1, and Journey 1 rebook-type touches floor at travel-completion + 1 day.

Context and supporting evidence: Both packs now carry the same contract on their faces: the J1 council (20 Jul) applied the rule as J1's working assumption, and the J2 pack now states the mirror (Process step 0, strategy.md section 5): while a won trip is inside S1, J2 owns every marketing slot; J1's habit/rebook sequence floors at travel-completion + 1 day; the T+0 win beat stays J1's as the entry hand-off. One Sheran ruling closes both packs.

Recommended option: Ratify as stated: while a won trip is inside S1, J2 owns every marketing slot; J1's habit/rebook sequence starts at travel + 1 day; the win beat itself (T+0 celebration + consent capture) remains J1's and is exempt because it is the entry hand-off.

Rationale: It is the direct application of layered rule 6 L1 (active mission owns the slot) to the one overlap that affects 710,616 trips a year.

Alternatives and trade-offs:

• Time-share the window (two journeys messaging one customer about one trip: the failure mode the journey_state field exists to prevent)

Dependencies:

• Rule 6 (ratified layer model)
• GRW-01 travel-completion gate rows

Risks:

• Unclosed, both journeys build sends into the same days and the cap arbitration decides by accident

Consequence of delay: Unclosed, both journeys build sends into the same days and the cap arbitration decides by accident.

Working assumption used in the journey pack(s): Both packs carry the same contract on their faces: while a won trip is inside S1, J2 owns every marketing slot; J1's habit/rebook sequence floors at travel-completion + 1 day; the T+0 win beat stays J1's as the entry hand-off. One Sheran ruling closes both packs.

Changed in the council round:

• J2 council (20 Jul 2026) changed context. Why: Without the mirror note the two packs read as independently proposing the same rule; the ratification should be presented as one decision covering both.

Provenance: agent-derived DRAFT. Source ids: J2-#10 (J2/decisions.json).

DL-053: M36_P ownership and consolidation into the M34_P winner

• Journey(s): J2
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Sheran (ownership) / Lisa (taxonomy)
• Recommended decision date: 2026-07-31

Decision required: Rule M36_P ownership (GRW-17's MoT vs GRW-16's T-1d window) and its consolidation into the M34_P winner.

Context and supporting evidence: M36_P is estate-mapped to GRW-17's moment but fires inside GRW-16's day-before window; the register records the merge/consolidate direction in the arc-handoff register and assigns the J2-level ownership call to Sheran. The M34/35/36/38 shared-code taxonomy fix sits separately with Lisa.

Recommended option: Register direction: consolidate M36_P into the M34_P winner and record arc ownership at the J2 handoff level; taxonomy fix proceeds in parallel with Lisa.

Rationale: The consolidation direction is already registered; the open step is the ownership ruling that lets it execute.

Alternatives and trade-offs:

• Leave M36_P standalone under GRW-16 ownership

Dependencies:

• Lisa's taxonomy fix on the shared BUN/UpgradeWon comm codes

Consequence of delay: M36_P keeps double ownership and the taxonomy fix on the shared BUN/UpgradeWon comm codes stalls.

Working assumption used in the journey pack(s): Direction registered: M36_P consolidates into the M34_P winner with arc ownership recorded at the J2 handoff level; the taxonomy fix proceeds with Lisa in parallel.

Provenance: agent-derived DRAFT. Source ids: DL-019 (decisions-baseline.json).

DL-054: Rollback floor X for the M31/M32 A/B + validate the M31_E offset

• Journey(s): J2
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Ivi + Lisa
• Recommended decision date: 2026-07-31

Decision required: Set the rollback floor X for the M31/M32 test-then-replace A/B, and validate the assumed M31_E offset.

Context and supporting evidence: The day-before creative A/B keeps current creative as control and replaces only on win or tie; the rollback floor is TBC with Ivi and already routed. Separately, the T-22h offset is confirmed for M32/M29/M30 but only assumed for M31_E (audit says validate).

Recommended option: TBC: needs Ivi's floor number; validate the M31_E offset in Braze before the A/B enters build.

Rationale: This is the estate's only proven-incremental family (+16.3%/+16.9%); the floor protects it and the offset check protects the test design.

Dependencies:

• Ivi (floor)
• CRM Braze check (offset)

Risks:

• Testing against a mis-timed control corrupts the read on the proven send

Consequence of delay: Testing against a mis-timed control corrupts the read on the proven send.

Working assumption used in the journey pack(s): none adopted: the M31/M32 A/B does not enter build until Ivi's floor number lands and the M31_E offset is validated in Braze.

Provenance: TBC (floor: Ivi; offset: CRM Braze check). Source ids: DL-021 (decisions-baseline.json).

DL-055: Persistent travel-day component: reskin Trips view vs net-new

• Journey(s): J2
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Design/Product
• Recommended decision date: 2026-07-31

Decision required: Decide the persistent travel-day component: net-new build or reskin of the Trips upgrade-confirmed view.

Context and supporting evidence: GRW-17's build is gated on this surface existing (approved for design, build gated on trip-state surface). The raid log records it as a proposed build that does not exist today, with interim anchors in use; one trip-state component is shared with GRW-16 (GRW-16 owns it-worked, GRW-17 owns platform-and-coach-today).

Recommended option: Agent-derived DRAFT: extend/reskin the existing Trips upgrade-confirmed view first (anchor to the existing mechanic), going net-new only if that view cannot carry the live platform/coach payload and the shared GRW-16 state.

Rationale: Anchoring to the existing mechanic is the standing recommendation-framing rule and shortens the path to unblocking GRW-17's build gate.

Alternatives and trade-offs:

• Net-new persistent component

Dependencies:

• Product Design capacity
• DL-047 travel-day arbitration

Risks:

• The consent-free fallback for non-push users has no home until this exists

Consequence of delay: The consent-free fallback for non-push users has no home on travel day.

Working assumption used in the journey pack(s): Extend/reskin the existing Trips upgrade-confirmed view first (anchor to the existing mechanic); net-new only if that view cannot carry the live platform/coach payload and the shared GRW-16 state.

Provenance: agent-derived DRAFT. Source ids: DL-028 (decisions-baseline.json).

DL-056: Fix the TrainSwap eligibility defect before the GRW-17 ladder ships

• Journey(s): J2
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Product/Eng
• Recommended decision date: 2026-07-31

Decision required: Fix the TrainSwap eligibility defect before the GRW-17 TrainSwap ladder ships.

Context and supporting evidence: Recorded as a hard gate in raid: the defect must be fixed before that path ships. It gates only the TrainSwap ladder, not the rest of GRW-17.

Recommended option: Keep the TrainSwap ladder out of the GRW-17 launch scope until Product/Eng confirm the fix; the rest of the journey proceeds.

Rationale: Scoping the ladder out decouples the journey launch from the defect timeline.

Alternatives and trade-offs:

• Hold all of GRW-17 for the fix (over-gates)

Dependencies:

• Product/Eng defect fix

Risks:

• Shipping the ladder against the defect sends customers into a broken swap flow

Consequence of delay: Shipping the ladder against the defect sends customers into a broken swap flow.

Working assumption used in the journey pack(s): The TrainSwap ladder stays out of the GRW-17 launch scope until Product/Eng confirm the fix; the rest of the journey proceeds.

Provenance: agent-derived DRAFT. Source ids: DL-030 (decisions-baseline.json).

DL-062: Journey 3 headline test = the ACT-02 bid-add reassurance holdout

• Journey(s): J3
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Sheran (ratify), Analytics (size)
• Recommended decision date: 2026-07-31

Decision required: Ratify the J3 headline test: the ACT-02 bid-add reassurance holdout. Everything else is a phase-2 row gated on the headline read and its own baseline.

Context and supporting evidence: Ruling 8: one headline test per journey release. J3 has six candidate tests across its cards. ACT-05 is separately ruled as a test, but its first phase is a diagnostic (analysis, not a send), so it does not compete for the headline slot.

Recommended option: Headline = dwell-triggered reassurance/speed-up at the bid-add step vs bare step, KPI first-bid rate among never-bid dwellers. Phase-2: ACT-05 arms (post-diagnostic), ACT-08 value-box A/B + risk-free relocation, ACT-22 steer, ACT-26 opt-in-width A/B, ACT-13 marketing-copy A/B.

Rationale: The bid-add step is the biggest single leak (48 to 18, ~34k/mo assumed pending verification) with the most specific evidenced barrier (bid-commit fear); it is the cleanest single causal read the journey can buy.

Alternatives and trade-offs:

• Make ACT-05 the headline (smaller cohort, and its ruling already sequences a diagnostic first)
• Make ACT-08's value box the headline (contested prior Maze evidence makes it a messier read)

Dependencies:

• Bid-add dwell event + bid_behaviour attribute (blocked list, decision 3)
• Analytics baseline/MDE/significance/runtime set up front

Risks:

• Dwell volume may underpower within Q3: agree leading indicators up front, not post-hoc

Consequence of delay: Five J3 test ideas compete for one cohort, the dwell-volume risk goes unexamined, and the headline read slips.

Working assumption used in the journey pack(s): Headline = the ACT-02 bid-add reassurance holdout, KPI first-bid rate among never-bid dwellers, with leading indicators pre-agreed for the dwell-volume risk; everything else is a phase-2 row gated on the headline read.

Provenance: agent-derived DRAFT. Source ids: J3-#4 (J3/decisions.json).

DL-065: J3/J5 window rule: 7-day episode, then hand back to Journey 5

• Journey(s): J3, J5
• Priority and urgency: P2
• Status: open; working assumption adopted in both the J3 and J5 packs (7-day episode window, J5 suppressed during owned episodes, automatic hand-back at close, re-entry on a new funnel session); ratification rides the DL-004 recovery-owner pass
• Recommended decision-maker: Sheran
• Recommended decision date: 2026-07-31

Decision required: Ratify the J3/J5 window rule: a J3 episode owns the abandoner for 7 days post-abandon, then hands back to the Journey 5 programme; re-entry requires a new funnel session.

Context and supporting evidence: Most J3 abandoners are also J5's never-transacted audience; without a window rule the same user gets funnel recovery and activation programme sends in the same week.

Recommended option: 7-day episode window (mirrors the loss-side 7-day gate and the J8 window shape); J5 suppressed during owned episodes; hand-back automatic at close.

Rationale: One owner per customer (ruling 4) applied at the J3/J5 seam; 7 days matches the existing loss-recovery gate so the estate has one rhythm.

Alternatives and trade-offs:

• 72h window (faster hand-back, but T2 at +22h plus a read window barely fits)
• 14 days (starves J5 of its own audience)

Dependencies:

• journey_state (decision 1)

Risks:

• Window length is a judgment call; revisit on the first read

Consequence of delay: Unratified, J3 and J5 both claim the abandoner in week one; a shared item under two verdicts is the exact drift DL-003 exists to stop.

Working assumption used in the journey pack(s): Both packs carry the identical contract: 7-day episode window, J5 suppressed during owned episodes, automatic hand-back at close, re-entry on a new funnel session; ratification rides the DL-004 recovery-owner pass.

Changed in the council round:

• J3 council (20 Jul 2026) changed status. Why: The window sets a contact-ownership boundary between two journeys and the J5 pack (finalised 20 Jul) keeps DL-065 ratification with Sheran while adopting the window. Closing it from J3's side alone would put one shared item under two verdicts, the exact drift DL-003 exists to stop. Both packs now carry the identical contract with the identical status, and it costs Sheran nothing extra: it is one line inside the DL-004 ruling he already has to make.

Provenance: agent-derived DRAFT. Source ids: J3-#7 (J3/decisions.json).

DL-070: AN-2 fold: one consolidated failure-reassurance send; box copy keeps 8 branches

• Journey(s): J4
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Sheran, on Analytics' AN-2 recommendation
• Recommended decision date: 2026-07-31

Decision required: Ratify the AN-2 fold: consolidate the tiny close-reason SENDS (technical, early-close, operator-withdrew) into one failure-reassurance message; keep the full 8-branch matrix as copy inside the in-app outcome box; keep the supply-returns notify track.

Context and supporting evidence: Analytics challenged the 8 bespoke close-reason branches as disproportionate: technical is 521 auctions (1,256 bidders)/yr and operator-withdrew 913 (2,071). The branching cost sits in the sends, not the box copy. The supply-returns track is separately validated: 97% of withdrawn auctions get a fresh same-route auction within 14 days.

Recommended option: Agent-proposed DRAFT, adopting Analytics AN-2: one consolidated failure-reassurance send (R3 becomes a variant of it, not a bespoke build); box copy keeps all 8 branches (trust-critical and near-free); R4 supply-returns ships as designed.

Rationale: Nobody builds a bespoke canvas for 1,256 people a year; but 'you lost' on an auction that never ran is the fastest trust-killer, so the copy branching must survive the fold.

Alternatives and trade-offs:

• Build R3 bespoke as drafted (rejected by the sizing)
• Fold the copy branches too (rejected: falsehood risk on Seatfrog-caused closes)

Dependencies:

• Payment-hold lifecycle (⑦) for the reassurance clause

Risks:

• Unruled, the build brief carries three sends where sizing supports one

Consequence of delay: The build brief carries three sends where sizing supports one.

Working assumption used in the journey pack(s): One consolidated failure-reassurance send (R3 becomes a variant, not a bespoke build); the in-app outcome box keeps all 8 copy branches; R4 supply-returns ships as designed.

Provenance: agent-derived DRAFT. Source ids: J4-D03 (J4/decisions.json).

DL-072: Pre-agree the ACT-23 decision rule (fold into ACT-11 below threshold)

• Journey(s): J4
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Sheran, threshold proposed by Ivi
• Recommended decision date: 2026-07-31

Decision required: Pre-agree the ACT-23 decision rule: if Analytics sizes knocked_out volume below a build-worthy threshold, fold the copy fix into ACT-11 and close the card.

Context and supporting evidence: ACT-23's Refine verdict (12 Jul) put a sizing gate before any build slot. Pre-agreeing the fold rule now means the sizing read executes itself instead of returning for another ruling.

Recommended option: Agent-proposed DRAFT: Analytics proposes the threshold with the sizing (events/yr at which a native half-screen earns an app-release slot); below it, ship the KO push copy fix inside ACT-11's Phase 1 and close ACT-23.

Rationale: Decision rules set before the read (data-strategy standard); avoids a second Gate-1 pass on a card that may not survive its own sizing.

Alternatives and trade-offs:

• Wait for the number then decide (an extra review cycle)

Dependencies:

• Analytics knocked_out sizing
• knocked_out event reliability (Eng)

Risks:

• Without the rule, a rare-event card occupies a Product slot ahead of the one-tap re-bid surface

Consequence of delay: A rare-event card occupies a Product slot ahead of the one-tap re-bid surface.

Working assumption used in the journey pack(s): Analytics proposes the threshold with the sizing; below it, the knocked-out push copy fix ships inside ACT-11 Phase 1 and ACT-23 closes.

Provenance: agent-derived DRAFT. Source ids: J4-D08 (J4/decisions.json).

DL-073: First-time-loser exhaust hand-off: split by transaction history into J5

• Journey(s): J4, J5
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Sheran, with the J5 pack author
• Recommended decision date: 2026-07-31

Decision required: Ratify the first-time-loser exhaust hand-off: never-transacted losers exiting the day-14 close hand to Journey 5's pool with first-timer flags; transacted losers hand to the lifecycle bands.

Context and supporting evidence: Both packs now carry the same contract: J5 (authored 20 Jul, council-reviewed) adopts the re-entry with the first-timer flags named (first_loss, close_reason, route), and the J4 pack now mirrors the flag enumeration on its exhaust node, Reference-tab boundaries and strategy hand-offs.

Recommended option: Agent-proposed DRAFT: split the exhaust by transaction history; carry first-timer flags (first_loss, close_reason, route) into J5's entry so its copy can use the bad-luck-not-rejection reframe.

Rationale: The hand-off is where the journey's biggest cohort (first-time losers) either gets the right next journey or disappears; one flag split fixes it.

Alternatives and trade-offs:

• Single exhaust to lifecycle bands (loses the activation hand-off)

Dependencies:

• Journey 5 pack entry definition (in authoring)

Risks:

• Unruled, J4 and J5 both assume the other owns this cohort

Consequence of delay: Unruled, J4 and J5 both assume the other owns the never-transacted loser cohort and the hand-off double-fires or gaps.

Working assumption used in the journey pack(s): Both packs carry the same contract: never-transacted losers exiting the day-14 close hand to Journey 5's pool with the first-timer flags named (first_loss, close_reason, route); transacted losers hand to the lifecycle bands under journey_state. The J4 pack mirrors the flag enumeration on its exhaust node; J5 adopts the re-entry.

Changed in the council round:

• J4 council (20 Jul 2026) changed context. Why: The cross-pack mirror is what makes the hand-off buildable; an unnamed flag set on either side would have left the two canvases guessing at each other's payload.

Provenance: agent-derived DRAFT. Source ids: J4-D10 (J4/decisions.json).

DL-074: Journey 4 headline test = the D+7 gate holdout

• Journey(s): J4
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Sheran, sized by Ivi
• Recommended decision date: 2026-07-31

Decision required: Approve the J4 headline test as the D+7 gate holdout (gated recovery vs powered no-send control among 7-day non-returners); everything else runs phase-2 on that read.

Context and supporting evidence: Ruling 8 mandates one headline test per journey release. J4 has five testable beats; the D+7 gate is the one carrying the £ claim (£92.70 LTV gap across 102,832/yr eligible) and the one whose prior reads are negative, so it needs the powered holdout before any claim regardless.

Recommended option: Agent-proposed DRAFT: headline = D+7 recovery vs holdout, primary KPI return-to-bid/BUN conversion among non-returners, guardrail overall bid volume (prior blanket read −11.8%) plus opt-out/complaint. Phase-2 queue: M21_E rework holdout, ACT-11 one-tap, ACT-33 branded alert (pull-forward decision rule pre-agreed), ACT-23 guided resolution, LA observational read.

Rationale: The gate is the journey's thesis; if it does not read, the phase-2 beats are re-prioritised anyway.

Alternatives and trade-offs:

• Headline on ACT-11 one-tap (Product-gated, cannot start first)
• Two concurrent headlines (contaminates the loss cohort, ACT-11's own test note warns against it)

Dependencies:

• J4-D02 (sizing definition)
• DL-018 (holdout structure)
• correlation-token P0 for the push beats

Risks:

• Unruled, five tests compete for one loss cohort and none reads clean

Consequence of delay: Five tests compete for one loss cohort and none reads clean.

Working assumption used in the journey pack(s): Headline = D+7 recovery vs a powered no-send control among 7-day non-returners; KPI return-to-bid/BUN conversion; guardrails overall bid volume (prior blanket read minus 11.8 percent) plus opt-out/complaint; the phase-2 queue runs on that read.

Provenance: agent-derived DRAFT. Source ids: J4-D11 (J4/decisions.json).

DL-075: Ratify the loss-close estate verdicts (T20_P, T1_E, T2_E/T30_E)

• Journey(s): J4
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Sheran
• Recommended decision date: 2026-07-31

Decision required: Ratify the loss-close estate verdicts: T20_P retire (overriding the estate Keep), T1_E delayed 30 minutes as unseen-fallback, and the one-comm-or-two question on T2_E/T30_E.

Context and supporting evidence: The close-moment cluster is reconciled in the register but three of its verdicts are TO-BE overrides of estate verdicts and are queued for Sheran's sign-off in the rows. T2_E/T30_E may be two live sends on one event (177/wk, confirmed).

Recommended option: Ratify as reconciled in the register: retire T20_P as a distinct loser variant (folds into the stripped L0), keep the T1_E 30-minute delay, and collapse T2_E/T30_E to one send once eng confirms whether both fire.

Rationale: These verdicts implement the approved ACT-07 shape (marketing stripped from the transactional layer); sign-off is the outstanding step, not analysis.

Alternatives and trade-offs:

• Keep T20_P per the original estate verdict

Dependencies:

• Eng check on whether T2_E and T30_E both fire on auction_closed

Consequence of delay: The loss-close estate keeps three unreconciled verdicts and QA discovers them late.

Working assumption used in the journey pack(s): The pack builds to the reconciled register verdicts: T20_P retires as a distinct loser variant (folds into the stripped L0), T1_E keeps the 30-minute unseen-fallback delay, and T2_E/T30_E collapse to one send pending the Eng both-fire check.

Provenance: agent-derived DRAFT. Source ids: DL-016 (decisions-baseline.json).

DL-076: Sign off the M21_E day-1 education-only rework (register discrepancy)

• Journey(s): J4
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Sheran
• Recommended decision date: 2026-07-31

Decision required: Sign off the M21_E day-1 loser email change, resolving the register-vs-approval discrepancy on its shape.

Context and supporting evidence: The clash register (5 Jul) says REPLACE: retire the day-1 send and fold into the 7-day-gated journey. The later Gate-1 approval card (8-9 Jul) says REWORK M21_E to day-1 education-only, and governance encodes day-1 education vs day-7 recovery as distinct jobs never same-day. M21_E's current incremental read is negative (minus 3.73% vs control, confirmed).

Recommended option: Ratify the Gate-1 approval shape (rework to day-1 education-only) as the later, superseding verdict; treat the register's retire verdict as overtaken.

Rationale: The approval card and governance post-date the register and are internally consistent; the negative incremental read argues against keeping the current commercial day-1 send, which both shapes remove.

Alternatives and trade-offs:

• Full retirement per the register (loses the education beat governance assumes)

Risks:

• Building to the wrong shape if the discrepancy goes unnoticed

Consequence of delay: The build proceeds to the wrong M21_E shape if the register-vs-approval discrepancy goes unnoticed.

Working assumption used in the journey pack(s): The Gate-1 approval shape (rework to day-1 education-only) is treated as the later, superseding verdict; the register's retire verdict is treated as overtaken.

Council alignment: STILL STANDS as a record-vs-record discrepancy: the register's retire verdict vs the Gate-1 approval's rework shape for M21_E. The pack treats the approval as the later, superseding verdict; Sheran signs off the shape.

Provenance: agent-derived DRAFT. Source ids: DL-017 (decisions-baseline.json).

DL-085: Extend the ruled capture cooldown ladder-wide (one shared decline counter)

• Journey(s): J5, J1
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Sheran; CRM (Lisa) implements
• Recommended decision date: 2026-07-31

Decision required: Extend the ruled ACT-13 capture cooldown (30d, max 2 asks/30d) ladder-wide as ONE shared decline counter across ACT-15, ACT-12, ACT-13 and ACT-24.

Context and supporting evidence: The ACT-24 adversarial pass (7c) found no cross-moment decline handling: a decline at any rung must suppress/soften the ask at every other rung. Ruled today for ACT-13 only (Q8).

Recommended option: Ratify the ladder-wide counter and resolve the recorded ACT-24 C1 conflict (BUN-card 're-prompt next session' vs the 30d cooldown, already logged in the consolidated baseline) in favour of the cooldown.

Rationale: Four surfaces asking independently is how consent UX burns trust at exactly the moments this journey needs it.

Alternatives and trade-offs:

• Per-surface cooldowns with a global 3-ask cap

Dependencies:

• Decline-counter attribute (journey-written, D7 pin)
• ACT-24 C1 ruling (baseline log)

Risks:

• Capture rate collapses from ask fatigue; opt-out rises

Consequence of delay: Ask fatigue collapses capture rate and opt-out rises as each surface re-prompts independently.

Working assumption used in the journey pack(s): One shared decline counter across ACT-15, ACT-12, ACT-13 and ACT-24 on the ruled 30-day / 2-asks ladder; the recorded ACT-24 C1 conflict resolves in favour of the cooldown (with DL-035).

Provenance: agent-derived DRAFT. Source ids: J5-D-08 (J5/decisions.json).

DL-092: T+24h slot arbitration order: advocacy > review > savings-tracker

• Journey(s): J6
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Sheran
• Recommended decision date: 2026-07-31

Decision required: Ratify the T+24h slot arbitration order: advocacy (when AND-eligible) > review (won-bid, unrated; private survey for no-upgrade) > savings-tracker default.

Context and supporting evidence: The GRW-29 card explicitly asks Sheran to arbitrate the post-trip slot (review ask vs re-bid/next-trip nudge); this pack proposes the answer and extends it to advocacy.

Recommended option: Adopt the order. Natural effect: first positive trip gets the review ask (advocacy is rule-banned there), second-plus positive trips get the referral ask; review re-offers on a later positive trip if unrated and advocacy did not fire.

Rationale: Advocacy is the rarest and highest-value eligible event and is already gated hard; review harvests the first-trip delight the rule leaves free; the tracker default keeps the slot warm without a second job.

Alternatives and trade-offs:

• Review always outranks advocacy (delays referral by at least one more trip)
• Fixed calendar rotation (ignores eligibility state)

Dependencies:

• JD6-03 gate
• D9 review-cap state

Risks:

• An eligible customer who is never asked to rate; mitigated by the re-offer rule

Consequence of delay: Two eligible messages compete for the one T+24h slot per trip and the review and referral asks collide.

Working assumption used in the journey pack(s): The order is adopted in the pack: advocacy (when AND-eligible) > review (won-bid, unrated; private survey for no-upgrade) > savings-tracker default, with the re-offer rule covering unrated customers.

Provenance: agent-derived DRAFT. Source ids: JD6-05 (J6/decisions.json).

DL-093: Commission personal_median_inter_txn_gap as a shared computed trait (J6 + J8)

• Journey(s): J6, J8
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Analytics (Ivi) build; Sheran signs the 1.0x/1.5x dials
• Recommended decision date: 2026-07-31

Decision required: Commission personal_median_inter_txn_gap as a computed trait SHARED with Journey 8, and time GRW-09 on it (fallback: base median 15-16d).

Context and supporting evidence: The cards time the nudge at the global median and float persona-based timing. Persona never sets cadence (confirmed); J8's slip thinking already treats the customer's own cycle as the unit.

Recommended option: One warehouse trait, two consumers: J6 nudges at ~1.0x the personal median; J8's slip window opens at ~1.5x. Brief it in both packs' Analytics sections as one build.

Rationale: One home per fact and half the build cost; the same rhythm that times the nudge defines the slip, so the two journeys can never disagree about whose customer it is.

Alternatives and trade-offs:

• Fixed 15-16d for all (ships now, is the fallback anyway)
• Wait for D2 churn score to carry cadence (couples J6 to a gated dependency)

Dependencies:

• Analytics build capacity; nightly sync path

Risks:

• Sparse history (n<3 txns) makes the trait noisy; the fallback rule covers it

Consequence of delay: J6's nudge and J8's slip window run on different clocks; the shared trait gets built twice or not at all.

Working assumption used in the journey pack(s): One warehouse trait, two consumers: J6 nudges at about 1.0x the personal median, J8's slip window opens at about 1.5x; fallback = base median 15-16d until the trait lands; briefed in both packs as one build.

Provenance: agent-derived DRAFT. Source ids: JD6-07 (J6/decisions.json).

DL-095: Power users in Journey 6: suppressed from nudges, eligible for review/advocacy

• Journey(s): J6
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Sheran (pends D1)
• Recommended decision date: 2026-07-31

Decision required: Rule power users' place in Journey 6: suppressed from the reflection push and habit nudge; eligible for review and advocacy under their own caps.

Context and supporting evidence: Blanket nudges cost -11.8% in holdout (confirmed) and GRW-01 A3 makes suppression the treatment at the win moment; but proven multi-win advocates are the natural referrer pool and the best rating source.

Recommended option: Suppress GRW-04 push + GRW-09 for the interim RW power cut; allow GRW-29 once per store version and GRW-28 once per open reward window (proposed additional cap: max 2 advocacy asks/yr for this cell).

Rationale: Over-sending is the harmful failure for this cohort; the two asks kept are the ones only this cohort can answer well.

Alternatives and trade-offs:

• Full J6 suppression for power users (loses the best advocates)
• No special handling (contradicts the holdout evidence)

Dependencies:

• D1 power definition (Sheran, open); roster decision (a) if the power journey folds into J6

Risks:

• If roster decision (a) lands elsewhere, these rows move; flagged, not assumed

Consequence of delay: Power users get the reflection push and habit nudge the over-send evidence (minus 11.8 percent) warns against.

Working assumption used in the journey pack(s): GRW-04 push and GRW-09 suppressed for the interim RW power cut; GRW-29 once per store version; GRW-28 once per open reward window with the proposed max 2 advocacy asks/yr; rows move if roster decision (a) lands elsewhere (flagged, not assumed).

Provenance: agent-derived DRAFT. Source ids: JD6-09 (J6/decisions.json).

DL-096: Episode contact cap for J6: 3 outbound sends (deviates from J8's 2-touch cap)

• Journey(s): J6
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Sheran
• Recommended decision date: 2026-07-31

Decision required: Ratify the episode contact cap: 3 outbound sends per post-trip episode (deviates from J8's 2-touch cap).

Context and supporting evidence: The J8 episode shape is the propagating pattern, but J6 carries three distinct, non-substitutable jobs (value proof, the arbitrated slot, the habit nudge). Under the DL-002 shape M39_E is a fourth touch in the window for some customers: it sits outside J6's episode cap but inside the weekly cap, with the same-day separation rule holding the per-day ceiling.

Recommended option: Cap at 3 with the spine's shape as the ceiling, all inside the flood guard; in-app surfaces and L0 excluded.

Rationale: Two touches would force choosing between the review/advocacy slot and the habit nudge every trip; the jobs are non-substitutable and each is once-per-trip capped already. The council restated the rationale after DL-002: the survey-rework justification no longer applies; the case rests on job distinctness alone.

Alternatives and trade-offs:

• Hold the J8 2-touch cap (drop the cadence nudge whenever the T+24h slot fires)
• 4+ (re-opens the pile-up this pack exists to close)

Dependencies:

• Backbone dials ratification (Class C 3/week, flood guard 4/day still pending)

Risks:

• Deviation precedent: other packs may ask for their own cap; mitigate by ruling it per-journey with rationale

Consequence of delay: Unratified, the deviation precedent is unmanaged (other packs may ask for their own caps) and the J6 build cannot fix its episode ceiling.

Working assumption used in the journey pack(s): The cap of 3 outbound sends per post-trip episode is adopted in the pack, resting on job distinctness alone; under the DL-002 shape M39_E sits outside the episode cap but inside the weekly cap, with the same-day separation rule holding the per-day ceiling.

Changed in the council round:

• J6 council (20 Jul 2026) changed context. Why: The cap was partly justified by the survey rework, which the DL-002 adoption removed; the council restated the case so it rests on job distinctness alone and states M39_E's cap interactions, keeping the ratification ask honest. Sheran ratifies the deviation as before.
• J6 council (20 Jul 2026) changed rationale. Why: Same restatement as the context change; the decision itself (cap at 3, Sheran ratifies) is unchanged.

Provenance: agent-derived DRAFT. Source ids: JD6-10 (J6/decisions.json).

DL-097: Journey 6 headline test = spine-level holdout, next-transaction within 60d

• Journey(s): J6
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Analytics (Ivi) design + Sheran approve
• Recommended decision date: 2026-07-31

Decision required: Approve the headline test: journey-level holdout on the whole spine, KPI = next-transaction rate within 60d of a positive trip; release 2 (share prompt, cadence test) gated on its read.

Context and supporting evidence: Ruling 8 allows one headline test per journey release; the cards independently proposed four separate tests.

Recommended option: One spine-level holdout; GRW-29's link fix reads pre/post outside the test; phase-2 rows as listed in the pack.

Rationale: The journey's claim is the spine, not any single send; a per-moment test family would under-power four reads instead of powering one.

Alternatives and trade-offs:

• Per-moment holdouts (the cards' default; under-powered and mutually contaminating)

Dependencies:

• Baseline + holdout sizing (Ivi)
• JD6-04 signal (defines the eligible population)

Risks:

• A null read stalls release 2; acceptable, that is what gating means

Consequence of delay: Release 2 has no gate and the spine ships without its one causal read.

Working assumption used in the journey pack(s): One spine-level holdout, KPI next-transaction rate within 60d of a positive trip; GRW-29's link fix reads pre/post outside the test; release 2 gates on the read.

Provenance: agent-derived DRAFT. Source ids: JD6-11 (J6/decisions.json).

DL-102: Restate the O4 reactivation denominator (band-level lift, not 1.5 percent on the pool)

• Journey(s): J7
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Sheran with Analytics (Ivi)
• Recommended decision date: 2026-07-31

Decision required: Restate the denominator of the O4 reactivation objective (0.24% to 1.5%).

Context and supporting evidence: 1.5% on ~600k implies ~9k reactivations. Confirmed evidence caps the programme: overall reactivation ~1% (push-only best 1.12%), ~40% of the pool unreachable, Churned one-time (the bulk of the stock) at 2.0% propensity.

Recommended option: Restate as band-level lift on the triggered-eligible pool (where Lapsed multi/PU run 20 to 30% historical reactivation), keep 1.5% as blended aspiration only, and book the clean suppression list as a co-equal deliverable. Value the result at £219.94 per reactivation (confirmed), never £22.75.

Rationale: Gate-1 should not sign a target the evidence already refutes; band-level targets are both achievable and honest to the 7-15x propensity spread.

Alternatives and trade-offs:

• Keep 1.5% on the full pool (guaranteed miss, poisons the programme read)
• Drop the % target for a £ target (loses comparability with the V2MOM)

Dependencies:

• DL-J7-11 pool reconciliation
• Analytics band sizing

Risks:

• A restated target can read as sandbagging; the 7-15x gradient evidence is the defence

Consequence of delay: The journey is judged against 1.5 percent on the wrong denominator and reads as a failure (or as sandbagging) either way.

Working assumption used in the journey pack(s): The objective is stated as band-level lift on the triggered-eligible pool (Lapsed multi/PU run 20-30 percent historical reactivation); 1.5 percent stays blended aspiration only; the clean suppression list is a co-equal deliverable; reactivations valued at GBP 219.94 (confirmed).

Provenance: agent-derived DRAFT. Source ids: DL-J7-06 (J7/decisions.json).

DL-104: Propensity band cut-offs and the LOW-band posture

• Journey(s): J7
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Analytics (Ivi), Sheran ratifies
• Recommended decision date: 2026-07-31

Decision required: Define the propensity band cut-offs and the LOW-band posture (suppression is the treatment).

Context and supporting evidence: The intensity tiers (HIGH full episode, MID standard, LOW no triggered touch) need numeric cut-offs from the reactivation-propensity model, which has not landed. The RFM matrix has no cut-offs in source ('Analytics calibrates').

Recommended option: Analytics proposes cut-offs from the model's validation, anchored to the observed gradient (Lapsed PU 29.8% to Churned one-time 2.0%); LOW band receives waves and the anniversary beat only, and feeds the suppression-list deliverable.

Rationale: The 7-15x spread is the whole case for score-informed win-back; without ruled cut-offs the tiers are vibes.

Alternatives and trade-offs:

• Proxy tiers from frequency band alone until the model lands (usable interim, stated as proxy)

Dependencies:

• D2 propensity model landing

Risks:

• Cut-offs set post-hoc to flatter the read; fix them before launch

Consequence of delay: Cut-offs set post-hoc flatter the read, and the LOW-band suppression posture stays unbuilt.

Working assumption used in the journey pack(s): Cut-offs are proposed from the model's validation anchored to the observed gradient (Lapsed PU 29.8 percent to Churned one-time 2.0 percent); the LOW band receives waves and the anniversary beat only and feeds the suppression-list deliverable; fixed before launch.

Provenance: agent-derived DRAFT. Source ids: DL-J7-08 (J7/decisions.json).

DL-105: REC-14 offer depth, incentive budget and the HIGH-propensity gate

• Journey(s): J7
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Sheran + Finance on Analytics sizing
• Recommended decision date: 2026-07-31

Decision required: Set the REC-14 offer depth, incentive budget and the HIGH-propensity eligibility gate.

Context and supporting evidence: The final attempt concentrates a one-time strongest offer on the recoverable slice only (prior art: M61P £5 push, 3.9% open, one-off). If the model cannot isolate a slice with materially better-than-1% odds, the right call per the card is to park the incentive and let suppression run.

Recommended option: No incentive spend until Analytics sizes the HIGH slice; offer depth and £ floor are a Sheran + Finance call on that sizing; the go/no-go is explicitly allowed to be 'no offer, suppression only'.

Rationale: Funding a sub-1% lottery is the named failure mode; the suppression list has value even if the offer never ships.

Alternatives and trade-offs:

• Fixed £5 legacy offer to all deep-lapsed (repeats M61P)
• No REC-14 at all (loses the clean end-state and the suppression discipline)

Dependencies:

• DL-J7-08 band cut-offs
• REC-T01 read

Risks:

• Incentive leakage to organic returners if the gate slips; hard-code the propensity gate in canvas eligibility

Consequence of delay: Incentive spend commits before the HIGH slice is sized, and incentive leakage to organic returners begins the day the gate slips.

Working assumption used in the journey pack(s): No incentive spend until Analytics sizes the HIGH slice; offer depth and the GBP floor are a Sheran plus Finance call on that sizing; the go/no-go is explicitly allowed to be 'no offer, suppression only'; the propensity gate is hard-coded in canvas eligibility.

Provenance: agent-derived DRAFT. Source ids: DL-J7-09 (J7/decisions.json).

DL-115: S1 recognition slot mechanic: Braze Content Card now, native tiles later

• Journey(s): J10
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Sheran recommends, Product co-approves (in-app operating model)
• Recommended decision date: 2026-07-31

Decision required: Mechanic for the S1 recognition slot: native dynamic tile vs Braze Content Card. A named mechanic decision type: surfaced, not silently chosen.

Context and supporting evidence: Sources assume differently: the portfolio row calls dynamic marketing tiles 'the approved Connected-Content surface'; the mots ACT-12 master says Braze Content Card in the For-you carousel (CONFIRMED Braze-controlled, 26 Jun); gate1-sequence marks the tile surface Product-led with no code. GRW-01's not-armed content-card arc is already 'placement approved 7 Jul, NEXT via dynamic tiles'.

Recommended option: Q3: Braze Content Card in the existing For-you carousel (no release, CRM-controlled, A/B-able, zero Eng build). Target state: the Product native dynamic-tile framework replaces the card slot when it ships; the swap criteria (release cadence, A/B control, render-time data access) go to Product at scoping.

Rationale: The J8 mechanics-table logic: Braze = no release + A/B-able, native = Eng-built + release-bound. A DEFINE-stage journey should not gate its pilot on an unbuilt Product surface; the card mechanic already exists and the 9 Jul ruling pre-authorises exactly one governed personalised card.

Alternatives and trade-offs:

• Wait for the native tile framework (pilot slips beyond Q3; no A/B control at launch)
• Both at once (violates the one-card ruling and doubles the build)

Dependencies:

• Product commitment or timeline for the tile framework (scoping)
• D-J10-05 instrumentation fix (a Content Card pilot is unreadable without it)

Risks:

• If Product ships tiles mid-test, the surface changes under the holdout: sequence via the experiment calendar (D-J10-07)

Consequence of delay: The S1 slot has no mechanic to build against; or Product ships native tiles mid-test and the surface changes under the holdout.

Working assumption used in the journey pack(s): Q3 = Braze Content Card in the existing For-you carousel (no release, CRM-controlled, A/B-able, zero Eng build); the Product native dynamic-tile framework replaces the card slot at target state on the stated swap criteria.

Provenance: agent-derived DRAFT. Source ids: D-J10-02 (J10/decisions.json).

DL-116: Ratify the Home slot map and its arbitration order

• Journey(s): J10
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Sheran + Product (surface), CRM (treatment rules)
• Recommended decision date: 2026-07-31

Decision required: Ratify the Home slot map and its arbitration order (S0 trip state > open-episode journey_state > S1 recognition > S2 opt-in banner > S3 tiles).

Context and supporting evidence: Pieces are already ruled separately: at most one eligibility-rotated card (Sheran, 9 Jul); the mid-Home opt-in banner belongs to the channel-infrastructure journey (Sheran, 14 Jul, J8 pack); the card obeys in-app card priority not the CRM cap (mots master); L1 state overrides (backbone). No single artefact orders them, and the S1 slot must arbitrate between three journeys' treatments (J5 new-user, J7 welcome-back, J10 returning).

Recommended option: Adopt the DRAFT slot map in strategy.md 3b as the single ordering, with the arbitration rule: lifecycle state resolves the S1 treatment; an open save/win-back episode (journey_state) outranks recognition; slots never stack two interruptions.

Rationale: Without an ordered slot map, the 'at most one card' ruling is unenforceable across three treatment owners, and a drifting customer could see a cheery welcome while a J8 save episode is open (an arbitration bug, not a copy choice).

Alternatives and trade-offs:

• Per-journey bilateral dedupe agreements (n-squared rules, the connective-tissue gap the full-set review already flags)

Dependencies:

• D-J10-01
• journey_state exposure at render time (D1 pin)

Risks:

• Product may own parts of the render order natively: confirm at scoping rather than assuming Braze controls the stack

Consequence of delay: Slots stack or arbitrate by accident, and Product's native render order gets assumed rather than confirmed.

Working assumption used in the journey pack(s): The DRAFT slot map is the pack's single ordering (S0 trip state > open-episode journey_state > S1 recognition > S2 opt-in banner > S3 tiles; slots never stack two interruptions), pending confirmation at scoping that Braze controls the stack.

Provenance: agent-derived DRAFT. Source ids: D-J10-06 (J10/decisions.json).

DL-117: J10 measurement frame: ON/OFF holdout on a shared experiment calendar

• Journey(s): J10
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Analytics (Ivi) designs and sizes; Sheran approves; Product co-signs the calendar
• Recommended decision date: 2026-07-31

Decision required: Commit the measurement frame: return-session to transaction rate vs a clean ON/OFF holdout, sequenced on a shared experiment calendar with Product.

Context and supporting evidence: This journey is the exact class the evidence flags as session-only risk (+108% sessions / -12% purchases, confirmed) and the -11.8% blanket-return precedent shows mis-targeting is net-negative. Concurrent Product tests (GRO-1612 dynamic tile, Firebase banner tests) contaminate funnel holdout reads this quarter (hub RAID, confirmed). Cohort and baselines are unsized (TBC: Analytics).

Recommended option: One headline test: recognition ON vs OFF, primary metric return-session to transaction rate, sessions demoted to guardrail; power/VIP excluded; organic-return contamination handled in the cohort definition (Analytics decides the method, CRM writes the rules). No incremental-£ claim before the holdout reads: attributed, not incremental, until the long-term holdout is live (Gate-1 ruling 1). Agree the experiment calendar with Product before build.

Rationale: Every predecessor card (ACT-12, GRW-09, REC-21) independently reached the same conclusion: gate the revenue claim on a purchase-rate holdout. The framework inherits the strictest common denominator.

Alternatives and trade-offs:

• Leading indicators only if Analytics declares the test underpowered within Q3 (agree up front, never post-hoc)

Dependencies:

• D-J10-05 instrumentation (DL-013); the precise primary-KPI definition (which bid/booking window counts against a returning session) BEFORE sizing; Analytics sizing (baseline, MDE, arms, runtime: all TBC, Ivi); Experiment calendar with Product/Analytics

Risks:

• GRO-1612 shipping mid-test invalidates the read; an unsized cohort may not power a Q3 read at all

Consequence of delay: GRO-1612 shipping mid-test invalidates the read; an unsized cohort may not power a Q3 read at all; no incremental claim is possible before the holdout reads.

Working assumption used in the journey pack(s): One headline test: recognition ON vs OFF, primary metric return-session to transaction rate, sessions demoted to guardrail, power/VIP excluded; attributed-not-incremental language holds until the long-term holdout is live; the precise KPI definition precedes sizing.

Changed in the council round:

• J10 council (20 Jul 2026) changed dependencies. Why: An undefined KPI cannot be sized or powered; the council added the definition as an explicit precondition (Analytics work item, no new approval).

Provenance: agent-derived DRAFT. Source ids: D-J10-07 (J10/decisions.json).

---

Band 4: later and phase-2

DL-019: Canonical eng-requests source: standalone vs pack subset

• Journey(s): cross-journey
• Priority and urgency: P2
• Status: open
• Recommended decision-maker: Sheran (ratify)
• Recommended decision date: 2026-07-31

Decision required: Confirm the canonical eng-requests source: the standalone q3-eng-requests-v2 (40 asks, full fields) vs the pack's 29-item subset.

Context and supporting evidence: Open Sheran question from 7 Jul with a recorded recommendation (standalone canonical, pack links it). Since then the live Airtable eng board (tbl35yQjmMkhrnKek) has become the operational surface for sprints, which strengthens the standalone-canonical case.

Recommended option: Source's recommendation: standalone canonical with the pack linking it; record the Airtable eng board as the live home so the one-home-per-fact rule holds.

Rationale: The eng board is already where records are created and updated (Darwin feed, computed traits); the pack subset would drift.

Alternatives and trade-offs:

• Pack subset as canonical (drifts against the live board)

Consequence of delay: Two eng-request records drift apart and builds pick the wrong subset.

Working assumption used in the journey pack(s): The standalone q3-eng-requests-v2 (40 asks, full fields) is treated as canonical with the pack linking to it; the Airtable eng board is the live home, per the one-home-per-fact rule.

Provenance: agent-derived DRAFT. Source ids: DL-025 (decisions-baseline.json).

DL-023: Power-user home (roster decision a): a cross-journey suppression rule-set with the backbone, not Journey 11

• Journey(s): cross-journey, J1, J6
• Priority and urgency: P3
• Status: open
• Recommended decision-maker: Sheran
• Recommended decision date: 2026-08-14

Decision required: Close roster decision (a): the power-user win-moment rows (A3 suppression-is-the-treatment, ACT-24 A3) stay inside Journey 1 as the fifth segment overlay; the standing 'don't pester' power programme is recorded as a CROSS-JOURNEY SUPPRESSION RULE-SET owned with the backbone (written up as a backbone annex, explicitly assigned), not folded in as a fifth entry view and not created as Journey 11.

Context and supporting evidence: Architecture J9 (power repeat winner, 20.7k confirmed, blocked on the power-definition collision AN-4) has no home in the canonical 10. Its win-moment behaviour is already signed inside Journey 1's matrix (A3 rows); the rest of the programme is suppression rules, not sends. J6's power handling (JD6-09) depends on this ruling: if a positive recognition/VIP treatment is ever proposed, it is a Journey 6 candidate.

Recommended option: Record as stated; assign the backbone-annex write-up explicitly so the seg5 walk does not stay packless; revisit only if a positive power-user treatment is proposed.

Rationale: A journey with no sends is a rule-set; forcing it into the roster duplicates the backbone. Keeping A3 in Journey 1 preserves the one-transition model (a power win is still a win; the treatment is silence).

Alternatives and trade-offs:

• Journey 11 (no sends to walk)
• Fold into Journey 6 (only a future recognition programme would belong there)

Dependencies:

• AN-4 / BRZ-83 rfm_power_user flag (the rule-set is unbuildable until one definition is wired)

Risks:

• The seg5 walk remains packless until the annex is written; assign it explicitly

Consequence of delay: The seg5 walk remains packless and the standing don't-pester power programme has no written home.

Working assumption used in the journey pack(s): The power-user win-moment rows stay inside Journey 1 as the fifth segment overlay; the suppression rule-set is recorded as a backbone annex awaiting explicit assignment; no Journey 11 is created.

Merge note (v1): Merged: roster decision (a) + J1-D11 (which answers it). No conflict.

Provenance: agent-derived DRAFT. Source ids: ROSTER-a (ROSTER.md section 4 (a)); J1-D11 (J1/decisions.json).

DL-024: North-star metric per bet (O1-O5) or objective level

• Journey(s): cross-journey
• Priority and urgency: P3
• Status: open
• Recommended decision-maker: Data/Analytics with Sheran
• Recommended decision date: 2026-08-14

Decision required: Decide whether each bet (O1-O5) gets a named north-star metric (transacting-user rate / ARPU) or stays at objective level.

Context and supporting evidence: Open Sheran question from 7 Jul, unruled since. Gate-1 conditionally approved the five-bet portfolio; the measurement framing per bet was left open.

Recommended option: Agent-derived DRAFT: assign one north-star per bet at the next measurement review, drawing from the leading indicators already permitted interim, so bet-level reads exist before holdout results land.

Rationale: Bets without a named metric drift to activity reporting; the interim-indicators ruling makes this cheap to do now.

Alternatives and trade-offs:

• Stay at O1-O5 objective level until the holdout is live

Dependencies:

• DL-018 holdout structure shapes what each north-star can claim

Consequence of delay: Bet-level reads have no named north-star when holdout results land, and O1-O5 report inconsistently.

Working assumption used in the journey pack(s): One north-star per bet is assigned at the next measurement review, drawn from the leading indicators already permitted interim.

Provenance: agent-derived DRAFT. Source ids: DL-024 (decisions-baseline.json).

DL-025: Reconcile the refund/disruption operational figures

• Journey(s): cross-journey
• Priority and urgency: P3
• Status: open
• Recommended decision-maker: Analytics + CS
• Recommended decision date: 2026-08-14

Decision required: Reconcile the refund/disruption operational figures (4,836/4mo vs 17,416/12mo vs ~12,359).

Context and supporting evidence: Gate-1 ruling 8 (closed) marked these TBC (Analytics/CS) rather than quotable fact. The reconciliation itself has not happened; journeys touching disruption and refunds keep inheriting the TBC label.

Recommended option: Analytics/CS produce one reconciled figure set with method notes; until then every pack keeps the TBC label (do not upgrade).

Rationale: Ruled labelling holds; the open action is producing the number that retires the TBC.

Dependencies:

• CS data pull

Risks:

• Disruption-adjacent journeys sized on wrong volumes

Consequence of delay: Disruption-adjacent journeys stay sized on three conflicting volumes.

Working assumption used in the journey pack(s): Every pack keeps the TBC label on the refund/disruption figures; no figure is upgraded until the reconciled set lands.

Provenance: source-ruled follow-through (Gate-1 ruling 8). Source ids: DL-033 (decisions-baseline.json).

DL-027: Confirm FNL-AB / GRW-27 / OPS-AR / OPS-DR sit intentionally outside the 10

• Journey(s): cross-journey
• Priority and urgency: P3
• Status: open
• Recommended decision-maker: Sheran
• Recommended decision date: 2026-08-14

Decision required: Unassigned carded masters FNL-AB, GRW-27, OPS-AR, OPS-DR: confirm they sit intentionally outside the 10 journeys.

Context and supporting evidence: Carded but in no journey; they read as funnel-AB / ops governance cards.

Recommended option: Confirm and apply as stated; a hygiene ruling, one line from Sheran (plus the one-word hub edit for the '#' column).

Rationale: Roster completeness and numbering clarity; both are cheap now and confusing forever if left.

Consequence of delay: Four carded masters float unassigned and resurface in every roster review.

Working assumption used in the journey pack(s): FNL-AB, GRW-27, OPS-AR and OPS-DR are treated as intentionally outside the 10 journeys pending Sheran's one-line confirm.

Provenance: agent-derived DRAFT. Source ids: ROSTER.md section 4 (d).

DL-028: Relabel the hub roadmap's '#' column as 'priority'

• Journey(s): cross-journey
• Priority and urgency: P3
• Status: open
• Recommended decision-maker: Sheran
• Recommended decision date: 2026-08-14

Decision required: Relabel the hub roadmap table's '#' column as 'priority' in deploy/q3-crm-hub/index.html.

Context and supporting evidence: It reads as a journey number but is the priority rank; the ambiguity caused the J7/J8 numbering confusion.

Recommended option: Confirm and apply as stated; a hygiene ruling, one line from Sheran (plus the one-word hub edit for the '#' column).

Rationale: Roster completeness and numbering clarity; both are cheap now and confusing forever if left.

Consequence of delay: The hub column keeps being misread as a count rather than a priority.

Working assumption used in the journey pack(s): none adopted: a one-word hub edit awaiting Sheran's go.

Provenance: agent-derived DRAFT. Source ids: ROSTER.md section 4 (e).

DL-044: Win celebration screen: commit or defer the product build

• Journey(s): J1
• Priority and urgency: P3
• Status: open
• Recommended decision-maker: Product + Sheran
• Recommended decision date: 2026-08-14

Decision required: Commit or defer the product build of the win celebration screen.

Context and supporting evidence: Recorded in raid as a proposed build that does not exist today, gating the full GRW-01 celebration experience; interim anchors are in use so the journey can ship without it.

Recommended option: Agent-derived DRAFT: ship GRW-01 on the interim anchors for Q3 and put the celebration screen on the product phase-2 track, revisiting once the habit-canvas read exists.

Rationale: The journey is approved and buildable without the surface; committing product build before the first incremental read would spend product capacity on an unproven lift.

Alternatives and trade-offs:

• Commit the build now alongside the canvas

Dependencies:

• Product roadmap slot

Consequence of delay: None material this quarter; the product-slot question resurfaces at phase-2 planning.

Working assumption used in the journey pack(s): GRW-01 ships on the interim anchors for Q3; the celebration screen sits on the product phase-2 track, revisited once the habit-canvas read exists.

Provenance: agent-derived DRAFT. Source ids: DL-029 (decisions-baseline.json).

DL-057: Re-scope the GRW-17 Phase 2 Live Activity investigation

• Journey(s): J2
• Priority and urgency: P3
• Status: open
• Recommended decision-maker: Eng
• Recommended decision date: 2026-08-14

Decision required: Scope the GRW-17 Phase 2 Live Activity control-plane investigation against the updated LA facts.

Context and supporting evidence: Raid logs this as investigation-first, Later tier, covering iOS and Android. The 15 Jul findings changed the premise: Braze push-to-start exists but is API-only (webhook shape), the token-holder question is open (DL-033), and Android Live Updates are notifications so they cannot serve reach.

Recommended option: Re-scope the investigation to inherit the J8 webhook shape and the settled platform strategy (iOS reach, Android experience-only for push-opted users); keep it Later tier.

Rationale: The J8 LA work answers most of the original investigation; re-scoping avoids duplicating it.

Dependencies:

• DL-030
• DL-033

Consequence of delay: None material; the Later-tier investigation simply re-scopes when picked up.

Working assumption used in the journey pack(s): The investigation inherits the J8 webhook shape and the settled platform strategy (iOS reach, Android experience-only for push-opted users); it stays Later tier.

Provenance: agent-derived DRAFT. Source ids: DL-031 (decisions-baseline.json).

DL-067: ACT-05 follow-through: disambiguation cut, then the T&L entry

• Journey(s): J3
• Priority and urgency: P3
• Status: open
• Recommended decision-maker: Analytics + Product (friction slice)
• Recommended decision date: 2026-08-14

Decision required: Run the ACT-05 follow-through: disambiguate price-hesitation vs payment-friction, then scope the test on the recoverable slice.

Context and supporting evidence: Gate-1 ruling 3 (closed) scoped ACT-05 as a test with no fixed build: first split the ~10.3k/mo abandoners into payment-friction (Product-owned, not CRM-recoverable) vs price-hesitation, then test in-moment slide-up vs T+1h push vs both. The disambiguation and T&L entry are the open follow-through.

Recommended option: Per the ruling: Analytics runs the disambiguation cut, then the test design enters the T&L roadmap on the genuinely recoverable slice.

Rationale: The ruling is explicit; only execution remains.

Dependencies:

• Analytics capacity
• DL-020 T&L owner

Consequence of delay: The recoverable slice stays unsized and a test could chase payment friction CRM cannot fix.

Working assumption used in the journey pack(s): Analytics runs the disambiguation cut first; the test design enters the T&L roadmap only on the genuinely recoverable slice.

Provenance: source-ruled follow-through (Gate-1 ruling 3). Source ids: DL-032 (decisions-baseline.json).

DL-118: Resolve the home-station data question before the Unclassified fallback relies on it

• Journey(s): J10
• Priority and urgency: P3
• Status: open
• Recommended decision-maker: Product + Data (Liz/Sam raised it)
• Recommended decision date: 2026-08-14

Decision required: Resolve the home-station data question before it is relied on as the Unclassified fallback anchor.

Context and supporting evidence: The ACT-12 card's fallback chain uses home-station when no recent route exists, but the team openly questions whether home-station data is worth exposing beyond search ('is the value strong enough to bother?', Liz/Sam via Miro). Unclassified share is 1.7% on the Ivi persona model (3 Jul, confirmed); the ~44% figure was a stale caveat (cross-journey figure clash 15). The fallback path is an edge state (model unclassified plus low-confidence reads, threshold TBC: CRM + Analytics), not a material-volume launch path.

Recommended option: Product + Data answer the exposure question at scoping. If no: the Unclassified fallback anchors on most-recent route only, and the card renders nothing for customers with no route history (an honest empty state beats a guessed one).

Rationale: Building personalisation on an input its own team doubts violates the evidence standard; the fallback design changes depending on the answer, so it precedes copy authoring.

Alternatives and trade-offs:

• Expose it and A/B the anchor later (spends Eng effort ahead of the value question it is meant to answer)

Dependencies:

• Scoping session attendance from Product + Data

Risks:

• Unresolved, the neutral-card path (an edge state, not ~44%) launches on an unvalidated anchor; the exposure question is a data-quality call, not a launch-volume risk

Consequence of delay: The neutral-card fallback path (an edge state on the corrected 1.7 percent figure, not about 44 percent) launches on an unvalidated anchor; a data-quality call gets made by default.

Working assumption used in the journey pack(s): If exposure is not approved at scoping, the Unclassified fallback anchors on most-recent route only and the card renders an honest empty state for customers with no route history.

Changed in the council round:

• J10 council (20 Jul 2026) changed context. Why: CROSS-JOURNEY-REVIEW.md figure clash 15 directs the J10 author to adopt the supersession J6 already adopted (JD6-12); Analytics confirms the 1.7% at scoping.
• J10 council (20 Jul 2026) changed risks. Why: Same figure correction; the decision stands at P3 with its urgency case restated honestly.

Provenance: agent-derived DRAFT. Source ids: D-J10-08 (J10/decisions.json).

DL-026: Brand-voice / TOV source of truth (guardrails after the J8 voice test)

• Journey(s): cross-journey
• Priority and urgency: P3
• Status: open
• Recommended decision-maker: Brand/PMM track with Sheran
• Recommended decision date: 2026-08-31

Decision required: Settle the brand-voice / tone-of-voice source of truth (no standing doc exists; the frog guardrails path is now defined).

Context and supporting evidence: The 7 Jul question was whether a TOV doc exists or the pack establishes it. The J8 round has since ruled the path: the J8 frog-voice test is the evidence engine, guardrails are a cross-journey asset on the brand/PMM track, and pack copy is the working reference until the test reads. The doc itself still does not exist.

Recommended option: Per the J8-round rulings: hold pack copy as the working reference, author the TOV guardrails on the brand track once the J8 voice test reads; other journeys inherit the guardrails question and do not re-run the voice test without a reason.

Rationale: The 20 Jul rulings answered the mechanism; authoring the artefact after the evidence lands is the only open step.

Alternatives and trade-offs:

• Author a TOV doc now from pack copy (pre-empts the test it is meant to learn from)

Dependencies:

• J8 voice test read

Consequence of delay: Packs write copy without a voice source of truth, and the J8 voice-test learning has no home to land in.

Working assumption used in the journey pack(s): Pack copy is the working reference; the TOV guardrails are authored on the brand track once the J8 voice test reads; other journeys inherit the guardrails question and do not re-run the voice test without a reason.

Provenance: source-ruled follow-through (J8-round rulings). Source ids: DL-035 (decisions-baseline.json).

DL-124: Parameterise the frozen pack template (branch label/timing) and normalise 'unopened' to 'unseen'

• Journey(s): J8, cross-journey
• Priority and urgency: P3 (P3 (before the next pack-format iteration, not before J8 lock))
• Status: open
• Recommended decision-maker: Sheran (format owner); agent executes
• Recommended decision date: next journey-pack format pass

Decision required: Fix the frozen pack template before the next format pass: parameterise the live-rail branch label and timing (currently hard-coded to J8's '+5-7 days, if Touch 1 goes unopened') and normalise 'unopened' to the ruled 'unseen'.

Context and supporting evidence: The 20 Jul QA sweep recorded that every rollout pack without a J8-shaped second knock truncated its walkthrough to four beats because the template renders J8's branch label on any step 5+. Separately, Sheran ruled the unseen definition 20 Jul; the hard-coded label still says unopened, so the defect propagates the retired wording.

Recommended option: One build_pack.py change: branch label, timing string and step threshold become content.json slots with J8's values as defaults; sweep unopened to unseen in the same pass.

Rationale: Cheapest single fix that restores full-length walkthroughs to up to nine packs and stops a ruled-out term re-entering every future build.

Alternatives and trade-offs:

• Leave as-is and keep truncating walkthroughs: honest but loses the branch storytelling that made J8 the benchmark.

Customer and business implications: Internal quality only; no customer impact.

Dependencies:

• Template edits must round-trip byte-identical on J8 (the existing build_pack.py validation gate).

Consequence of delay: Nine packs keep four-beat walkthroughs; retired wording persists in the benchmark.

Working assumption used in the journey pack(s): J8's own walkthrough is correct for J8; the defect only bites on reuse.

Provenance: agent-derived DRAFT (council). Source ids: J8/decisions-delta.json (new, council round 20 Jul 2026).