← Q3 journey review packs

Executive recommendation: the Q3 lifecycle journey programme

1. Overall strategic strength

The honest verdict: this is a strong, coherent portfolio that is ready to be ruled on, not a finished machine ready to ship. Ten journeys now describe one customer timeline with no journey contradicting another (the council pass dissolved all six inter-pack conflicts to zero), every design choice traces to a labelled figure, and each journey carries one headline test with pre-agreed guardrails. What it is not, yet, is proven: only one send in the whole estate has a causal, incremental read behind it (the GRW-16 day-before family, +16.9%/+16.3% confirmed), the biggest denominators are still unreconciled (DL-009), and delivery runs through two capacity bottlenecks (Analytics and the unhired Braze developer). The strategy deserves approval; the value claims deserve the tests the packs have designed for them.

What makes it strong. Three things, in order of durability:

1. Evidence discipline. Every figure carries a confidence label, statements are classified on the six-tier evidence scale (confirmed through requires-approval), attributed revenue is never presented as incremental, and the packs state honestly what does not exist (missing events, unbuilt surfaces, consultations not held). The councils found no unsupported claims to strike in any pack.
2. Dissolved conflicts. Before the council pass, five customer windows were contested by more than one journey and roughly a dozen headline figures circulated in two versions. After it: every previously contested boundary carries one synthesised proposal, identical in both packs involved, awaiting a single Sheran ruling. Detail in PORTFOLIO-STRATEGIC-ASSESSMENT.md section 1.
3. One engine, not ten journeys. The journeys are windows over one customer timeline, connected by shared scores (the churn and propensity model family), one arbitration stack (journey ownership state, contact caps, the rule that transactional always outranks marketing), and named hand-off contracts that move a customer between owning journeys with the data the receiving journey personalises on.

Where it is thinnest. The evidence base is wide but shallow on causality: the portfolio's revenue history is attributed, not incremental, and two of the historical causal reads are negative (blanket recovery minus 11.8%, both confirmed), which the packs correctly treat as design constraints. The core pool sizes and baselines exist in two or three versions each until the Analytics definitions sprint lands. And six genuine end-to-end gaps remain where no authored or built thing exists (section 5). None of this argues against approval; it argues for the sequencing in section 6, which puts definitions and instrumentation ahead of every powered claim.

2. Journeys ready for approval vs requiring material revision

None of the ten requires material revision. That is the plain, load-bearing result of the council pass: every verdict is a form of approve-with-conditions, and no council found a strategy weakness that survives its conditions. What "conditional" means in practice: the councils already applied their improvements directly to nine packs (edited, rebuilt, validated; Journey 8's edits are proposals only because its pack is mid-review with the team), so the conditions are not rework. They are named Sheran rulings plus named data or engineering deliveries, all already in the log. Approving a journey at Gate-1 means signing its strategy and its headline test; nothing enters build until its listed gates clear.

3. Highest-priority decisions: the Band-1 stack

The log's Band 1 holds 16 items: rulings that shape more than one journey or sit against hard dates. They fall into four natural sessions, three of them already anchored to real dates. Full items live in the log; only ids, dates and the cost of delay are given here.

Session 1: the J8 walkthrough (Tue 22 Jul, recommended). Rules DL-029 (landing destinations, extended to a default-destination contract for every pack). Delay: the Braze build spec stays blocked even after lock, and destination debates multiply per pack. The same session should put a date on the D2 model delivery (section 5, risk 1). The lock itself then follows on DL-031 (numbers due EOP Mon 21 Jul) and DL-032 (per-team verdicts), both Band 2 but on the same clock.

Session 2: the J10 scoping session (Thu 24 Jul, recommended). Rules DL-114 (what Journey 10 IS). Delay: every other pack's landing table stays "to confirm" and DL-029-style decisions multiply per journey. Its companions DL-115 to DL-117 (mechanic, slot map, measurement frame) rule in the same room.

Session 3: the wave-1 package (to Sheran by 24 Jul; freeze ~Fri 25 Jul, assumed). Eight items ruled as one package because the Summer Leisure Wave is the portfolio's only hard external date: DL-125 (readiness gate and slip-vs-ship authority; delay: the date decides by default and wave 1 ships degraded), DL-108 (incentive posture; delay: a discount ships by default), DL-109 (claim standard; delay: creative drafts against two contradictory standards), DL-006 and DL-007 (the anniversary interim home and the decommission vehicle; delay: double-messaging burns the wave's one clean read), DL-008 (contact-cap dials; delay: opt-out spike risk in week one), DL-009 (the reconciled audience and baseline table, owner Ivi; delay: wave 1 is judged against a baseline nobody agreed), DL-010 (lapse threshold family; delay: the wave targets on an unreconciled overlay and its read is not reproducible).

Session 4: the arbitration sitting (target w/c 27 Jul, assumed; no ruled date exists). One sitting closes every formerly contested boundary: DL-005 (the paired J1 + J2 ratification; delay: J1's READY status stays misleading and J2 cannot go to review as a journey), DL-002 + DL-089 (the post-trip slot and J1/J6 boundary, one ruling for both packs; delay: the habit canvas and the J6 spine cannot both finalise), and DL-003 + DL-004 + DL-060 (the recovery one-pass, after Lisa's entry-criteria audit; delay: the estate's highest-attributed canvas keeps three verdicts and 13+ moments point at no recovery owner).

4. Most valuable opportunities

Five things in the portfolio are worth materially more than any single journey.

1. The shared home framework (J10). Every journey's deep links land on the app home surface, so ruling one state model, one slot map and one governance rule closes the landing question for all ten packs at once instead of ten times. It is also the only reach mechanism for the ~26% of active users who cannot be messaged on any channel (confirmed, surface-only read).
2. The cadence trait family. One warehouse build (the personal inter-transaction gap trait, DL-093, plus the positive_trip and savings traits) serves J6's nudge timing, J8's slip window and J2's value proof from a single definition. The evidence says persona never sets cadence (confirmed), so personal cadence is the portfolio's real personalisation lever, bought once.
3. The win-moment owner surface. The first win is the strategy's thesis lever, and the win-surface contracts (DL-079/081/085) guarantee one celebration and one ask at the customer's highest-trust moment, with the 31.8k/yr transacted-never-bid onramp (confirmed) dual-listed rather than double-built. Cheap to ratify, expensive to retrofit.
4. The decommission programme. The journeys replace a legacy estate whose only causal reads are negative (blanket recovery minus 11.8%, confirmed). A sequenced retirement runbook (home ruled at DL-007), with the constraint that retirements land before the replacing journey ships, converts every launch from an addition into a swap: contact load falls as capability rises.
5. The double-read measurement standard. J7's design separates journey lift (within-propensity-band holdouts) from targeting value (a model-vs-random read). Without it, a positive result could be entirely the model finding people who were coming back anyway. Adopted portfolio-wide, it is what lets the programme claim incrementality honestly, the exact thing the current estate cannot do.

5. Principal risks

1. The D2 model family is a four-journey single point of failure. J7 cannot ship, and J8 thresholds, J6's slip hand-off and J9's anniversary re-home all wait on one Analytics model whose brief was due end of Q2 and has not landed (confirmed per the packs). In place: J8's threshold subset is separately deliverable (DL-031), J6/J8 run a fixed 60-day fallback, J9 runs the anniversary interim. Proposed: date the delivery at the J8 walkthrough and keep "deliver or date it" as a standing agenda line.
2. Analytics concentration. The definitions sprint, all sizing, the model deliveries and the wave-1 power calculation route through one person (Ivi), who submitted no feedback in the J8 round (logged risk). In place: EVIDENCE-VALIDATION-PLAN.md packages everything as one prioritised commission, not ten ad-hoc requests. The risk stands until capacity is confirmed at commissioning.
3. The Braze hire gates build throughput. J7's build, decommission execution and general canvas throughput assume it; no date exists. In place: sequencing keeps gated journeys authored in parallel so no strategy work waits on the hire; the exposure is build dates, and it should be priced into any date Sheran commits.
4. Legacy estate discipline. Until the stop rulings and batch-audience exclusions (DL-012) land, every go-live stacks on the estate it replaces. In place: the retirements-land-first ordering constraint is carried in three packs and DL-012 is a launch precondition; ruling DL-007 gives the runbook a single owner.
5. The six end-to-end gaps (portfolio assessment section 2): the disruption service comm is unauthored while every journey yields to it (disruption is 64.2% negative sentiment on 12,359 tickets, confirmed); the power-user programme has no home (DL-023); the not-armed majority has no habit channel until the card placement and its instrumentation fix ship (DL-036 + DL-013); no-channel dormant customers have only unbuilt organic recovery; survey and referral infrastructure is broken where three journeys depend on it (a P0 link fix); and the opt-in banner moment needs an owner before J8 build QA (DL-123). Each has a named proposal; none has yet been actioned.
6. The unproven-incremental caveat. Everything except GRW-16 is attributed, and last-touch attribution is not incrementality. In place: the two-versions ban on figures, one headline test per journey, the holdout structure (DL-018) and the double-read standard above. The discipline to hold: no pack borrows GRW-16's figure, and no attributed number is ever presented as a result.

Also watched: the consent-classification precedent chain (DL-034 rules first, four send classes inherit it; ruled wrong it propagates four ways) and the process risk that figure labels and decision ids decay in consolidation passes, now countered by standing coherence rules (portfolio assessment section 5).

6. Recommended next steps and sequencing

Dates labelled (recommended) or (assumed) are council-proposed; the rest are in the log.

Fortnight 1, w/c 20 Jul and 27 Jul: rule.

• Mon 21 Jul: Analytics delivers the J8 numbers (DL-031, EOP). Sheran commissions the validation plan to the six owner groups as one package.
• Tue 22 Jul (recommended): J8 walkthrough; Sheran rules DL-029 and dates the D2 delivery. Per-team verdicts follow and Gate-1 lock is declared (DL-032); the Live Activity cell is held, not blocking.
• Thu 24 Jul (recommended): J10 scoping session; Sheran + Product rule DL-114 to DL-117 and harden DL-013 as launch-blocking. The wave-1 ruling package (section 3, session 3) goes to Sheran the same day.
• ~Fri 25 Jul (assumed): wave-1 freeze, only through the DL-125 readiness gate; Lisa + Analytics run the checklist; slip beats ship-degraded.
• w/c 27 Jul (assumed): the arbitration sitting (section 3, session 4) closes the paired J1/J2 ratification, the post-trip and recovery boundaries, and each pack's headline test. Lisa's M22_P entry-criteria audit lands before it.
• Fri 31 Jul: J8 ships, on the batch exclusions (DL-012) having landed first.

Fortnight 2, w/c 3 Aug and 10 Aug: clear the build gates.

• Eng/PMM close the close-event audit (DL-001); Lisa + compliance rule the consent cluster DL-034-first; Eng lands the card-instrumentation fix and Analytics verifies it in the warehouse (DL-013); J1 and J2 enter build as their gates clear.
• Analytics completes the definitions sprint core (the reconciled pool and baseline table, the J8 residuals, the return-rate restatement, the lapse thresholds: validation plan AN-1 to AN-4).
• J6 goes to team review and its six-ruling Gate-1 agenda; the J7 consent-age Legal read lands (DL-122, due 7 Aug); the decommission runbook starts executing under its ruled home.

Fortnight 3 onward, from w/c 17 Aug: build and read.

• J3 Phase 1, J4 and J5 enter build as their sizing and event gates clear; J9 wave 2 runs only on wave 1's logged read; the J10 recognition pilot starts once DL-013 is verified. J7 builds when D2 and the hire land, and not before. Band-4 hygiene items batch to one short session (dates 14 Aug in the log).

The standing operating rhythm thereafter, per journey and per fortnight: pack to its review circle (team review, the J8 pattern) then a changes round (edits folded in, figure labels travelling), per-team sign-off recorded, Gate-1 lock declared by Sheran; estate retirements land before each replacing journey ships; Analytics runs Test & Learn reads on the fortnightly cadence with Sheran arbitrating results interim (DL-020), and every read writes back to the Test & Learn roadmap before the next phase gate opens. Nothing ships, and no claim graduates past its label, without passing that loop.

Companions: PORTFOLIO-STRATEGIC-ASSESSMENT.md (the portfolio thesis and sequencing detail), DECISIONS-LOG.md (v2, every item in full), EVIDENCE-VALIDATION-PLAN.md (the validation commission), the ten J<n>/COUNCIL-ASSESSMENT.md files, ROSTER.md (canonical numbering). System of record: Airtable base appN7dd7fgbyJ8EYg; this document cites, never restates, its records.