← Q3 journey review packs

Journey 9 council assessment: Dormant, seasonal reactivation waves (Layer 2)

Verdict

CONDITIONAL APPROVE (agent-proposed DRAFT). The problem, purpose and outcome are clearly defined; the AS-IS is honest (a graveyard plus two ruled-stop broadcasts); the TO-BE is deliverable and CRM-only; the evidence-to-choice chain is strong and correctly labelled throughout. The design already refuses the three classic reactivation mistakes: it never sums the two pools, never claims persona predicts propensity (7.1% vs 7.3%, confirmed), and never lets attributed revenue stand in for incrementality. The conditions: the wave-1 ruling package (claim DL-109, incentive DL-108, REC-06 vehicle DL-007, anniversary home DL-006, cap class C DL-008, threshold family DL-010) and the canonical baseline plus power calculation (DL-009) before the freeze. This is the only journey with a hard external calendar; the readiness gate below governs it.

Strongest points (keep, do not dilute)

1. Pool honesty: the 296.3k churned-transacted pool carries the money case (£219.94/reactivated, confirmed); the 820k never-transacted tier is reach, never a revenue line; wave arithmetic on the 820k is banned in the pack.
2. Suppression-first design: triggered outranks scheduled, the -11.8% power-user lesson is a hard exclusion, and the journey_state batch-read mandate is applied to this journey before wave 1, not just the triggered ones.
3. The REC-06 re-scope executes Sheran's 14 Jul STOP ruling instead of letting the Gate-1 circle sign a card the ruling already overruled.
4. One headline test with a pre-agreed decision rule: wave 1 vs its no-contact holdout is the go/no-go funding read for the whole calendar.
5. The campaign coupling is explicit and one-directional: the campaign reuses Q3 segmentation, contact rules and measurement; wave reads write back to the Test & Learn roadmap.

Changes applied (each: what · why/evidence · who · where · capability · outcome · action needed)

1. Beginner's Luck copy conformed to the substantiation rule. What: "Most first-timers win their first auction" replaced with "7 in 10 first-time bidders win their first auction" in all three draft-copy instances plus the wave-grid guardrail note. Why: 68.6% (confirmed) is the win rate among COMPLETED first auctions, a self-selected denominator; "first-timers" over-claims to the cold pool and carries ASA/CAP exposure (DL-082, the J5 rule). Who: audience A (Cooling/Cold never-transacted). Where: T1 sends, both frames. Capability: none new. Outcome: regulatory safety and honest activation framing. Action: none; conforms to the standing rule (claims owner naming still open, DL-082).
2. Wave-1 readiness gate added (new decision J9-DC01). What: wave 1 freezes only when BRZ-83+BRZ-84, the journey_state batch read, the baseline/holdout/power calc, and claim-and-incentive-ruled Gate-2 creative are ALL true; otherwise the wave slips; sole de-scope = Frame-A-only if the D8 anchor feed alone is missing. Why: an untiered or unmeasured send recreates the estate this engine replaces and burns the one clean funding read; the hard late-July date makes ship-degraded the default failure mode. Who: every wave-1 customer plus the holdout. Where: entry, before the freeze. Capability: DL-012, DL-009, BRZ-83/84. Outcome: a readable go/no-go instead of a blast. Action: REQUIRES APPROVAL, Sheran (slip-vs-ship moves the campaign calendar).
3. Fourteen-field personalisation and contact completeness table added. What: the standard's fourteen fields stated explicitly on the Reference tab with a summary in strategy.md; marketing/service/operational/product contact reviewed together (L0 always outranks; product owns post-return; no service contact fires on this pool). Why: the standard requires all fourteen explicitly; the pack held them in substance but not as the checklist. Who: all treated customers. Where: Reference tab. Capability: none new. Outcome: reviewable completeness. Action: none.
4. Stale cross-pack claims corrected. What: "Journey 7's author has not seen this split" replaced in five places with the recorded state: the J7 pack exists, the boundary rule is AGREED across J5/J7/J9, and two conflicts are logged for Sheran (anniversary interim home DL-006; REC-06 vehicle DL-007, where the cross-journey review also proposes a decommission programme outside both journeys); REC-21 wording upgraded from "unowned beat" to the three-pack convergence on J7 (DL-014). Why: the consolidated log and cross-journey review (both 20 Jul) supersede the pack's authoring-time premise; presenting a resolved-shape disagreement as unseen would misdirect the review. Who: churned-transacted (anniversary) and post-return customers. Where: cross-journey boundaries, exits. Capability: none. Outcome: the Gate-1 circle reviews the real decision state. Action: Sheran rules DL-006 and DL-007; ratifies DL-014.
5. Decision-ID mapping to the consolidated log added (footer, Reference tab, strategy.md status): J9-D01 to D12 mapped to their DL homes so every item has one home and reviewers land on the live record. Why: one home per fact; the log renumbered on merge. Action: none.
6. Map consistency fix: the TO-BE map's audience-A second touch now says Cooling-only (Cold takes a single touch), matching the wave grid. Why: the map is the build spec; a mismatch ships as a bug. Action: none.

Working assumptions adopted (labelled, with what would break them)

• Anniversary slot runs under wave governance with per-cycle mutual exclusion until D2 lands (the DL-006 synthesis); breaks if Sheran rules for J7's interim ownership.
• Frequency band (one-time/multi/power) is the audience-B priority axis; persona overlay is message fit, not propensity (DL-110, closed to the pack; Gate-1 sign-off of the pack ratifies the grid); breaks if the D11 trait cannot be built for wave 1 (falls back to persona × LTV as carded).
• Sunset-send posture: ~12m working threshold pending the Analytics definition; the 820k tier stays capped at a partial test behind the Cooling read and deliverability guardrails (DL-112, closed conservative).
• Write-back loop: Ivi logs holdout reads, CRM logs guardrails, no wave 2+ without wave 1's logged read (DL-113, closed as operating cadence).
• Cap class C ~3/week remains ASSUMED (DL-008 must ratify before wave 1); the wave T2 yields to the campaign's weekly drop if the worst-case arithmetic does not fit.

What genuinely requires approval (Sheran, before the freeze)

1. The wave-1 claim standard: substantiated anchor leads CRM; "up to 60%" market-level only, post-substantiation (DL-109).
2. The incentive posture: no monetary incentive in wave-1 CRM; strike or ratify the ACT-31 master's £5 line (DL-108).
3. The two recorded J7/J9 conflicts: anniversary interim home (DL-006) and REC-06 decommission vehicle (DL-007).
4. Cap class C ratification plus the worst-case contact arithmetic (DL-008); the threshold-family reconciliation incl. BRZ-84 (DL-010).
5. The readiness gate itself and slip-vs-ship authority (new item J9-DC01); REC-21 to J7 (DL-014, an easy ratification, three packs concur).

Evidence gaps (REQUIRES VALIDATION)

• Top gap: no canonical baseline, pool definition or tier-level sizing (DL-009; INITIATIVE SIZING row 22, Analytics/Ivi). Three baselines circulate (0.84% organic, 0.24% legacy KPI, the ~1% ceiling) on at least two pool definitions, exactly where the funding case sits; addressable opt-in for the 296.3k pool is unknown; and the power calculation at a ~1% response ceiling may show wave 1 is underpowered before it is designed. Nothing is committable without this.
• The M40-M42 WIN-BACK FINAL sequence's true live state (re-audit, CRM/Lisa; its £5 incentive contradicts the proposed posture).
• The "up to 60%" ASA substantiation vs walk-up First fares (open campaign action) before any CRM secondary use.
• The unengaged-tail threshold definition (Analytics) behind the sunset send.

Log deltas: decisions-delta.json beside this file (1 new, 1 changed, 4 closed). Pack rebuilt and validated PASS, 20 Jul 2026.