← Q3 journey review packs

Journey 8 (Active user: first slip, early lapse-save, DEF-20): council assessment

Verdict: LOCK CONDITIONAL

RECOMMENDATION: declare Gate-1 lock once (a) the D1 per-persona trigger thresholds land (the single launch-gating Analytics item in DL-031, due EOP 21 Jul), (b) landing destinations are ruled at the team walkthrough (DL-029), and (c) per-team verdicts are recorded (DL-032). The LA consent ruling and token-holder answer (DL-030/DL-033) gate only the Live Activity test cell, not the core journey: lock without them and hold the LA cell. All 8 feedback records are answered (5 Applied, 1 Declined, 2 awaiting Analytics numbers); no council finding reopens any ruling.

Completeness check: all fourteen personalisation and contact fields are STATED in the pack. Strongest: eligibility/suppression (ten entry checks, suppression-as-treatment matrix rows), trigger (D1/D2 with fallback), data/tech (D1-D7 pins with kind, owner, status). Stated but resting on opens: contact priority (backbone cap dials pending ratify), consent (LA ruling open, honestly flagged), measurement (holdout size and MDE with Analytics). None missing.

Strongest points (keep these in every rollout pack)

• AS-IS honesty: the estate audit on the surface (LAPSING50D measured minus 10.2% vs control, verdict Stop; the working M62 switched off anyway; broadcasts with no holdout). EVIDENCE-BACKED INTERPRETATION done right.
• Ruled definitions embedded with provenance and dates (unseen, usual_route, slip episode, all "ruled 20 Jul, Sheran"); agent-drafted copy labelled as such.
• The data taxonomy (computed trait / journey-written / live) with ownership on every pin, and the journey-by-journey Analytics brief panel.
• Test discipline: one headline test (voice A/B), phase-2 candidates explicitly gated, per-variant guardrails, a decision rule to agree before launch, and a qual twin on a brand bet.
• One home per fact throughout; suppressed audiences named with the journey that owns them instead.

Proposed edits (all seven fields; each is proposed, not applied: pack is mid-review)

1. Unseen wording sweep. What: replace residual "unopened" with the ruled "unseen" (live-rail branch label and LTIME strings, the walk's trace line, the matrix commuter drop-off cell "No open/tap"). Why: Sheran ruled the unseen definition 20 Jul and retired non-opener wording; the live tab still contradicts it. Who: every T2-gated customer (build QA reads this). Where: Live tab rails plus matrix WHEN/drop-off cells. Requires: copy edit only. Improves: correct T2 gate build, no wrong canvas condition. Action: CRM/agent applies in the changes round on Sheran's go. [proposed, not applied: pack is mid-review]
2. Status cell vocabulary. What: reword the exec-summary Status "Strategy APPROVED (Gate 1, clean)" to "DEF-20 card approved; Gate-1 lock pending per-team sign-off (this review)". Why: it currently contradicts the Process tab (YOU ARE HERE at step 5, lock at step 6) and DL-032. Who: every reviewer judging what is still open. Where: Campaign brief exec summary. Requires: copy edit. Improves: no false "already locked" reading. Action: same changes round. [proposed, not applied]
3. Consent-total caveat restored. What: re-attach one line to the matrix caption's 47,836 / 18,477 / 11,327 totals: sourced from the CI schema with purchase-based recency as the activity proxy, Analytics to confirm as official sizing. Why: the caveat existed (33rd pilot pass, 14 Jul) and was lost when the permission table was deleted (v56); the caption now reads as clean confirmed. Who: anyone sizing cells or caps from it. Where: matrix caption, Campaign brief. Requires: copy edit; the confirm rides DL-031. Improves: honest-figures rule holds. Action: changes round. [proposed, not applied]
4. One label for the 20% vs 0.5% figure. What: carry the T&L label (assumed, industry benchmark, Analytics validating) into the exec summary bet, the Quant grid, Objective/KPI, and the AS-IS churn-exit caption, where it currently reads as fact. Why: the pack's most load-bearing economic claim has one label in one home and none in three. Who: Gate-1 signatories. Where: four spots, Campaign brief plus AS-IS map. Requires: copy edit; cohort validation added to DL-031 (see delta). Improves: figure-label integrity. Action: changes round. [proposed, not applied]
5. Delete the stale content-card flag. What: remove "FLAG: the content-card leg inherits the Braze instrumentation gap" from the NOT-push-opted matrix row. Why: the content-card save leg was dropped when Sheran ruled banner-only (14 Jul); the flag re-imports a dependency this journey no longer has (DL-013 keeps it where it belongs, J10/J1). Who: no-channel segment builders. Where: matrix suppression cell. Requires: copy edit. Improves: no phantom build dependency. Action: changes round. [proposed, not applied]

Also worth folding in when edits reopen: a one-line note on the landing register that "the personalised Home screen" is the Journey 10 framework surface (DEFINE stage, per DL-029), so the walkthrough rules destinations knowing the surface itself is unowned today.

Working assumptions (labelled, with what confirms or breaks each)

• WORKING ASSUMPTION: commuter ~14d slip threshold; leisure score-led later. Confirmed/broken by Analytics cadence validation (rides the journey, DL-031).
• WORKING ASSUMPTION: save-at-slip ~20% vs ~0.5% churned (industry benchmark). Confirmed/broken by the own-cohort read (delta to DL-031).
• WORKING ASSUMPTION: the eight permission combinations collapse to five treatment states; the LA reachability rule (LA only where nothing else is held) is a working rule for ratification, stated as such in the pack.
• WORKING ASSUMPTION (council, on the M44/M45 stop): treated as proceeding per the pack's recommendation; it still needs Sheran's confirm and the Braze audience verification listed in the change table.

What genuinely requires approval

Per-team verdicts and the Gate-1 lock itself (Sheran, DL-032) · landing destinations (team walkthrough, DL-029) · LA consent position and webhook start shape (Sheran plus CRM/Eng, DL-030; token holder DL-033) · the M44/M45 stop and the reachability rule ratification (Sheran). Nothing else on the pack needs formal authority; the rest is decided strategy or Analytics delivery.

Evidence gaps (REQUIRES VALIDATION)

1. Top gap: the DL-031 bundle, due EOP 21 Jul, specifically the per-persona D1 trigger thresholds (the one launch-gating number) plus holdout size and MDE. Method: Ivi's warehouse run against the production churn model. Note the logged risk: Analytics submitted no feedback in the round.
2. Unscored-group count: customers the model cannot score get nothing until day 60 by design; the size of that blind spot is unknown (fairness as well as coverage).
3. Episode volumes and persona splits (classifier BRZ-83 not yet in the warehouse); VIP cell power.
4. Benchmark cohort validation (edit 4 above).

Template-level findings (the format the other nine packs inherit)

• Parameterise the live-rail branch label and timing. The frozen template hard-codes "+5-7 days, if Touch 1 goes unopened" on step 5+; every rollout pack without a J8-shaped second knock truncated its walkthrough to four beats to avoid false timing. Fix in build_pack.py before the next format pass, and normalise unopened to unseen while in there.
• Figure labels must travel with every restatement. The 20%/0.5% case (edit 4) and the consent-total caveat loss (edit 3) are the same defect: consolidation passes drop labels. Rule for rollout: a figure's confidence label is part of the figure.
• Status vocabulary. Distinguish "moment card approved" from "journey pack Gate-1 lock" in the exec-summary Status cell; the two-gate ambiguity will recur in all nine packs.
• Decision-id hygiene. CROSS-JOURNEY-REVIEW.md cites baseline ids (DL-001 landing, DL-002/003 LA, DL-004 churn family) that now resolve to different items in the consolidated DECISIONS-LOG.json (correct: DL-029, DL-030/033, and the churn-family rows). One renumber sweep, or a "baseline ids" footnote, before the docs circulate together.
• Carry forward as strengths, not fixes: suppression-as-treatment matrix rows, the ruled-definition provenance pattern, and the Analytics-brief-in-one-place panel.