For: the marketing team’s review of the Gate 1 strategy decisions; the rulings are Sheran’s. The one takeaway: all ten journey packs are rebuilt to one programme standard and live with their films, so each decision below can be reviewed with the full pack and film beside it. The decisions run in ship order, each with the recommendation, its evidence, and the strongest credible alternative. Every ruling is genuinely open: nothing here is written as decided.
Ruling these first makes several journey rulings one-liners.
| The decision | Recommendation, with its evidence | Strongest credible alternative | Where to look |
|---|---|---|---|
| The shared episode flag and broadcast stand-down. Commission journey_state as one shared convention: every owning journey’s canvas writes it; waves and batch sends read it. | Commission it now as the engine’s first governance build. It is already a launch gate on the first-slip journey and the wave engine, and every triggered journey’s holdout stays clean only if broadcasts stand down for owned episodes. | Per-journey ad-hoc exclusion segments: works once, then drifts, and every new journey renegotiates it. | Journey 9’s matrix (the stand-down row); Journey 8’s launch gates. |
| The welcome-back owner. The win-back journey fires and measures the dormant welcome-back; the returning-home journey renders the surface. One owner, decided. | Journey 7 owns the moment and its measurement; Journey 10 renders it as one state of the recognition slot, under a rendering contract (the shape the decisions log already sketches). Both packs deliberately carry the same question so one ruling settles both. | Journey 10 owns it outright: cleaner surface ownership, but the win-back loses its one phase 1 ship and its measurement home. | J7 pack, Decisioning matrix (the welcome-back row) · J10 pack, the same question from the other side. |
| The post-win slot arbitration (one of the four standing arbitration jobs). | Ratify: the pre-travel state owns the slot, and rebook-type touches floor at travel completion plus one day: the position the decisions log carries, which Journey 2 already builds on as a working assumption. | The win journey holds the slot to its own clock: simpler per journey, rebuilds the pile-up the slot rule exists to stop. | The decisions log’s post-win row; J2’s pack carries the working assumption. |
| The opt-in machinery’s owner. The opt-in banner has no owning journey or matrix row, and the ruled capture cooldown (30 days, two asks) needs extending ladder-wide as one shared decline counter. | Name the owner and extend the shared counter across every capture surface. Three uncoordinated permission asks measured on the never-transacted journey are the cost of leaving it unowned. | Leave capture per journey: every journey re-solves it, and the counters never share memory. | J5 pack, the permission ladder row (its master-record commission is below) · the opt-in banner rows on J8. |
| Two standing rules cited as policy but never ruled: power users suppressed from hesitation recovery, and blocked-on-data moments briefed as blocked, never built. | Give both one logged ruling each. Both are already applied across four packs; the ruling costs a minute and closes a provenance gap the log itself flags. | Leave as working rules: nothing breaks today, and the gap resurfaces at every audit. | Cited in the J3, J6, J7 and J10 packs. |
| The decision | Recommendation, with its evidence | Strongest credible alternative | Where to look |
|---|---|---|---|
| The loss-recovery moment’s verdict. The journey is complete through Gate 1, but its priority-one moment (loss recovery at the close) carries no verdict in the master. | Record the verdict the review effectively gave it (the close rework and day-three recovery are ruled and built), so the master matches the record. | Re-open it: only if the morning surfaces something the eight review rounds did not. | Pack: the close rows on the Decisioning matrix · film chapter at the close. |
| The headline-test approval. The day-3 gate holdout (gated recovery against a powered no-send control among day-3 non-returners) awaits the formal go. | Approve, sized by Analytics: the design is on the pack, the counting bases are reconciled, and the 46.2% do-nothing baseline makes an unheld read worthless. | Fold into a journey-level holdout: fewer cells, but the day-3 question is the one the model actually asks. | Pack: the Analytics brief’s headline-test card. |
Moments without a master verdict: loss recovery (the close), above. The other four (Live Activity, outbid recovery, the T-30 alert, knocked-out) carry verdicts; the two Refine verdicts’ refinements are tracked on the pack.
| The decision | Recommendation, with its evidence | Strongest credible alternative | Where to look |
|---|---|---|---|
| The launch date. CACI are building the journey in Braze now against the pack; the go-live note needs a date only Sheran can give. | Set it now, on the Tuesday check-in’s canvas state: the build is real, the entry data landed 27 Aug, and every surface already says launch date to confirm. | Hold for the pack-to-build reconciliation: safer, costs a week. | Pack masthead and CRM brief carry the to-confirm note. |
| The CACI thread reply (the newly-slipped entry condition and the £10 floor / £1,000 cap on the savings figure) sits drafted, awaiting send. | Send it: the drafted answer matches the ratified entry exactly, and the build waits on it. | Amend the floor and cap first if the review wants different guardrails. | The Journey 8 data-points email thread with CACI and Analytics. |
| Gate-1 lock. Per-team verdicts on the updated pack were never recorded. | Declare the lock at this review: the early lapse-save moment carries an Approve, the design response is approved, and Engineering’s validation found nothing to build. | Chase written per-team verdicts first: tidier, slower, and the teams already reviewed. | Pack: the process tab’s gate strip. |
| The decision | Recommendation, with its evidence | Strongest credible alternative | Where to look |
|---|---|---|---|
| Sheran’s review of the rebuilt pack, and the team share. The covering-note draft is held; the pack and sheets are live. | Walk the pack and release the share: the known-gaps note is drafted, and Design and CACI wait on the briefing surfaces it points to. | Hold the share until the survey instrumentation lands: it gates the read, not the spine. | The pack, end to end; the build-now sheet. |
| Power users’ place in the journey: suppressed from the reflection push and nudge, eligible for review and advocacy under their own caps. | Ratify as written: it is the recorded design, pending only this ruling. | Full suppression: simpler, forfeits the advocacy the power segment is best at. | Pack: the matrix’s power rows · film, the power-user run. |
| The positive-trip definition needs its operational sign-off (inputs queryable at fire time). | Commission the sign-off from Analytics this week: phase 1 runs the reduced gate, and the full definition is the phase 2 upgrade. | None credible: the reduced gate is already accepted and read by the holdout. | Pack: the Engineering brief’s gate rows. |
| The decision | Recommendation, with its evidence | Strongest credible alternative | Where to look |
|---|---|---|---|
| Ratify the recovery spine’s shape. The 28 Jul ruling gave this journey the pre-bid recovery; tonight’s pack authors it: deepest step stamps the episode, two touches then silence, day 7 hands back. | Ratify as authored: it reuses the ruled first-slip machinery (two-touch cap, unseen rule, ownership flag) instead of inventing new rules, and 13-plus moments across the estate stop referencing an unauthored owner. | Author the recovery per moment: the status quo in the old cards, and it guarantees stacking and contaminated reads. | Pack: the day-after group on the Decisioning matrix · film, step 5. |
| Day-one timing for the first recovery touch, superseding the one-hour working assumption. | Ratify day one: 31% of conversions arrive unaided after the first hour, and the 1-hour funnel converts 9.3% against 13.6% at 7 days, so same-hour contact pays for returns that would happen anyway. | A timing arm in the test if Analytics wants the hour settled empirically. | Pack: the recovery row’s why column · film, step 5’s evidence panel. |
| The power-user suppression as a standing rule for this journey (no hesitation treatment, no recovery; interim flag until the value-model re-cut). | Ratify: an imperfect suppression beats chasing the best customers with confusion copy, and the interim flag over-suppresses, which is the safe direction. | Wait for the recency-frequency-value re-cut: blocks all dwell recovery for a definition question. | Pack: the suppressed rows · film, the power-user run (nothing fires, by design). |
| The headline test: the bid-screen reassurance holdout, read as first-bid rate among never-bid stallers. | Ratify the pick; the sizing waits on Analytics reconciling the two step definitions (the 90-day and six-month reads put the biggest leak on different screens, and the fix follows the location). | Make the confirm-abandon work the headline: smaller cohort, and its ruled first phase is a diagnostic, not a send. | Pack: the measurement position on the Campaign brief. |
Moments without a master verdict, stated at their recorded positions in the pack: the bid-add stall, the confirmation-abandon test, the search-results value answer, the price-rejection exit, both dead ends, the alert engine, and the shared in-lead capture. The pack states each plainly; none is written as decided.
| The decision | Recommendation, with its evidence | Strongest credible alternative | Where to look |
|---|---|---|---|
| The win-surface ownership. One celebration per win event needs one owner per win type (auction won, instant win, auto-win, buy-now), with Product co-approving the in-app halves. | Rule the first-win frame as a copy branch on whichever surface owns each win type, never a third modal: the shape the decisions log sketches, and the cheapest build. | One new dedicated first-win screen: cleaner story, a Product build the evidence does not demand. | Pack: the first-win row · film, the final step. |
| Commission the permission ladder’s master record (asks, surfaces, copy, shared-memory rules in one signed design). | Commission it now: three uncoordinated asks for one permission set are live today, 56% of instant winners hold no push permission at the moment they win, and nothing on the ladder should ship piecemeal. | Fix the worst ask first (the cold system prompt): faster, and it recreates the pile-up one surface at a time. | Pack: the ladder row and its clash callout. |
| The two money questions: the £5 consent-incentive lead on the ladder, and the fresh-start offer for first losers the recovery missed. | Rule direction now, gate spend on Analytics sizing and the compliance view of consent incentives: both are proposals in the record, and both surfaces say so. | Rule both out: defensible on trust-first grounds; forfeits the one honest reach at the proven-intent cohort. | Pack: rows 5 and 7 on the matrix. |
| The education test gets decided (the standing rule: running tests get decided). | Direct Analytics to declare the winner or fold its control into the onramp holdout this week: the session-slot resequence and the headline test both wait on it. | None credible: leaving it running is the thing the standing rule exists to stop. | Pack: the session-slot row’s gate. |
Moments without a master verdict: the in-lead capture (shared), the always-on content card, and the dormant seasonal waves (the wave engine’s half lives on Journey 9). The push prompt and first-win activation carry Approve verdicts already.
| The decision | Recommendation, with its evidence | Strongest credible alternative | Where to look |
|---|---|---|---|
| Ratify the entry contract: live entry only by the day-60 hand-off, plus a one-time metered, propensity-ranked backfill. Never an independent day-count, never a blast. | Ratify as authored: it makes the win-back trigger-driven where every predecessor was calendar-driven, and the launch gate has shrunk to one sync payload extension. | Add a standing day-count entry as belt and braces: quietly rebuilds the calendar clock the design retires. | Pack: the entry row · film, step 1. |
| The live £5 card’s disposition: the territory’s only live message, unowned and unread. | Fold it into the welcome-back card’s offer state for the band that earns one, executed in the go-live change: one offer story, one surface, and the episode’s reads stay clean. | Retire it outright: cleanest, and it forfeits the one live touch until the new build ships. | Pack: the estate row · the as-is tab’s record. |
| The final attempt’s offer depth and band cut-off (with Finance, on Analytics sizing). | Rule the shape now (one offer, once, then hard suppression) and gate the depth on the sizing: the historic £5 push (3.9% open) is the cautionary read, not the model. | No offer at the end: honest, and the goodbye loses the one reason it converts anyone. | Pack: the final-attempt row. |
Moments without a master verdict: churned win-back, won-back final attempt, the reactivation moment, and the app re-open welcome-back (the shared ruling above). The pack states each at its recorded position.
| The decision | Recommendation, with its evidence | Strongest credible alternative | Where to look |
|---|---|---|---|
| The guardrail ruling. Drop 1 unsubscribed 0.71% in four days against a 0.28% benchmark. Does drop 2 (Edinburgh) send as designed, send changed, or hold? | Send changed: the softened, occasion-led tier frame in tonight’s pack is the response (the read says the current frame over-asks), sequenced after the York holdout read lands. | Hold the series entirely for the York read: safest, and it idles the calendar the engine is built on. | Pack: the wave row and the design brief’s softened frame · the as-is tab’s measured table. |
| The York read’s consequence rule. The series is gated on the holdout read: agree now what weak and strong mean. | Adopt the recorded gate: a weak read stops the remaining drops; a strong one triggers the targeting review. Deciding the rule before the read keeps the decision honest. | Read first, decide after: invites exactly the post-hoc reasoning the gate exists to prevent. | Pack: the measurement position · Analytics ask 1. |
| The sunset send for the roughly-twelve-months unengaged: one last email, then removal from every marketing audience. | Approve pending the Legal read (shared with the win-back journey’s deep-tail rule): 634,504 churned customers already hold no channel, and the consented tail decays toward them. | Quiet removal without the goodbye: cheaper, and it wastes the one honest touch the tail would ever notice. | Pack: the sunset row · film, the unengaged state. |
Moments without a master verdict: the dormant seasonal first-win waves and the always-on consolidation (the old midweek-plus-weekly family, now the wave engine’s job). The pack states both at their recorded positions.
| The decision | Recommendation, with its evidence | Strongest credible alternative | Where to look |
|---|---|---|---|
| Commission the Product scoping session. This journey sits at the earliest gate of the ten: nothing is signed, and the pack is the proposal that goes into the session. | Commission it this week, with the pack as the pre-read: the session decides the state model, the slot map, the rail’s one occupant (Product’s own tile experiment runs on the same real estate), and the card-against-native split. | Ship the Braze card without the session: it fits phase 1, and it spends the one slot before the slot map exists. | The whole pack, briefly: it is one proposal. |
| The Home slot map and its arbitration order (trip state, then open episode, then recognition, then the opt-in banner, then tiles). | Ratify the order as the standing home-surface rule: it is the same one-voice-per-moment logic every journey’s stand-down uses, applied at render. | Leave ordering to per-card priority values: how the current eleven-card pile happened. | Pack: the state-resolve row and the map’s slot frame. |
| The recognition slot’s mechanic: native dynamic tile against Braze content card. A named mechanic decision type: surfaced, never silently chosen. | The Braze card for phase 1, the native rail at the Product build: CRM-controlled and testable now, the richer surface when the render-path data exists. | Hold everything native: one coherent build, nothing testable this quarter. | Pack: the recognition-card row’s mechanic note. |
Moments without a master verdict: the returning-user personalised landing (this journey’s own moment, parked Q4 in the master but proposed here for the scoping session) and the shared app re-open welcome-back.
| The decision | Recommendation, with its evidence | Strongest credible alternative | Where to look |
|---|---|---|---|
| The win-to-second-win moment’s verdict. The programme’s thesis moment (the first-to-second-transaction cliff: month-on-month retention falls from 20.15% to 7.47% at that boundary) carries no verdict in the master. | Rule it at this review: the confirmed job line (first-to-second-transaction) and the pack’s ruled beats are ready to sign; the cliff is the programme’s own sizing evidence. | Defer to a later session: nothing else waits on it, but the programme’s thesis moment deserves its ruling first. | Pack: the win-moment rows · film, the win chapter. |
| The win-moment headline test: the one-job rework (realised-value-led, referral stripped, single next action) against today’s five-ask send. | Approve the pick; Analytics designs. The five competing asks on today’s send are the evidence. | Test the day-after habit-touch timing first: the smaller lever. | Pack: the headline-test row. |
| The transactional-classed win receipt doing marketing work needs its privacy-regulations class ruling before the rework ships. | Commission the compliance read now (with the consent-forked CTA block as the exhibit) so the rework does not stall at build. | Strip the marketing work from the receipt: compliant by construction, and the moment’s best real estate goes unused. | The decisions log’s compliance row; the pack’s win-receipt row. |
Also queued from the log: the auction-termination event contract audit (does an instant win emit the shared won event), and the win-celebration screen’s commit-or-defer with Product.
| The decision | Recommendation, with its evidence | Strongest credible alternative | Where to look |
|---|---|---|---|
| The 22-month ride-timing test gets decided (the standing rule: running tests get decided). | Direct the decision on the read Analytics holds: the rule exists for exactly this test. | None credible. | Pack: the pre-travel window rows. |
| The pre-travel headline test: the reassurance-reframe A/B, one interaction design across the three cells that can hit one trip. | Confirm the pick and commission the single design: three uncoordinated cells on one trip is the risk the design closes. | Run the day-before holdout alone first: cleaner, slower. | Pack: the measurement rows. |
| The unconsented day-before email leg: suppress or shift. | Shift to the consent-free in-app surface for the unconsented slice, keeping the email for the consented: reach without the compliance exposure. | Suppress outright: cleanest, and the anxiety-peak moment goes unanswered for the slice that needs it most. | Pack: the day-before rows. |
Moments without a master verdict: the day-before anxiety peak, morning-of, the mid-window beat, and the travel-day experience. The Darwin-feed posture, the seat-swap boundary and the travel-day modal scope are queued log rulings behind them.
What the master’s Gate 1 field holds today: 8 verdicts across 61 moments (four on Journey 4’s auction moments, the first-slip save, the instant-win confirmation, the push prompt, and first-win activation). 26 work-now moments carry none. Two record-hygiene fixes owed: write Journey 6’s 30 Aug rulings into the field (the three spine moments plus the review prompt’s consider-testing ruling), and write Journey 4’s close-moment verdict. Every other blank is a genuine open ruling listed above, journey by journey. Parked, operational and Q4 moments sit outside Gate 1’s scope and are deliberately not listed.
Sources: the moments master pulled 2 Sep; the decisions log; each pack’s question register. This page decides nothing: it compiles. Every pack and film links to its live page.