← StrategyJourney 8, defendWorking draft

The churn-prevention journey

The churn-risk model scores every active customer nightly on their likelihood to lapse; the first sign of a slip starts the journey the next day.

This journey’s strategy
There is no subscription to cancel: a customer who is leaving simply goes quiet, and once lapsed most are gone for good. The churn-risk model spots the first sign of a slip, so we step in with one personalised reason to come back while they are still cheap to keep.
  1. One window we own: the 60 days from the first sign of a slip to the hand-off to win-back, where today nothing happens until a generic email at day 50.
  2. The model decides who enters and when, never a calendar timer; a customer it has not flagged is sent nothing.
  3. Two touches, then rest: one personalised reason to come back on their usual route, a second only if the first went unseen, then silence until the hand-off.

The problem  Nothing watches an active customer for the first 49 days of a slip, and the one thing we then send makes it worse.

Why it happens Lapse is never announced. The record shows two drivers: missed auctions (2,310 missing-notification mentions) and eroded belief that bidding is still worth it (price perception, 1,952 mentions, 66% negative), compounded by win value that varies trip to trip. Some of it is a decision, some of it is drift, and none of it reaches us as a signal.

The behaviour we are driving They come back and transact within the 60-day window. Any transaction, an auction, a buy-now or an instant win, ends the episode that instant.

The job to be done  “Give me a reason before I forget you exist: one useful message, one charming nudge, then respect.”

How entry is decided

The churn-risk model re-scores every active customer's behaviour nightly; a shift in their churn band, never a calendar timer, is what opens the journey.

Sessions, searches, bids, tripsChurn-risk model, re-scored nightlyFirst sign of a slip: the save, next dayReturn, or day 60: win-back

The numbers that decide

1,841the audience: established actives have entered a save episode
TBCthis journey's metric: Slip-to-return rate; the baseline is sized with Analytics
£380,118 a yearnet to Seatfrog; £1,652,685 gross. Upgrades only, measured behind a 20% control

Q4 marketing model, Business Case tab, exported 29 Sep 2026: volumes checked by Analytics 8 Sep 2026; organic and target rates measured against holdout and confirmed by Analytics; a buyer's month valued at £30 an upgrade, 1.29 upgrades in a first-purchase month and 1.67 after; net is gross at the 23% take rate.

The business case and sizing · £380.1k net a year (£1652.7k gross)
Model rowWorked volume a monthRateValue per active monthGross a yearNet a yearQ4, drivenQ4, in quarter
active buyers at risk (64.7k marketable, Sept)
DEFEND
11,194
About 11,000 buyers enter the flow each month on the band state change (Braze counts; Michael confirmed 10 Sep these are monthly volumes). Re-entries count as new episodes, which is how the built journey works.
27.3% vs 21% organic£30 an active month (one upgrade) × 6.82 months × 12 a year£1,385,119£318,577£80,299£23,356
session active users at risk
DEFEND
9,082
About 9,000 session users enter the flow each month on the band state change (Braze counts). Re-entries count as new episodes.
6.5% vs 5% organic£30 an active month (one upgrade) × 6.82 months × 12 a year£267,566£61,540£15,511£4,512

Q4 2026: £95,810 net on the driven basis (Oct £32,284, Nov £31,243, Dec £32,284); £27,868 landing inside the quarter. Driven: months active credited to Q4, pro rata by days live from the launch, the basis the Q4 V2MOM carries. In quarter: only the months that fall in October to December (Tony, 10 Sep 2026: the forecast Marketing submits). Net is gross at the 23% take rate. Launch 1 October 2026.

Still to confirm The at-risk segment is being redefined (item 19): the churn score cannot flag active buyers. The 61 to 90 day segment's rates were measured 8 Sep, rough until the holdouts mature (item 6). Two segments, below; the churn-risk score is live

Ladders to Reactivation (Defend): month-three retention 18.0%. Metric moved M3 retention %. Evidence B, benchmark.

The planned case, a 30% lift in conversion, as presented to the board on 23 Sep 2026 (the Sizing tab behind the model's Business Case tab): volumes checked by Analytics (Ivi, 8 Sep 2026); organic and target rates measured against holdout and confirmed by Analytics; a buyer's month valued at £30 an upgrade, 1.29 upgrades in a first-purchase month and 1.67 in later months (the 20 Sep change; the earlier £172k counted one upgrade a month); net is gross at the 23% take rate (agreed 8 Sep 2026). Source: Q4 2026 Marketing Model, Business Case (output) and Sizing tabs, exported 2026-09-29 (the live model). Every figure is incremental of the customers who would have returned anyway: worked volume × (target rate − organic rate) × months active × times a year × value per active month.

The journey, today and proposed

channels down, moments across; a card keeps its place between Today and Proposed, so the change is the only thing that moves. Open a card for the words, the trigger and who builds it; filter by builder and outcome. Opens on the returning commuter, still quiet. Every live send, listed here.
State
Show for
Outcome
On Proposed (Today is the picture before any ruling):Reworked, replaces the live sends it namesNewRetires, struck through in placeContinues (grey: not yet ruled)Waits for Phase 2Engineering fix, nothing to writeApp engineeringBackend engineeringBrazeAnalyticsDesign

What changes

one row per moment: what ships now, what waits for engineering, and who builds it, with the change type in the engineering board’s words

Each engineering-row chip is coloured by the ask: APP AND JOURNEY OWNED COMPUTED TRAIT · ANALYTICS READY · NOTHING TO BRIEF

The momentPhase 1, ships nowPhase 2, waits on engineeringWho builds it
on the map: 1 · The save message
Touch 1, the save email

email The generic day-50 email becomes a personalised save: their usual route and a true from-price where held, the recognition line for power users, and the generic…

Engineering rows D1D3D4D6

Nothing waits

Braze · Braze buildAnalytics · computed traitDesign · copy and creative
on the map: 1 · The save message
Touch 1, the save push

push The generic day-50 push becomes the same personalised save in one glanceable line, deep-linked to their route.

Engineering rows D1D3D6

Nothing waits

Braze · Braze buildAnalytics · computed traitDesign · copy only
on the map: 2 · The second knock
Touch 2, the second knock

push, email The old canvas’s week-two repeat, same content again to anyone without a session, becomes one channel-switched knock gated on unseen, with the episode closing after it…

Engineering rows D3D7

Nothing waits

Braze · logic changeDesign · copy and creative
on the map: 3 · Designed silence, against today's canvas
The estate around the episode

The seasonal weekly stands down for open save episodes.

Engineering rows D6D7

Nothing waits

Braze · logic change
No usable channel

in-app message The only message is capture a channel first.

Nothing waits

Marketing · copy only
The control group

The old canvas’s control proved the negative read; the new one proves the replacement earns its place.

Nothing waits

Analytics
Not eligible at entry

Nothing: the boundary is the design.

Nothing waits

Marketing · copy only

The words are on the journey map above. A D number opens that row of the Engineering brief; the engineering board uses the same numbers. Live sends not named here continue unchanged. Before anything in this journey goes live: One lapse-owner: the old canvas stops as this journey starts; Broadcasts stand down for owned episodes; The day-60 hand-off exists before anyone reaches it.

What needs deciding

the questions the build waits on, from the matrix’s register; the full list with every owner is on the reference
The questionOwner
Reachable audience by channel for the lapsing band, and daily entry volume at the proposed threshold.Blocks launch Analytics: Ivi and Ben
The band or score for Braze selection: which score movement opens an episode, per persona.Blocks launch Analytics: Ivi and Ben
The retiring canvas’s exact entry segment, its re-entry setting, and its control split: one console look at the entry step.Lisa, and the Braze developer
The retiring canvas’s conversion setting.Lisa, and the Braze developer
Confirm where every button lands: sign the register 'Where every tap lands' on this matrix against the deep-link availability list, one pass for the journey (Sheran, 6 Sep 2026, ruling J5-12).Product
Working detailPack version: 14 Oct 2026, 13:07 (London)