Win-back: relevance or silence, never a blast.
“Show me it's relevant to a trip I actually take, make coming back effortless, and if you can't do either, leave me alone.”
After day 60, nobody owns the customer.
Why it happens Churn is an event, not a fade: a specific unresolved grievance, a feels-rigged conclusion, or price drift against remembered cheap upgrades. A generic miss-you message answers none of them, and the return door itself sometimes fails.
The behaviour we are driving Tap through to their own route with a live price and bid within 14 days; or walk back in unaided and be recognised. Either way a transaction closes the episode and hands the comeback to Journey 1.
A propensity-to-return score reads the dormant pool daily: it decides who is worth a knock, who is left in peace, and later whether a reason needs an offer.
A belief until Analytics restates the per-reactivation value (the £219.94 basis is under challenge) and reconciles it with the response ceiling; the Q3 marketing model's segment sizing, 8 Jul 2026 basis, mapped in the business case of 4 Sep 2026.
| Model row | Marketable | Worked volume | Rate | Value per customer | Gross a year | Incremental a year |
|---|---|---|---|---|---|---|
| Lapsed transactors, 91 to 240 days, value-tiered win-back (RECOVER) | 49,300 | 3,500 contacted a month | 5% vs 3.5% organic | £55 × 2.5 months × 12 a year | £288,750 | £86,625 |
Q4 2026 estimate: £7,357. Annual incremental £ × days live in Q4 ÷ 365 × the Phase 1 share (100%). Proposed launch 1 December 2026; the date is Sheran's to rule.
Ladders to Reactivation: reactivation rate 4.8% to 6%; MAU transacting 4.5%. Metric moved Transacting users %. Roadmap type New revenue. Evidence B, benchmark. Also moves Reactivation rate (30-day). Shared The anniversary send fires inside the churned pool counted under Journey 9.
DRAFT, first pass, 4 Sep 2026. Basis: Q3_Marketing_Model (24).xlsx, Initiative Sizing tab, 8 Jul 2026 basis; figures incremental of organic return, credited over average active months; rates are benchmarks until holdouts replace them. Incremental of organic return, credited over the average active months; every rate is a benchmark until a holdout replaces it.
Each engineering-row chip is coloured by the ask: APP AND JOURNEY OWNED COMPUTED TRAIT · ANALYTICS NEW EVENT · ENGINEERING BUILDS READY · NOTHING TO BRIEF
| The moment | Phase 1, ships now | Phase 2, waits on engineering | Who builds it |
|---|---|---|---|
on the map: 1 · The route-anchored win-back The route-anchored win-back | braze push, email After day 60 of silence nobody owns the dormant customer and almost nothing reaches them. | Nothing waits | Braze · Braze buildAnalytics · new model, computed traitDesign · copy and creative |
on the map: 3 · The comparison anchor The comparison anchor | email, braze in-app message One concrete reason to look again: the walk-up fare on their route today against what their last upgrade cost. | Nothing waits | Braze · Braze buildAnalytics · computed traitDesign · copy and creative |
on the map: 2 · The welcome-back landing The welcome-back landing | braze content card A returner stops landing on the same generic cold start as everyone else: the card recognises the absence, restates that the account is as they left… | Nothing waits | Braze · Braze buildAnalytics · computed traitDesign · copy and creative |
on the map: 4 · The final attempt, then a genuine end The final attempt, then a genuine end | email, braze push One last try that says so: what has changed on their route, live prices this week, and after this we stop asking. | Nothing waits | Braze · Braze buildAnalytics · new model, logic changeDesign · copy and creative |
on the map: 5 · The anniversary message The anniversary message | email, braze push A year on, the one owned reminder of what they liked: the same trip, this year’s real prices, against the walk-up fare. | Nothing waits | Braze · Braze buildAnalytics · computed traitDesign · copy and creative |
on the map: 2 · The welcome-back landing The live incentive card | The card retires in the go-live change, under the archiving rule. | Nothing waits | Braze · Braze build |
on the map: 1 · The route-anchored win-back Unlikely to return | The weekly-blasted deep tail is now deliberately left alone: churned one-time customers are about 80% of the stock and about 2% ever come back, so a… | Nothing waits | Analytics |
on the map: 1 · The route-anchored win-back Likely to return anyway | The one harm the record has measured is chasing customers who were coming back anyway (a bid-volume holdout on active power users read minus 11.8%). | Nothing waits | Analytics |
on the map: 1 · The route-anchored win-back The organic returner | Every read in this journey must clear organic return, not silence; sending win-backs to people already back both wastes sends and poisons the read. | Nothing waits | Marketing · copy only |
| The email deep tail | The deep tail stops being emailed on stale consent: sender reputation is a shared resource and the tail costs more than it could ever return. | Nothing waits | Braze · decision needed |
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.
| The question | Owner |
|---|---|
| Rule the welcome-back owner: this journey fires and measures the card; the returning-customer home journey renders the surface. One owner, decided. | Blocks launch Sheran, at Gate 1 |
| Size and meter the one-time backfill wave: the persuadable band only, ranked by what a return is worth, route-stale customers left out, with the double-read holdout design. | Blocks launch Analytics: Ivi and Ben |
| Run the frequency-band read (reactivation rate by transaction-frequency band across lapsed and churned), then size the two reverse-holdouts and the within-band holdouts from it. | Blocks launch Analytics: Ivi and Ben |
| Restate the 1.5% reactivation objective’s denominator, and reconcile the pool counts across sources (279,876 verified against the circulating figures). | Analytics: Ivi and Ben |
| The 24-month email deep-tail rule: confirm the working assumption (no engagement and no consent refresh in 24 months means no email). | Legal |
| Asked 6 Sep 2026: restate what the £219.94 per-reactivation value measures; until then no offer depth is priced and the journey’s revenue figure stays labelled as a belief. | Analytics: Ivi and Ben |
| 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 |
| # | The ask | Hand back | Start here |
|---|---|---|---|
| 1 | The propensity-to-return band into the daily sync, on the standing pattern: the frequency band on day one, the model’s output when it clears its backtest; this one extension is the whole phase 2 launch gate. | The band in the sender, refresh cadence stated. | The Engineering brief’s plumbing band. |
| 2 | The route, from-price and anniversary traits, nightly, for this cohort. | The traits live, coverage stated. | The traits band on the Engineering brief. |
| 3 | Size the backfill waves and the double read: holdouts within bands (journey lift) and banded waves against a batch arm (targeting value), with the underpowered-band risk stated. | The wave plan and test design. | The measurement position on the Campaign brief. |
| 4 | Restate the 1.5% objective’s denominator and reconcile the pool counts across sources. | The restated objective and one pool number. | The hero figures on the Campaign brief. |
| 5 | Deliverable-reach counts per channel: consent is not reach (roughly 1.8x overstatement; warehouse push consent three months stale). | Reach ceilings per band and channel. | The suppressed rows on the Decisioning matrix. |
Every read must clear organic return: about 33,000 lapsed customers a month come back unaided and convert at new-customer rates.