Before the trip: reassure, wayfind, bank the value. The customer's four questions (did it work, which platform, which coach, was it worth it) go unanswered on every channel while the window sells taxis and swaps over them. One arc per trip answers all four in order and demotes every sell below the reassurance, with disruption yielding everything to the service layer. The stake is won-trip lifetime value: £115.62 against £22.92 when the trip goes wrong.
Today beside what is proposed at every moment, with the trigger, the message, the guardrails and what each moment needs. Filter by phase, by new or changed, and by the channels the customer holds.
Open the walk →▶The film of this journey as the customer receives itEvery segment, outcome and channel state, played as the customer meets it.
Open the film ↗On-train is the most negative stage in the voice of customer, the anxiety starts at purchase, and every proactive morning send today is a sell. The won trip's questions (did it work, which platform, which coach, was it worth it) go unanswered on every channel, while the window carries a 22-month undecided timing test, a duplicate reminder from an unidentified trigger, and a catering email measured on opens.
“Did it work? Which platform, which coach, and was it worth it?”
The next behaviour we want: the trip taken calmly on the upgraded seat, the value banked, and no support contact raised, because every question was answered before it was asked.
The matrix spec's per-row evidence; the voice-of-customer and value reads.
Reassurance never sells; the yield order ranks every commercial send below wayfinding; nothing fires on a disrupted trip; no opt-in ask anywhere in the arc; no saving figure renders until substantiated; anchored sends respect the 04:30 floor.
| Objective | Protect the won trip’s value: answer the customer’s four questions (did it work, which platform, which coach, was it worth it) before they are asked, and stop the window selling over them. |
| The journey | One arc per trip: the wait email for the long-lead tail, the day-before reassurance (push-primary), the morning wayfinding line with its lock-screen pin, the persistent trip state everything lands on, and the on-board modal that banks the win. The estate’s sells join the yield order below all of it. |
| Who enters | Every upgrade winner at purchase, on the trip clock. The 18 percent with no proactive channel live on the trip state; disrupted trips leave marketing entirely. |
| How it exits | Departure, then trip end (the post-trip journey opens); refund or cancel exits instantly and flips the state; the first-win journey’s episode resumes after travel with its rebook floor. |
| How we know | The anxiety-contact rate and CSAT, with the email A/B’s non-inferiority floor as the guardrail; every attributed figure in the window stays labelled attributed until tested against the estate-wide holdout. |
Every row’s full logic lives on the Decisioning matrix: one home per fact.
| Team | The whole ask in one line |
|---|---|
| Sheran, at Gate 1 | Ratify the post-win slot assumption and the deleted recovery send; rule the catering call; direct the timing-test decision. |
| Analytics | Decide the timing test, size the non-inferiority floor and the long-lead slice, define the deflection baseline, and substantiate the saved amount. |
| The Braze developer | The arc, the yield order, the clamp, the episode flag and the stand-down. |
| Engineering | The interim platform-and-coach source (it decides the wayfinding phase), the refund and cancel path, the fire-time disruption flag, and the zombie’s caller. |
| Product | The trip-state surface ruling, the on-board modal, the pin’s re-scope, and the swap-defect date. |
| Design | Six assets on the Design brief tab: three reassurance beats, three value surfaces. |
| The service track | Author the disruption service comm: the window’s top experience gap. |
*Sources. Value and contact reads: the measurement tree and support analysis, confirmed. The voice-of-customer percentages: the confirmed analysis. The proven-incremental reads: the estate revenue audit, April to June 2026, applying to the family’s narrow eligibility (the reach sizing states its true coverage). Attributed figures are labelled attributed, never incremental.
Phase is derived from each row, never authored: engineering work that is a new event, feed or surface is Phase 2; new wording, a deletion, an Analytics trait or Braze wiring is Phase 1. The deck walks Phase 1 in situ.
| The treatment | Who gets it | What stops it: caps and suppression | When nothing is sent |
|---|---|---|---|
| The wait email | Winners whose trip is seven or more days out | Marketing layer, inside the arc's once-per-trip-per-send rule. Suppression Refund or cancel exits the arc; live disruption yields the whole arc to the service layer. | For short-lead winners the answer is do nothing: the day-before beat is hours away. This email exists only where the wait is long enough for doubt. |
| The day before | Every reachable winner at T-22h | Reassurance class within the arc; once per trip. Suppression Refund or cancel exits; disruption yields; the episode flag stands broadcasts down. | The beat is reassurance. A customer whose trip is tomorrow does not need a cross-sell; the arbitration order enforces that downstream. |
| The morning | Push-armed winners on travel morning | Reassurance class; the single correction re-issue is cap-exempt by ruling. Suppression Disruption yields the arc: a disrupted trip gets service comms, not a cheerful platform line. | Where the payload is unavailable at fire time the push does not send a vague fallback: the trip state carries the journey and the push stays silent. A wayfinding push without wayfinding is a sell slot in disguise. |
| The live pin on the lock screen | iOS winners from T-2h | Not a send once started; the start event sits in the reassurance class. Suppression Disruption flips the pin's content to the service state, not a marketing message. | Where the pin starts, the push suppresses: one wayfinding answer per customer, never two. |
| On board | Winners on a non-disrupted trip, once on board | One render per trip. Suppression Disruption hard-suppresses; refund or cancel exits. | On a disrupted trip the honest action is silence from marketing: the service layer owns the customer. |
| The persistent trip state | Every winner, the whole window | Not a send. Suppression None: it is never consent-gated and never suppressed; its content changes state instead. | The state is the do-nothing answer done properly: solving the need in the product beats adding a message, and most of this journey's sends are alerts pointing at it. |
| The day-before email family, kept as the control | Its current narrow eligibility, unchanged | As today. Suppression As today, plus the arc's episode exclusions as they land. | Do nothing, deliberately: this row's treatment is the absence of change. |
| Ride cross-sell to the station | Winners whose trip starts in the ride partner's coverage cities | Commercial class within the arc's order. Suppression Suppressed on disrupted trips; loses same-window collisions to wayfinding and logs the loss. | The sell is legitimate; its rank is not. Below operational and reassurance, always. |
| Ride cross-sell from the station | Winners arriving at one of the 78 Veezu-covered stations (console-confirmed filter), personalised on their home station | Commercial class. Suppression Disrupted trips suppress: a delayed arrival is exactly when this push currently fires into anger. | At arrival the reassurance job is done; a relevant sell is acceptable if it ranks correctly and shuts up during disruption. |
| The train-swap reminder | Ticket buyers on swap-capable operators; upgrade winners only when the same customer also bought the ticket | Commercial class; re-entry after 2 days with no exit criteria is reviewed as part of the same pass. Suppression Disruption yields; loses same-window collisions to wayfinding. | For a customer holding a won upgrade on the same trip, the honest question is whether a swap sell belongs in their morning at all; the kill threshold answers it with data. |
| The legacy swap push | Unknown: whatever cohort its unidentified trigger still selects | Outside every rule until its caller is known. Suppression None known. | A send nobody can explain does not get to keep firing at customers because switching it off takes effort. |
| The catering email | Same-day purchasers on one operator's services, email opted in | Commercial class; joins the order. Suppression Disruption yields. | If the reassurance beats already carry per-operator inclusions, a separate catering email may be the estate answering a question the journey already answers. |
| Decision | Owner | Where it stands |
|---|---|---|
| How the journey is measured: the holdout, the baseline and the read | Michael, with CACI | On the Test and Learn record |
| # | The ask | Hand back | Start here |
|---|---|---|---|
| 1 | Decide the 22-month timing test on the read you hold: which arm wins, or does it close without a winner? One decision retires two arms. | The decision. | The ride pick-up row on the matrix. |
| 2 | Size the non-inferiority floor for the day-before email A/B on the proven family’s baseline, and state the family’s true coverage. | The floor and the reach sizing. | The measurement position on the Campaign brief. |
| 3 | Define the deflection baseline: does the trip state, made prominent, already deflect most of the 1,753 anxiety tickets? The wayfinding push’s marginal case depends on it. | The baseline. | The wayfinding row. |
| 4 | Size the long-lead slice (won seven or more days out): go or no-go on the wait email’s test power. | Go or no-go. | The wait-email row. |
| 5 | Substantiate the saved amount (the walk-up anchor is unconfirmed) and build the persona register trait, nightly. | The figure and the trait. | The on-board modal row. |
Conversion is behaviour, never opens: the catering email’s open-based success is the standing instrument fault, and every canvas converts on purchase or behaviour from day one.