Four verticals, one product. The decisions behind one trip.
The prototype shows what I would build. This page shows why: where each vertical enters the flow, what I rejected and on what grounds, and which bets deserve a cheap probe against real investment. Written for a rail retailer expanding into hotels, car hire, insurance and media.
The unit of design stops being the ticket and becomes the trip.
A rail retailer that sells only tickets competes on price and app quality. A rail retailer that holds the whole trip competes on something harder to copy: it knows where you will be, when, and what is missing. Two tickets to Edinburgh with two nights between them is a fact worth money, and the customer benefits when the product acts on it well.
Every decision below follows from one principle: new verticals attach to the trip a customer has already committed to. They never interrupt the decision that funds the whole business, which is buying the ticket.
Hotels
Enters: the trip screen, inside the gap it fills
- The decision
- Hotels appear after both tickets are chosen, as part of the trip timeline: "2 nights in Edinburgh" with three options ranked by walk time from the arrival station, not by star rating or sponsored placement. After a four-hour train, what you are buying is the distance to your bed. Skipping is one tap and carries no guilt copy.
- Rejected
-
- Hotel teasers on the results page. They compete with the hardest decision in the flow. The cost lands on rail conversion, the guardrail metric.
- Post-purchase email attach. Relevance decays in minutes; by the evening the customer has booked on another app.
- A full-screen hotel interstitial after checkout. Coercive, reads as an upsell wall, and trains customers to sprint past everything between them and the ticket.
- Measure
- Attach rate and skip rate on the trip screen; rail checkout completion as the guardrail that must not move.
Insurance
Enters: the payment screen, one tap, defaulted off
- The decision
- Trip cover is a single toggle at the moment of payment, priced in plain sight, with cover explained in one sentence and no asterisk. It protects the decision the customer just made, which is why it belongs at payment and nowhere else.
- Rejected
-
- A pre-ticked box. A dark pattern and a regulatory risk in the UK. Short-term take rate, long-term trust damage across every vertical.
- A dedicated insurance step. Adds friction to every checkout to serve the minority who want cover.
- Selling it after purchase. Too late to protect anything; reads as an afterthought because it is one.
- Measure
- Take rate, checkout completion delta, and refund complaint rate. A high take rate with high complaints is worse than a low take rate.
Car hire
Enters: a fake door on the trip screen, deliberately
- The decision
- Car hire in the prototype is a dashed card that opens a sheet saying, honestly, "we are testing this". Demand for car hire from a rail audience is unproven; inventory, supply deals and support are expensive. A fake door measures intent for the cost of one card, and the honest framing turns the probe into a conversation with customers rather than a trick on them.
- Rejected
-
- Building a real car hire flow first. The classic way to spend two quarters learning what a two-week probe would have told you.
- A dishonest fake door that looks bookable. Learns slightly more, costs trust, and trust is the one asset shared across every vertical.
- Measure
- Tap rate on the card, and taps converting to a declared "I'd use this". Set the threshold that triggers real investment before the probe runs, not after.
Advertising and media
Enters: not the booking flow. The confirmation and trip views, later
- The decision
- Media is not in the prototype, on purpose. Its economics are different: the first three verticals sell to the customer; media sells the customer's attention. The only placements that could carry it without damaging the core are the confirmation screen and the trip view, where a relevant partner offer can read as a service. The probe is house inventory first: promote your own hotels and cover in those slots, measure blindness and complaints, and only then price the space for partners.
- Rejected
-
- Placements inside search or results. Selling the customer's attention while they are trying to spend money with you taxes the core to feed a side bet.
- Third-party display from day one. Unproven slots at unproven prices, with the brand damage arriving before the revenue does.
- Measure
- Engagement with house placements, complaint and opt-out signals, and only then partner yield.
What keeps four verticals feeling like one product.
The core never pays
Nothing attaches before the ticket is committed. Rail conversion is the guardrail metric for every experiment in every vertical.
One visual grammar
Blue marks the core journey, coral marks the new, one hierarchy over both. Four verticals read as one product, not four apps stitched together.
Price honesty everywhere
Totals are final, fees are inside the number, and the fare or rate is named. A rule from rail that every vertical inherits unchanged.
Trust language is shared
Cancellation and refund wording is identical across trains, hotels and cover. Trust is the connective tissue; it cannot vary by vertical.
Every attach is skippable
One tap, no interstitials, no guilt copy. The customer who says no this trip is worth more than the attach forced on them today.
The trip is the home
Everything bought lands in one timeline with one reference. The confirmation is the front door of the trip, and the natural home for what comes next.
Not every bet deserves the same treatment, and deciding which is which is most of the job. My reading of the four:
Insurance is the quiet winner: high margin, near-zero marginal cost with a partner underwriter, and the design risk is trust rather than demand. Run it now and measure complaints as hard as take rate.
Hotels are the big prize and worth real investment, staged: start with affiliate inventory to prove the attach point, and go deeper only when the trip-screen attach rate clears a threshold agreed before launch.
Car hire is cheap to test and unproven in this audience, so it gets a fake door and nothing more until the signal argues otherwise.
Media has a real prize but different economics and real brand risk, so it learns on house inventory before a single partner slot is sold.
Numbers first, opinions second.
- Rail checkout completion. The guardrail. If any vertical moves it down, that vertical redesigns or dies, whatever its own numbers say.
- Attach rate per vertical, at the attach point. Not blended across the app, where a good placement hides a bad one.
- Skip and complaint signals. Skips, refund complaints, support contacts per attach. The cost side of every attach ledger.
- Fake-door conversion. Taps to declared intent, against a threshold set before the probe runs.
- Return visits to the trip view. The measure of whether the trip object is real to customers, and the foundation the next vertical stands on.