The benefit was growing.
Operations couldn't keep up.
Every cancellation became a phone call. Cost scaled with the product's success.
A premium travel subscription offered "Cancel for Any Reason" — cancel for any reason whatsoever. On paper, total autonomy. In practice, to use it, the customer had to call.
At the worst possible moment — flight cancelled, trip undone — the person had to relive the problem out loud, hold on the line, and hope the agent got it right.
CFAR adoption was growing fast, and the contact center carried high daily volume from this flow alone. Each phone-mediated cancellation cost a few dozen euros between agent time and rework — a bill that scaled linearly with the product's success.
The problem wasn't a bad experience. It was the impossibility of scaling without breaking.
Visibility was the obvious answer.
And the wrong one.
The team wanted to give the benefit more prominence. The calls said otherwise.
The instinct was visibility: banner at the top of the app, notification when a flight is delayed, highlight on the booking confirmation. More exposure, more adoption.
To understand the problem for real, I worked with research and CS to get access to cancellation call transcripts and the agents handling them every day. I read dozens of them.
Raises visibility, but surfaces a cancellation option to people who'll never need it.
Reduces handle time, but the customer still has to call. Scales the problem, doesn't solve it.
The customer wasn't confused. They were scared.
Research corrected the question.
I was looking for a usability problem. I found a belief problem.
CFAR didn't have a usability problem. It had a belief problem.
I expected fear of the process — "can I do this on my own?". It wasn't that.
The fear was about the money. People called because they didn't believe the value would actually come back to them.
This reframed the project. The work wasn't to design a cancellation flow — it was to prove, at every step, that the money was safe. Solving the belief was what took the call out of the equation.
The first hard call:
relevance over visibility.
Showing a cancellation option to an ineligible user creates false hope — worse than no option.
Growth wanted CFAR everywhere. Design's direction went the opposite way: gate the flow on eligibility, checked silently before the screen rendered.
There was also a budget reason. CFAR is a liability — every trigger is a refund going out, within a budget sized for predictable volume. Broad visibility increases triggers and pressures that ceiling: the same "grow until it breaks" as the contact center, now on the financial side.
I traded reach for reliability, eyes open on the cost: an eligible user who doesn't open the app at the right time may never discover the benefit. A benefit that surfaces but can't be used destroys more value than one some people never see.
CFAR everywhere. More exposure = more perceived value = better subscription retention.
Gate the flow on eligibility, checked silently before the screen rendered.
The flow: each screen
disarms a fear.
The journey isn't a sequence of screens. It's the progressive dismantling of distrust.
Research wrote the script: the flow was designed against the questions a user asks in panic, in the order they ask them.
Fear / 01
"Is this for me?"
Eligibility resolved before the first screen. If you reached the flow, you can use it. The silent check was expensive — we negotiated scope with engineering until the gate fit the technical budget.
Fear / 02
"Will I lose money if I choose wrong?"
The refund choice screen. The hardest in the project — and the next section is only about it.
Fear / 03
"What if there's a catch?"
Pre-confirmation shows the exact amount and what happens next. When the value depends on the airline, the screen admits the uncertainty there, with context. Copy written with content and legal.
Fear / 04
"It's done — did it actually work?"
The success screen puts the refund amount as the headline, big and unambiguous. The direct answer to the fear research revealed: the money came back, and it's here.
Fear / 05
"What if it errors out right now?"
When cancellation fails, the flow loops back without losing context. The user resumes where they were — a dead end at a moment of crisis sends the case straight to the phone.
The complete flow: entry → refund choice → confirmation modal → success (Free Trip / Prime) → error state. Red path marks the negative / cancellation funnel route.
The screen that
proved me wrong.
I bet on credit prominence. The A/B knocked down my hypothesis — and I rebuilt the screen.
The refund choice — cash or travel credit — was the hardest screen in the flow. And I started it on the wrong side.
The hypothesis belonged to the whole team, and I bought it: highlight the credit. The reasoning was solid — credit stays in the ecosystem, cash leaves it, and the refund budget was already a known concern. Giving credit more visual weight seemed good for the business on every front.
Instead of going purely by logic, we took both versions — credit highlighted and equal weight — into a production A/B. It ran for weeks, with real traffic, enough time to measure not just the on-screen choice, but what happened after it.
The result contradicted the hypothesis that was also mine.
The credit-highlighted version converted more to credit. The number the team wanted to see. But weeks of follow-up showed the rest of the bill — exactly the budget pressure we wanted to avoid.
Credit conversion on the screen. The number the team wanted to see.
Post-cancellation doubt calls. Plus more benefit triggers — the budget pressure we wanted to avoid.
Prominence didn't convert more conviction. It converted more clicks. And clicks without conviction call back later.
The prominence captured users who chose by salience, not by understanding the option. Doubt wasn't eliminated — it was postponed to after the choice, where it cost a call and a refund, not a second of hesitation on screen.
That's when I went back. I redesigned the screen for equal visual weight: cash and credit with the same treatment, the choice returned to the user. It converts less to credit in the short term — and was the right call, because the prominence gain came with a support and liability cost that consumed it.
The decision didn't close with me alone. The A/B was reviewed with PM, CS, and finance at the table — CS confirming the support cost, finance confirming the budget pressure. The data aligned three areas with different starting interests.
Seniority wasn't defending the right screen. It was accepting, in front of the team, that the screen I had proposed was the wrong one.
The contested screen — equal visual weight
The product took over
the work that was the people's.
Agents stopped mediating trust manually.
The flow shipped complete across all three platforms. Aside from the A/B on the refund screen, the project didn't have broad formal instrumentation — measurement was concentrated on the decision that most divided the team. The post-launch impact signal came from where the problem was born: the contact center.
One design decision deliberately cuts against short-term conversion: the pre-confirmation exposes the credit's uncertainty before the commit, which likely reduces immediate conversion in the flow. Trade-off accepted — fewer conversions now, fewer calls and less regret later. A flow that converts by hiding uncertainty isn't better; it's deferred debt.
Agents reported a noticeable drop in CFAR calls. And the calls that remained changed in nature: they stopped being "is this going to work?" and became one-off questions.
Each avoided call is one of those few dozen euros that no longer goes out — and, more than the cost, it's the work of convincing the customer that left the person and went to the product.
It was this invisible work — done by a human, one call at a time — that kept the benefit from scaling. The flow didn't improve support. It made support unnecessary.
Walkthrough · The full experience
Six screens.
One loop.
From the cancellation intent to "your money is back" — and the CTA on the last screen that returns the user to the start of the journey.
01 / 06
Modify trip menu
The last CTA opens the app wallet
What the product taught.
Good research corrects the question, doesn't just answer it. I came in convinced it was a usability problem and even started sketching a flow on that hypothesis, before the transcripts showed the problem was about belief.
A team hypothesis isn't better than an individual one — it's just harder to abandon. Credit prominence had everyone's consensus. It took an A/B with real traffic to show the consensus was wrong, and it took swallowing the pride of having proposed the version that lost.
Every UX decision has a technical price, and part of the work is negotiating it. The eligibility gate only existed because it was talked through with engineering, not imposed on it.
Senior design solves a cost problem, not just an experience one. Trading visibility for relevance, and prominence for informed choice, were UX decisions — but what justified them, in the end, was what each one did to the cost of operating the benefit.