A UK payments startup came to us with a specific, expensive problem: people wanted the product, signed up for the product, and then quietly gave up before they ever used it. The funnel did not leak. It haemorrhaged.
How it started
The brief arrived as a single dashboard screenshot: a sign-up funnel with a healthy top and a cliff in the middle. Marketing was doing its job — thousands of people arrived intending to open an account. Somewhere between “create account” and “first payment”, most of them left.
The team’s instinct was to add. A progress bar to reassure people. A “why do we need this?” tooltip on every field. A save-and-resume feature so abandoners could come back. All reasonable. All treating the symptom.
We asked for one thing before writing any code: the recording of ten real sessions where someone dropped out.
What we found
Onboarding was nine screens. Two of them collected information the product did not need until much later — some of it never. One screen asked for a document upload on mobile, then rejected the photo with an error message written for a compliance officer, not a human holding a phone in a supermarket queue.
The cliff was not one bad screen. It was the accumulation: every extra field was a small reason to stop, and there were dozens of them before the user got to do the single thing they came for — move money.
What we built
We ran the same delete pass we run on our own products. Every field had to justify being asked for now rather than later. Most could not.
- Nine screens became three. Identity, one funding step, done.
- Verification moved to the background. The user starts using the product on a provisional limit while checks run, instead of staring at a spinner that owns their first impression.
- The document-upload step was rewritten around the phone camera, with error messages a tired person can act on (“too dark — try again near a window”).
- Everything the product genuinely needed later got asked for later, in context, when the user had a reason to care.
Nothing here was clever. It was subtraction, applied with discipline, on someone else’s product instead of our own.
Results
The client’s growth numbers are theirs to publish, not ours. What we can say plainly: onboarding completion moved in the direction everyone wanted, and the largest single category of support ticket — some variant of “I’m stuck at verification” — fell to a trickle, because the user was no longer stuck behind verification to begin with.
The most useful metric was quieter. The team stopped debating which reassuring element to add next, because there was far less left to be anxious about. A short flow does not need a progress bar. It needs to be short.