How to reduce onboarding drop-off without weakening KYC
When a regulated check is where new users disappear, the usual advice is to make the check faster. Often the check isn't the problem. At elyps, 8 in 10 new sign-ups left at one screen, and every KYC requirement stayed. What changed was when people met it.
Most advice on KYC drop-off is about the check itself: a faster document scan, fewer fields, a smoother selfie step. That's worth doing when the check is where people struggle. But in a regulated product you can't remove the check, and it often isn't where people leave.
At elyps, a French-Belgian digital bank where I led product as the first product hire, roughly 8 in 10 new sign-ups abandoned onboarding in the weeks after launch. Here's how we found where they left, why, and what we changed without touching a single compliance requirement.
Find the screen before you fix the step
The first instinct was the obvious one: KYC is too much friction, so simplify it. Product analytics said something more useful. The drop-off wasn't spread across the funnel. It sat almost entirely on one screen, the step right before identity verification started.
That distinction matters. People who quit halfway through a document scan are telling you about the scan. People who leave before they start are telling you about the moment. Cut your funnel step by step and compare the screen before the check with the check itself. If most of the loss happens before anyone has opened the camera, a faster scan won't reach it.
Why a shorter form is the wrong first move
When one screen loses most of your users, the reflex is to remove steps. Sometimes that works. When the step is a legal requirement, it's off the table, and the effort goes into negotiating with compliance instead of learning anything.
The more durable fix is usually about when and how you ask, not what you ask. That reframe is the reason to write a test card before you touch the product: it makes you name the real hypothesis instead of the symptom. At elyps, the experiment was live four weeks after the insight.
Ask what the moment looks like for the user
So we stopped asking how to make KYC lighter and started asking why people left at that point. The main problem wasn't the verification. It was the context people were in when the app asked for it.
Our hypothesis was about timing. Verifying your identity takes a document within reach and a few quiet minutes, and someone signing up in a spare moment doesn't always have either. If people could choose a better time, and we reminded them when it came, more of them would finish.
Change when, not what
We wrote it down as a test card before building anything. We believed that letting users defer KYC to a more convenient time, and reminding them at home, would reduce drop-off on the pre-KYC screen and increase KYC completion. We would call it a success if overall onboarding churn fell by at least 10%.
The change itself was small:
- A "Do it later" button on the screen before KYC.
- Tapping it led to a notification consent screen that explained the deal: a reminder would come that evening.
- At 8pm local time, a push notification with a deep link straight back to the KYC step.
Nothing in the verification changed. Same documents, same checks, same order inside the flow. The only thing that moved was when a user met it.
Measure the whole funnel, not just the screen
A deferral button can make one screen look better while the funnel gets worse: people tap "later" and never come back. So we measured three things: drop-off on the pre-KYC screen, KYC completion within 24 hours, and overall onboarding churn. The last one was the one that counted.
The test ran for four weeks. Overall onboarding churn fell from roughly 80% to 60%, a 25% relative decrease. Notification opt-in rose too, because people finally had a reason of their own to say yes. "Do it later" became the default for every new user, and notification consent screens started leading with the user's benefit instead of compliance or marketing language.
Across the whole engagement, elyps' activation conversion rose by 30%. The elyps case study has the test card and the learning card.
What to settle with your compliance team first
None of this is a way around KYC. It's a way to stop treating every part of onboarding as fixed because one part is. Before you design anything, sit down with compliance and separate the two lists:
- Which checks does the regulation require, and before what? Before an account exists, before money moves, before a card ships?
- Which parts are product choices that only look like requirements: the order of screens, when the check starts, the wording, whether a user can stop and come back later?
The first list is fixed. The second is where your experiments live.
Other changes to test on the same screen
The elyps result came from asking when and how, not what. The same question suggests other tests. Each is a hypothesis, not a fix, so write a test card for the one your data points to:
- Tell people what they'll need before the check starts, so nobody opens the camera without the document.
- Let people stop and pick up where they left off, instead of starting the flow again.
- Ask for notification permission with the user's benefit first, as the elyps consent screen did.
Where to start this week
Pull the funnel step by step for last month's sign-ups and find the screen where most people leave. If it's the one before a check you can't change, you have the problem elyps had, and probably a cheaper fix than a new verification vendor. Then find out why people leave there, write the test card, and decide what success means before you build anything.
This is the loop behind the Funnel Conversion Sprint: product analytics for where users drop off, interviews for why, and a scoped experiment for whether the fix works. The broader method is in how to improve user activation.
