Case study

How Northgate Exchange cut onboarding drop-off without loosening its checks

A mid-size European exchange was losing most new sign-ups before their first deposit. Reordering the flow, not weakening it, recovered them.

Northgate ExchangeCrypto exchange5 min read
+38%
Onboarding completion
4 min (from 11)
Median time to verify
-27%
Manual review rate
-41%
Verification support tickets

The problem

Northgate Exchange is a fictional name for a composite of mid-size, EU-registered spot exchanges. Like its peers, it has to verify every customer before allowing deposits, and like many of them it had built that process in layers over several years as rules changed.

By early 2026 the result was a sign-up flow of fourteen screens. New users were asked for an email, a password, a phone number, a full address, an occupation, a source-of-funds declaration, a photo of an identity document and a selfie, all before they had seen the trading interface. Fewer than a third of people who started registration completed it. The largest single loss was at document capture, where nearly half of attempts on older phones failed on the first try and were routed to manual review, which took up to a day.

The compliance team was clear that none of the checks could be dropped. The product team's question was whether they all had to happen in that order, at that moment.

What they did

The team mapped each piece of information collected against the specific regulatory obligation that required it, and the point in the customer lifecycle at which it was actually needed. Three changes followed.

  • Progressive collection. Account creation was reduced to email, password and country. New users could then explore the interface, view markets and set up two-factor authentication. Identity verification was required before the first deposit, as before, but was presented as a single clearly explained step with a time estimate. Source-of-funds questions were moved to the deposit thresholds at which they are required.
  • Better document capture. The document step was rebuilt with live edge detection, glare warnings and an immediate check that the image was readable before upload. Users whose capture failed were shown the specific reason and allowed to retry on the spot. Electronic identity schemes were added as an alternative in the countries where they are available.
  • Save and resume. Progress was stored so that a user interrupted halfway could return by email reminder to exactly where they had stopped, on any device.

The compliance function signed off each change, and the firm's risk thresholds, screening providers and review criteria were left untouched.

Results

Over the twelve weeks after launch, the share of started registrations that reached a verified account rose by 38 percent. Median time from starting verification to approval fell from eleven minutes to four, largely because fewer submissions needed a human. The manual review rate dropped by 27 percent, and support tickets about verification fell by 41 percent.

Fraud and chargeback indicators were monitored weekly against the previous flow and showed no deterioration. That was the condition on which the project had been approved.

What others can take from it

The instinct when onboarding leaks is to ask compliance what can be removed. The more productive question here was when each item is needed. Regulation generally specifies what must be known before a given activity, which leaves room to sequence.

Second, a great deal of drop-off blamed on regulation was really a camera problem. Telling a user immediately why a photo failed recovered more sign-ups than any change to the form.

Third, the team measured the right outcome. Completion was tracked to a verified, funded account, and fraud metrics were reviewed alongside it, so that an improvement in one could not hide a decline in the other.

Latest stories

All stories

Worth a look

All guides