The direct debit form at Lendo

The short version

The problem

The direct debit step at the end of the insurance purchase asked for the account number in two fields, and dropping out there left a live policy with no way to pay.

What I did

Researched how every Swedish bank app writes account numbers, with screenshots from colleagues where I had no account. Then changed the form to one field.

What it was worth

Completion of the step up by an estimated 10%, from memory rather than a recorded number.

The call

A small form fix that cost almost nothing, because the new insurance system was going to replace the step anyway.

What changed

The direct debit step came at the end of the insurance purchase at Lendo, after the customer had already chosen a policy. It asked for an account to take the payments from. Dropping out there did not cancel anything. It left the policy live with no payment method, so the insurer had to chase the customer by post, and by the time the papers landed the insurance had faded in the customer's mind and churn on those policies was high.

The form asked for two things. Clearing number in one field, account number in another. That looks harmless until you look at where the number comes from.

The number comes from the customer's mobile banking app. So I researched how every Swedish bank writes account numbers. Some show two parts. Most show one combined string, and most offer a copy button for the full number.

I did not have accounts at every bank, so I asked colleagues to check theirs and send screenshots. No budget, no tooling, just asking. The picture was consistent. A customer copies one string from their bank app, and our form demanded they split it into two boxes by hand.

The fix followed the evidence. One field, accepting the number the way the bank app hands it over.

The thorough answer existed. An open banking connection, where the customer picks their account and never types anything.

That was already on its way. The new insurance system was going to replace this step with an open banking signup, so an integration built now would have been throwaway work. The one field form helped customers straight away and cost almost nothing, which is the right size of fix when a bigger platform change is coming anyway.

Completion of the step rose by an estimated 10%. That is from memory, not a recorded measurement, and I would rather leave it as an estimate than dress it up. The change did not sell more policies. It kept more of the ones already sold.

The method is the part I would repeat anywhere. Find where the input comes from, gather evidence cheaply, change the form to match, and save the big build for when it will not be thrown away.