Improved click rate by 50%

Context

Bankdata builds digital banking products used by a large share of Danish banks. The design culture was mature in many ways — but almost entirely intuition-driven. No split testing, no structured way to validate design decisions with data. I set out to change that.

The problem

Getting new tooling approved at Bankdata is genuinely hard. Security reviews, managerial sign-off, procurement cycles — the friction is high enough that teams default to building in-house rather than adopting proven third-party solutions. Split testing was no exception. Rather than fight the approval process, I worked within it: I championed the development of an in-house A/B testing solution the local team could own and reuse, with an architecture designed to eventually roll out across the wider org.

The first test

The inaugural experiment was deliberately simple: a button relocation. Simple enough to be a clean test, meaningful enough to matter. The result was a 50% uplift in clicks — statistically significant. At the scale Bankdata operates, that translates to a potential increase in leads worth upwards of 17M DKK annually for our partner banks. We don’t control what advisors do with those leads — but we now control the evidence that design decisions move that needle.

The uncomfortable follow-up

Having built the capability, I became one of its most vocal critics — not the tool itself, but the instinct that produced it. We spent considerable time solving problems that dedicated third-party products have long since solved, time that could have gone toward delivering actual value to users. I’ve since pushed hard for the org to move faster on third-party adoption and raise its tolerance for the approval overhead. The split testing story is my clearest example of why: we got the result, but at a cost that a tool like VWO or Optimizely would have absorbed for us.

What I learned

Building the capability was the easy part. The harder — and more valuable — shift was recognizing when building is the wrong instinct, and having enough credibility from having done it to make that case convincingly.