A fintech organization

01 · A scattered picture

In fintech products the first obstacle to growth is usually not the product but the opening experience. When we started, a mobile-heavy user base was walking into a 14-screen onboarding flow; one in three users abandoned the digital identity step halfway through.

The second problem was less visible. Product, marketing and risk teams were getting different answers to the same question from different tools. There was no shared picture of which screen lost users or which segment got stuck; meetings opened with data disputes and closed on opinion.

02 · Measure first

For the first month we designed nothing. We instrumented every step of the existing flow: collection was standardised on one schema with Segment, and a single transformation model everyone shared was built on BigQuery with dbt. Drop-off points were marked as data, not guesses.

By the end of that month the teams were looking at the same funnel for the first time. It became clear the sharpest drop was on the document upload screen, and a third of photos taken with mobile cameras were failing quality control.

03 · Rebuilding the flow

Onboarding was reduced to a staged structure. The account opens in the first three steps; identity verification kicks in after the user has seen value. Document upload was redesigned around the phone camera: a live framing guide, instant quality feedback, a flow that completes one-handed.

Every change shipped as a hypothesis and was watched against data for a week. Measurable steps were chosen over a grand redesign; we rolled back two experiments that did not work and recorded them as cheap learning, not loss.

04 · One source

A single dashboard was built on Looker Studio where product, marketing and risk read the same definitions. Activation messages fire from behavioural data through Iterable; a stuck user gets one reminder, at the right moment, on the right channel.

After five months KYC completion was up 41 percent, onboarding time was cut by more than half, and time to first transaction dropped under 8 minutes. More importantly, the teams now talk from the same data in a weekly rhythm.