Finova's iOS onboarding was a 14-step linear flow built for our web product and ported to mobile without adaptation. Users had to verify identity, link a bank, set a PIN, enable notifications, and complete a KYC form before they could do anything. Drop-off was concentrated at two points: the bank-linking step (43% drop) and the KYC form (31% drop among those who reached it). The core problem: we were asking for maximum commitment before delivering any value.
Three themes emerged clearly. First: users didn't trust the app enough to hand over their bank credentials before seeing any value — they wanted proof the product worked before committing. Second: the KYC form felt like a government form, not a product — dense, jargon-heavy, with no progress indicator. Third: the notification permission ask came too early (step 3) and was framed around us ("stay updated") rather than them ("know the moment your transfer arrives"). We also discovered that users who completed a single simulated transfer in a demo flow before linking a bank had 2.3× higher completion rates in an informal test run by a previous PM — this was buried in a Notion doc.
We restructured onboarding around value-first: show a demo transfer before asking for any credentials, defer bank linking until after the user experiences the product, and break the KYC form into conversational micro-steps with progress indicators. The notification ask was moved to post-first-transfer with a value-anchored message. We shipped in two phases: the value-first reorder (4 weeks) and the KYC redesign (8 weeks).
The most contested decision was deferring bank linking. The Growth team wanted to keep it early to maximise the number of linked accounts (a key metric). I argued that a linked-but-churned account was worthless, and that improving the activation funnel would drive more net new linked accounts than forcing the link earlier. We aligned by agreeing to measure "linked accounts active at D30" rather than just "linked accounts" — which gave both sides a shared definition of success. We also debated whether to build a full demo mode or just a video. We built a lightweight interactive demo (not a full fake account) because our research showed users wanted to feel the product, not watch it.
The KYC redesign required sign-off from our Compliance and Legal teams, which added 5 weeks to the timeline. The first version of the conversational KYC was rejected because one question's wording didn't meet a specific regulatory requirement in Germany. We had to work with Legal to rewrite the question flow while preserving the UX quality — which took two more rounds of review.
I spent too long trying to build consensus on the value-first approach internally before running an experiment. We debated the bank-link timing for 6 weeks before I pushed to just test it. The test ran for 2 weeks and gave us a clear answer. I should have moved to experimentation 4 weeks earlier. The lesson: when you have a reasonable hypothesis and the risk of a bad outcome is reversible, run the test and let data end the debate.