← All notes

How we moved onboarding activation from 17% to 58%

2 min read Onboarding · Metrics · Mobile

When I joined Startup Studio in December 2023, onboarding activation sat at 17%. Out of every hundred new users, eighty-three left before reaching the first useful action in the product. Ten weeks later, the number was 58%. Here is what we did — and what we did not.

Analytics first, mockups second

The first week was not spent in Figma. It was spent in Amplitude and Hotjar. I pulled the screen-by-screen funnel and overlaid a heatmap. Two screens were honestly eating 60% of the drop-off: phone verification and choosing interests from a list of 24 categories.

This step sounds obvious, yet it is exactly where most “onboarding redesigns” fail. The team sees a bad number and starts repainting illustrations. I have watched teams burn months that way.

Four decisions

  1. Verification went to the background. Previously the user typed the SMS code on a dedicated screen and waited. Now they enter the product immediately and verification runs in the background. If the code does not arrive, a soft top notification — not a blocker.
  2. Interest selection became optional. From a mandatory step we turned it into a “bonus” after the first useful action. Anyone who wants a tuned feed can still do it, but not upfront.
  3. Onboarding shrank from 8 to 5 screens. Every “but we have to tell them about X” was weighed against drop-off. Half did not survive the conversation.
  4. A progress bar wherever a step remained. A simple top bar. Without it, the user has no idea how close to the end they are.

What we did NOT do

  • We did not touch the visuals. The design system stayed.
  • We did not add animation. No success-state particle bursts.
  • We did not rewrite copy from scratch. Three headlines changed; the rest stayed.

When a metric moves, the temptation is to credit everything. In reality, 2–4 decisions move it. The other thirty are noise.

What I would do differently

We rolled the new onboarding to 100% after a two-week A/B at 50/50. Today I would run the A/B for a month — D7 retention only stabilised at the end of week two, and a couple of days earlier could have given a falsely flat trend.

Also: I would have shipped a post-onboarding nudge towards interest selection from day one. We added it a month later, after seeing that 70% of users skipped it and never returned.