Redesigning the Onboarding Experience
How I reduced drop-off by 40% by rewriting every screen in a 12-step onboarding flow for a fintech app.
- Role
- Lead Content Designer
- Timeline
- Jan – Mar 2024 · 3 months
- Tools
- Figma, Notion, Dovetail, Maze
- Company
- Fintech Startup
Overview
This project began with a single alarming metric: 62% of new users were abandoning our 12-step onboarding flow before reaching step 4. The product team had already ruled out architecture issues — the problem was the language.
I was brought in as the lead content designer to audit every screen, identify the root causes of drop-off, and rewrite the flow from the ground up. The work spanned three months and touched every layer of the product's first impression.
The Problem
Users weren't abandoning because the product was hard to use. They were abandoning because the product was hard to trust — and hard to understand. Screens dense with legal language, passive voice, and noun stacks were creating friction at every step.
Process
Content Audit
I audited all 12 screens against three criteria: Clarity (could a user explain what this screen wants?), Trust (does the language feel safe?), and Momentum (does it create a sense of progress?).
Eight of twelve screens failed on at least two criteria. The worst offender was screen 3 — a data permissions screen written entirely in passive voice and noun stacks.
Audit scoring matrix across all 12 onboarding screens
Support Ticket Analysis
I pulled 340 support tickets tagged 'onboarding' from the previous quarter. Three questions appeared in 80% of them: 'Is this safe?', 'Why do you need my information?', and 'How long is this going to take?'.
These three questions became my design principles. Every rewrite had to answer at least one of them — ideally proactively, before the user had to ask.
Clustered support ticket themes from Dovetail analysis
Rewriting & Pair Writing
I rewrote every screen using a consistent structure: a headline with one job in active voice, a body that answered the most likely question proactively, and a CTA that described what would happen next — not just 'Continue'.
For screens with legal content, I pair-wrote with the compliance team in Figma to find plain-language alternatives that still passed their review. It was a negotiation, but a productive one.
Before/after comparison for screen 3 — data permissions
Outcomes
Drop-off at step 4: 62% → 22%
Overall onboarding completion: 38% → 64%
Support tickets tagged 'onboarding': down 55% in 30 days
A support ticket that read: 'I actually understood what you were asking for and why. That never happens with these things.'