Design case

Onboarding redesign for a subscription AI photo app

This is a design exploration for a subscription AI photo app, whose name is withheld. It compares five onboarding concepts that each test a different mechanism before the paywall. No concept has run on real traffic yet, so the case shows the method and the outputs and reports no results.

Published November 12, 2026. Part of the open-source growth skills.

Problem

The question

The app had an onboarding of five screens: a splash, photos and video, prompt to video, a ready set, and reviews. The screens sold the product before sign-up but did not use the person's own choices, and the next step was a paywall.

The question was which onboarding mechanism makes people more ready to start a trial.

Skills applied

Skills used

onboarding-web2app-funnel
The concept set and the brief to the design tool.
View on GitHub
paywall-audit
The handoff screen into the paywall, covered in the paywall case.
View on GitHub

Method

How the work went

Map the existing flow and write down the job of each screen. Keep the existing flow on the board as the baseline.
Pick one strategic question: should the person see value, invest effort, or see themselves in a result before the paywall?
Write five concepts, each with a different mechanism and a one-sentence hypothesis, then build each as an interactive prototype on a design board.
Concept A, quiz and personalisation: two questions (goal and styles), a wait screen while the plan is built, then a personal plan that leads to the paywall.
Concept B, interactive demo: every screen demonstrates the product, the person drives the demo, and the last screen repeats their choice and says plainly that a plan comes next.
Concept C, stories and reviews: the shortest route to the paywall, with full-screen stories that look like the ad the person came from.
Concept D, profile and likes: the onboarding looks like the person's future profile, which they assemble, with reactions arriving on the photos.
Concept E, avalanche of likes: reactions are the hero and the photos are the excuse. The person taps a post action and a stream of reactions builds up.
Run a second pass on each concept that makes the person do something (pick styles, drive the demo, assemble the profile, press post). Concept E got a third pass that reused the interaction ideas from the paywall work.

Output

What was produced

One baseline and five concepts, 11 interactive prototypes in total across three passes.
A hypothesis per concept, a screen list with the job of each screen, and the point where each concept hands off to the paywall.
Social proof placeholders, marked as needing real figures before any release.

Why these five

They test different mechanisms: effort (A), proof by demonstration (B), ad continuity (C), identity (D) and social reaction (E).
D and E share a mechanism, so they would run as a pair after the first round.

Results

Results: not measured yet

This case is a design exploration. No version has been run on real traffic, so there are no conversion, trial start, retention or refund figures to report. The numbers on any mock-up are illustrative placeholders.

No concept has been tested with real traffic.
The source material has no measured conversion, trial start, retention or refund figures.
Any ratings or counts on the prototypes are illustrative placeholders, not claims about a real product.

Planned test

Split the same paid traffic across concepts A, B and C first. The primary metric is trial starts per ad click, with refund rate and day 1 and day 7 retention as guardrails. The stop rule is written before launch.

Request an audit Back to the open-source skills

Let’s talk