Why onboarding is important in D2C apps.

I want to be precise about what was wrong, because the flow was not bad. Basics, body measurements, activity, primary goal, a target date, diet, "have you tried other apps," "where did you hear about us." Haptic ruler pickers. An editable plan. A safety cap on aggressive calorie targets. Nobody complained about it, which is exactly why it took me three months to see it as a problem.

The moment it became one was July 2026. We had decided to charge, and the paywall had to live somewhere. Putting it at the end of a flow that built no commitment would have been asking strangers for money.

Two sources, one week. First, the subscription-app data. RevenueCat's benchmarks across 115,000 apps put about half of all paid conversions on day zero, in the first session, and show hard or trial-inclusive paywalls converting around five times better than freemium.¹ Their onboarding writing makes the counterintuitive point directly: most onboarding is too short, because value and projection screens build momentum rather than friction.² Noom runs over a hundred screens and converts through progressive commitment.

Second, I captured the complete first-run of Cal AI, MacroFactor, and MyFitnessPal, screen by screen, and laid ours next to them. Cal AI's 34 screens are engineered for one moment: a soft paywall at the peak of commitment, after your name is on a plan with a goal date.³ MacroFactor pre-justifies its price with an honest "here's our business model" screen. MyFitnessPal extracts data and dead-ends at "create account."

The one-line verdict I wrote at the top of the teardown: ours is the most polished and honest of the four, and the most commercially naive. We had the credibility edge, peer-reviewed sources on the plan reveal that none of the slick apps could match. What we lacked was structure: the mechanics that turn minutes of answering questions into a commitment.

v2 went live on July 6 with the first paywall. Café Cal has had paying subscribers since. That is the outcome the problem statement asked for, and I am not going to dress it up with a conversion percentage from a base of a few hundred installs.

What I will report is what the August 10 audit found in this exact flow, because it is the honest second half of the story. The paywall had no back, skip, or close, and quitting at it destroyed all twenty answers. The projection rendered an empty card for four of the six goals. The money screen had the old brand name on it. All three were fixed within days. The barrier question became multi-select on August 22 after watching people pick one reason when they clearly had three.

Length is not friction if every screen gives something back. I had been proud of eight questions. The research and the teardown both said the same thing: people will answer twenty if the output is visibly theirs, and the projection they see at step fifteen is what makes them believe the plan at step twenty-one.

Build the paywall's exit before its entrance. I designed the moment to ask for money and forgot to design the moment someone says no. The audit caught it. Next time it is in the first sketch.