Onboarding
How many screens, what to ask, where the permission prompts go, and whether sign-up belongs before or after the first moment of value.
How many screens
The one number that matters is how many screens stand between the first open and the first thing the app does for the user. Completion runs Not verified (Retired: no traceable primary study. Do not use as evidence.) Unverified attribution (retired) 2026 and collapses to Not verified (Retired: no traceable primary study. Do not use as evidence.) Unverified attribution (retired) 2026. That is not a small difference: it is the difference between most of your installs seeing the paywall and most of them never getting there.
Three to five is the published band, and the reason is not attention span. It is that every screen before value is a screen the user is paying for with trust they have not yet been given a reason to extend. Screens after value are paid for by the value.
| Metric | Figure | Source |
|---|---|---|
| Onboarding benchmark | Not verifiedRetired: no traceable primary study. Do not use as evidence. | Unverified attribution (retired) 2026 |
| Onboarding benchmark | Not verifiedRetired: no traceable primary study. Do not use as evidence. | Unverified attribution (retired) 2026 |
| Onboarding benchmark | Not verifiedRetired: no traceable primary study. Do not use as evidence. | Unverified attribution (retired) 2026 |
| Onboarding benchmark | Not verifiedRetired: no traceable primary study. Do not use as evidence. | Unverified attribution (retired) 2026 |
Unverified attribution (retired) 2026: Retired September 2026. Do not quote these onboarding numbers or treat them as evidence.
What to ask, and why quiz onboarding works when it does
Question screens — "What is your goal?", "How often do you…?" — appear in most of the top-grossing apps in health, education and finance. There is no published measurement of their effect in isolation, so this section is practitioner consensus and says so. What the consensus agrees on:
- Questions earn their screen only if the answer changes something visible: the plan the app builds, the number on the paywall headline, the first screen after onboarding. A question whose answer is never used is intake, and intake is what the below-40% band is made of.
- Each answer should be one tap. A screen with a text field is a screen with a keyboard, and a keyboard is where people leave.
- Three to five questions, then a moment that visibly uses them — a personalised plan, a projected result, a chart — and then the paywall. The paywall's job is easier when its headline can quote the user's own answer back.
- Reassurance beats information. "You're in the top 20% of people who start this" outperforms a paragraph about features in every practitioner account, though no one has published a controlled measurement.
Measure it
Sendonboarding_step with a 1-based step and a name on every screen. The Funnel report then shows which screen loses people, not just that onboarding does. That is the only way to know whether a question is earning its place.Progress and sign-up
Progress indicator
The cheapest change in this playbook: a progress indicator cut first-step drop-off Not verified (Retired: no traceable primary study. Do not use as evidence.) Unverified attribution (retired) 2026. People finish what they can see the end of.
Sign-up
Three fields instead of seven abandons Not verified (Retired: no traceable primary study. Do not use as evidence.) Unverified attribution (retired) 2026. Sign in with Apple fills them for you. And where the product allows it, sign-up belongs after the first moment of value, not before — an account is something people create for a product they already want.
Permissions: notifications and tracking
Never on first launch. A prompt that arrives before the app has done anything is answered by reflex, and the reflex is no. Ask for notifications at the moment a notification would obviously help — after the user sets a reminder, a goal, a delivery — with one screen of your own explaining why before the system prompt.
For App Tracking Transparency, expect about a third to allow it: 35% (Q2 2025 industry average; 34% in 2023, 34.5% in 2024) Adjust 2025. That figure counts only people shown the prompt; across all devices it is far lower, because many never see it. It also means attribution will be partial whatever you do, which is one reason the App Store's own download figures — which Broctic reads — matter more than an SDK's.
Adjust 2025: Vendor of attribution; sample size not stated. The same metric with an 'all devices' denominator (Singular, 2024) comes out around 14%, so the denominator is the whole story with ATT figures.
What Apple says
Apple 2026 asks that onboarding be brief, skippable where possible, and that it not be used to display licensing details or feature lists — the app should teach by doing. It is guidance, not measurement, but it is consistent with the numbers above.