How many questions should an onboarding personalization quiz ask?

Across 119 quiz apps, the average is 3.9 captured question screens with a p90 of 10 and a long tail to 65.[1] Don't anchor on the median of 1 — it reflects apps with only one or two captured quiz screens, not one-question quizzes.[1] For a real personalization quiz, plan for roughly 4 to 10 questions and validate depth against your vertical.

Quiz apps average 3.9 question screens (p90 of 10, max 65) — Lazyweb Research, July 2026.

Lazyweb Research · n=119 · Published 2026-07-07

onboardingsignupux-patternsexperimentsmobile

The distribution

Across 119 quiz apps, captured question screens per app average 3.9, with p90 at 10 and a maximum of 65.[1]

Question screens per appApps
1 captured62 of 119[1]
2–433 of 119[1]
5–911 of 119[1]
10+13 of 119[1]

The median is 1 only because 62 apps have a single captured quiz screen. The methodology explicitly warns against headlining that median — use the average, the p90, and the leaderboard to characterize real depth.[1]

How to apply it

If you want light personalization, 2 to 4 questions matches the modal cluster (33 apps).[1] If you want a full quiz-to-plan flow, the depth players sit at 10+ (13 apps).[1] Going past ~10 questions puts you in the top decile — justify every screen, because each one is a drop-off opportunity.

Map each question to a downstream personalization it drives. If a question doesn't change the plan, feed, or paywall, cut it.

Caveats

These are captured question screens from LLM vision descriptions (~94% coverage), not a guaranteed count of every question a user answers.[1] Apps with a single captured screen may run longer quizzes that weren't fully captured, so the low end is a floor. Use the average (3.9) and p90 (10) as your planning band, not the median.

The numbers

StatComputed from
avg 3.9 question screens, median 1, p90 10, max 65; 62 apps 1, 33 apps 2-4, 11 apps 5-9, 13 apps 10+quiz_question_count_distribution across 119 apps
Methodology. Universe: 119 quiz apps in the Lazyweb vision corpus. Method: count of distinct captured quiz-question screens per app via regex over vision_description, July 2026. Caveat: median of 1 reflects single-capture apps, not one-question quizzes; use average and p90.

Sources & citations

  1. [1] Lazyweb Research analysis of 119 apps (Lazyweb onboarding-quiz corpus), July 2026. Per-app distinct captured quiz-question screens; avg 3.9, median 1, p90 10, max 65; buckets 62/33/11/13. Median is depressed by single-capture apps; do not headline it.

Source: Lazyweb Research — proprietary analysis of real, in-market app screens. Cite as Lazyweb Research, 2026-07-07.

Related questions

Explore the underlying screens, flows, and A/B tests inside Lazyweb. More research