Product evidence
Asking is an intervention — the quiz is the commitment, not the data. But effort only converts to attachment if the user finishes and sees the receipt.
The research, distilled
Across 40,000+ analyzed households, families asked a single purchase-intent question — no ad, no pitch — bought cars at 3.3% versus 2.4% for those never asked. The mere-measurement effect has since been meta-analyzed across 116 tests (d+ = 0.24): small, consistently real, and strongest (d+ = 0.29) when people predict their own behavior.
Morwitz, Johnson & Schmittlein (1993), Journal of Consumer Research 20(1); Wood et al. (2016), PSPR 20(3)Residents asked to predict whether they'd volunteer three hours for the Cancer Society over-predicted their virtue — then 31% actually volunteered when called back, versus 4% of people simply asked directly. Sherman called it a self-erasing error. Every "What's your goal?" tap is that instrument at scale, which is why the user should tap it themselves, not get it preselected.
Sherman (1980), Journal of Personality and Social Psychology 39(2)Builders bid 63% more for boxes they assembled themselves — but when people failed to finish, or watched their work get taken apart, the effect vanished: "labor leads to love only when labor results in successful completion." A quiz abandoned at question fourteen is worse than no quiz: effort charged, receipt never issued. The payoff screen is the completion artifact, not decoration.
Norton, Mochon & Ariely (2012), Journal of Consumer Psychology 22(3)Moving account creation from the front of the flow to after the first lesson raised daily active users by about 20%, Duolingo reports; softening the wall further — a quiet "Later" instead of a red discard button — added another 8.2% DAU. The commitment questions (goal, daily minutes) stayed up front; the extraction question ("give us your email") moved after value.
Gotthilf, First Round Review (company-published A/B results)Everything in this deep-dive
Original data
We sampled 35 complete onboarding flows plus 18 individual screens from top iOS apps via Mobbin — August 2026 — and tallied every one by hand. Explore the patterns, then see how each flow compares to the overall takeaway.
The most respectful personalization flow sampled: 3 questions with an honest "1/3" step count, Skip on every screen including the notification primer, zero account requirement.
Short mascot-guided quiz that pays off with a FREE generated starter plan of 7 tiny habits — the quiz-to-plan payoff with no gate and no paywall.
Delivers a full lesson before asking anything; signup is "CREATE PROFILE / LATER" framed as saving progress, with a streak-commitment device instead of a paywall.
One clear goal question, then immediate value — a usable 5-minute session offered during onboarding with a "Not Now" escape hatch.
Full content feed before any ask; signup is fully deferrable and the notification ask is a soft dismissible in-app banner, not a system-dialog ambush.
You build a real budget plan first, then sign up to save your progress — and the paywall carries an unusually honest explicit auto-renew consent checkbox with the renewal date.
Privacy-forward outlier: "Your data, your choice" granular consent toggles (analytics OFF by default) and a "Maybe later" on the account pitch — respect encoded as defaults.
The maximalist quiz done well — privacy reassurance up front, milestone encouragement, empathetic sensitive questions, a "79% match" payoff — but it is still a 22-screen, no-skip quiz that ends in an account gate.
Offers Skip and a live percentage progress bar (rare courtesies), but still ends with the classic hostage move: "Create an account to save and get your recommended programs".
The most persuasion-dense flow in the sample: a 23-screen quiz interleaving social proof, Harvard/Oxford/Cambridge authority cards, and empathy feedback, closing with a finger-signed commitment contract — questions serving the sale, not the user.
The classic quiz-plan-gate sequence: 6 questions, then "Your plan is ready. Unlock it now" — your answers converted into a hostage held behind signup.
"Your plan is ready. Unlock Calm for free" — the plan-picker paywall arrives before any delivered value, though the "Remind me 2 days before renewal" toggle is a genuine trust device.
The opposite failure: a bare name/email/password form into an empty "You haven't enrolled in any courses (yet)" state — no questions, no personalization, no value before or after signup. 2015-era onboarding.
A raw ATT tracking dialog with no primer as effectively the first interaction — the only naked permission ambush in the sample, and ideal "before" footage against any primed flow.
Signup before any value, then straight into a trial-timeline paywall with annual preselected — account and monetization arrive as one shove.
Hard native StoreKit subscription sheet ($19.98/yr) before any habit or value is established — no questions asked, just the toll booth.
How to read this data: Mobbin curates top-tier, design-forward iOS apps, so percentages describe what leading apps do, not the App Store at large. Queries are relevance-ranked, not random: quiz-seeking queries inflate the 46% quiz rate by construction, so cross-cutting stats (progress-bar style, priming rate, skip rate within quiz flows) are more trustworthy than base rates. Flow previews are evenly-spaced screen subsets, so intermediate screens like "preparing your plan" loaders may go unobserved — the 17% fake-work figure is a floor. Tallies made from visual inspection of preview images on 2026-08-10. Full per-flow observations are in the audit.
Apply it to your app
This prompt distills everything above into instructions for an AI coding session (Claude Code, Cursor, or similar). It interviews you about your app first — so nothing changes until it understands your context — then audits every question against the research and implements the fixes with your design system.
You are a senior product engineer applying peer-reviewed research to my app's onboarding — specifically the question/personalization layer. Grounding: asking is an intervention (mere-measurement effect: one purchase-intent question raised real car purchases 3.3% vs 2.4% across 40k+ households — Morwitz 1993; meta-analyzed at d+ = 0.24 across 116 tests, strongest for self-prediction questions), commitment via active choice (a goal the user taps themselves sticks harder than one preselected), the IKEA effect's boundary condition (effort converts to attachment ONLY on completion — an abandoned quiz is worse than no quiz), the personalization paradox (Aguirre 2015: personalization from covertly-collected data tanks engagement; overt asking flips it positive), and Duolingo's published A/Bs (moving signup AFTER value: ~+20% DAU; softening the wall further: +8.2%).
BEFORE YOU CHANGE ANYTHING, ask me and wait for answers:
1. Walk me through your current first-run flow screen by screen (or paste the route/component list). Where is account creation? Where's the first moment of real value?
2. What can your product actually personalize today based on user answers (content, plan, difficulty, defaults)? Be honest — this determines how many questions we keep.
3. What's your stack and is there an existing onboarding/quiz component system?
4. What permissions do you need (notifications, tracking, location), and what user-visible job does each perform?
5. B2C habit product, utility, or B2B tool? (The right question count differs.)
THEN audit the flow against these rules and show me the plan before coding:
- Every question must pass the double test: something visibly changes based on the answer AND the user sees it change. Questions that only feed a marketing profile get deleted or deferred.
- Keep it to ~3 questions in the critical path (goal, context, constraint). Question one is the commitment question — the user's goal, tapped by them, NOT preselected (this is the one screen where defaults are wrong).
- Ship the receipt: a payoff screen — "Here's your plan, built from your answers" — that quotes their goal and visibly uses their answers. This is the completion artifact the IKEA effect requires. No payoff screen = you charged effort and issued nothing.
- Move signup after value, reframed as saving: "Save your plan" instead of a front-door toll booth. Support Apple/Google sign-in.
- Prime every permission ask with a benefit screen that quotes the user's own answers ("Tuesdays, 7am — want a reminder?"), then fire the system dialog. Never ambush.
- Skip must exist on every question, with sensible defaults on skip.
- No hostage sequence: never gate the quiz results behind payment. If a paywall follows onboarding, the user gets their personalized value first.
- Progress: honest step count ("1 of 3") — short flows disclose length; hiding length is a long-quiz manipulation pattern.
THEN implement incrementally with my design system, behind a flag if possible. Measure: flow completion, question-level drop-off, activation (first value moment), signup rate at the new position, permission grant rate, and day-7 retention. Note that asking goal questions itself shifts behavior (mere-measurement), so any "goal-setters retain better" readout is confounded — if we want truth, randomize whether the question is asked.
Close with the list of questions you removed and where their data collection moved (or why it died).
See it, click it
Stride, a fictional running app: click through the 17-screen toll-booth flow, then the 6-screen version where your answers visibly build the plan.
From Build With Kris
Subscribe to catch the teardown when it drops — the science, the audit, and the Stride rebuild, screen by screen.