Onboarding flow
The screen sequence a new user goes through between first launch and becoming an actual user of the product — typically including some combination of welcome, account creation, permissions, personalization quiz, paywall, and first task. The onboarding flow is the product’s rails to its Aha moment.
What the data actually says about length
The default advice is “keep it short.” Per the survey in I Studied 1,460 Onboarding Flows (video) (~1,460 flows / ~986 apps), this advice doesn’t match the data:
- Average length: 25 screens.
- Longest categories: finance, music/health/fitness, education — finance dominates the top of the distribution (7 of 10).
- Several of the longest flows belong to the most successful products — Duolingo’s hits ~60 screens before signup; Bite Pal’s hits 61.
- Web onboarding is ~21% shorter than iOS because mobile has to absorb additional permission and paywall screens.
The mechanical takeaway: length isn’t the variable. Whether the flow has texture, pacing, and felt progress is the variable. A 60-screen Duolingo flow with first-lesson satisfaction beats a 4-screen flow that drops the user into an empty state.
The recurring structure
The Mobbin survey identifies the same shape across many strong flows:
sign up → set up your account → aha moment
…with onboarding screens interleaved to bridge the gaps.
The patterns that distinguish good from forgettable
(All from I Studied 1,460 Onboarding Flows (video) — see that page for the worked examples.)
- Sell the outcome, not features — show the product running; let users try the core experience before signup if possible.
- Make it feel human — small founders’ notes, birthday acknowledgments, CEO videos at milestones; signals of intentional craft.
- Personalize, but make it earn its keep — 23% of apps personalize during onboarding (only 7% of AI apps); the strong ones show the user what their answers built immediately.
- Place the paywall after personalization — 22% of apps run a paywall during onboarding; the better ones pair it with the personalized plan and social proof.
- Make long onboarding feel short — animations, mascots, copy texture, completed-lesson satisfaction.
- Teach in context — tool tips, real-time validation, populated empty states, persistent checklists.
- Prime before OS prompts — a custom screen before the iOS notification permission popup raises accept rates.
- Split signup forms across screens — e.g., House +15% conversions.
A/B-tested deltas worth remembering
| Change | Delta |
|---|---|
| Headspace: multi-intent goals (pick more than one) | +10% trial conversion |
| Dollar Shave Club: conversational quiz copy | +5% subscriptions |
| Grammarly: tailored pricing recommendations from quiz | ~+20% plan upgrades |
| Mural: popups/banners → six-step checklist | +10% week-1 retention |
| House: signup form split across screens | +15% conversions |
These aren’t guidelines — they’re examples that flow design has compounding leverage when done well.
The prescriptive complement — Tim Gabe’s five patterns
The Hidden App Growth Killer (video) is the prescriptive pair to the Mobbin descriptive survey. It opens with a sharper framing claim:
Apps lose 77% of users within 3 days — not because they lack features or have bad marketing. Because of their onboarding.
Tim Gabe proposes five patterns, each attached to a named psychological principle and a real product:
| Pattern | Mechanism | Product |
|---|---|---|
| 1. Deliver value in the first 60 seconds | Aha moment / eureka effect | Breathwork drops users into a guided session immediately |
| 2. Three steps, no input fields | Zeigarnik effect (motivational variant) + chunking | Stomper’s three-step welcome — “sometimes onboarding isn’t about doing, it’s about preparing the user smoothly” |
| 3. Let users interact with the core flow | Trial-and-error learning | Sudoku gives you a puzzle with hints rather than teaching the rules |
| 4. Personalize the product feel early | Familiarity principle | Speechify sets up voice/tone/speed preferences during onboarding |
| 5. Show visible progress | Goal-gradient effect + Peak-end rule | Marathon’s simple top progress bar |
Together with the descriptive Mobbin survey, the picture is now:
- Descriptive layer — what high-performing onboarding looks like across ~1,460 flows (length distribution, paywall placement, personalization frequency, structural patterns).
- Prescriptive layer — five specific moves, each attached to a named mechanism, that aim the flow at the Aha moment and away from the early-dropoff cliff.
The descriptive survey says length isn’t the variable; texture and pacing are. The prescriptive video sharpens that to: deliver value in 60s, chunk what remains into achievable steps, let users interact with the real thing, give them light personalization, show progress. The two don’t conflict — Tim’s five patterns are roughly how to do well on the dimensions Mobbin’s survey measures.
Hodent’s onboarding-plan methodology
Per The Gamer’s Brain (book) (Chapter 13), Celia Hodent adds a third layer beneath the descriptive (Mobbin survey) and prescriptive (Tim Gabe’s five patterns) layers above: a concrete design tool for actually building the sequence. She opens with the same urgency the 77%-in-3-days stat above captures, citing SteamSpy data directly: “it is not rare to see that about 20% of the audience is already gone after only one hour of play,” which is why she calls the first hour critical to the first-time user experience (FTUE).
Her method: build a spreadsheet listing every element a player must learn, one per row, with nine columns — Category (system it belongs to), Priority (how important), When (a rough integer bucket, e.g. “1” = first mission or first 15 minutes), Tutorial order (a unique sortable number, e.g. 1.1, 1.2, 2.1), Difficulty (simple/moderate/difficult), Why (what makes learning this meaningful to the player — e.g. “If I do not learn how to craft a weapon, I will be killed by the monsters”), How (UI-only, learning-by-doing, dynamic tutorial text), Narrative wrapper (the story that will carry the sequence once sorted), and UX feedback (early test results or anticipated friction points). Sorting by tutorial order reveals the plan. She illustrates the method with a worked example using features from Fortnite.
The design principle underneath the spreadsheet ties directly back to Cognitive load: distributed learning is more efficient than massed learning, so the harder a feature is to teach, the fewer other features should be taught alongside it — and two difficult features should never be taught back-to-back. This gives the “texture and pacing, not length” finding from the Mobbin survey above a concrete production method: the spreadsheet is how texture and pacing actually get designed, rather than just observed after the fact.
The counter-pattern — when progress cues backfire
The same Tim Gabe video adds the important boundary condition. Pattern 5 (show visible progress) only works when the user is already motivated to reach the destination. If they haven’t felt value yet, revealing the step count is demotivating:
If someone sees 12 steps before they’ve even felt a single hit of value, they’re gone.
TypeForm omits step indicators by default in long-form templates for exactly this reason. The general rule: progress cues motivate the committed user and discourage the uncommitted one. Show them only after the user has crossed the Aha moment threshold.
A second counter-pattern — when effort is the feature
The Twisted Psychology Behind Top 1% Apps (video) complicates Tim’s own earlier “value first, friction later” rule. Starcrossed asks for birth date, exact birth time, birthplace, and other inputs before generating a soulmate prediction and revealing a paywall. Tim argues that the long quiz is not incidental friction: it makes the result feel calculated, personal, and earned by stacking the Barnum effect with Effort justification.
The two Tim sources point in opposite directions only if every step is treated as the same kind of friction. A better diagnostic distinguishes:
| Type of effort | User’s model | Expected effect |
|---|---|---|
| Investment | “This answer will make my result more relevant.” | Anticipation, ownership, possible Effort justification |
| Tax | “The app is making me work before it proves value.” | Cognitive load, skepticism, abandonment |
This does not establish that longer quizzes convert better. The source reports no A/B test, and Aronson and Mills’s 1959 experiment studied severe initiation into a group rather than digital forms. The falsifiable design hypothesis is narrower: a step can earn its place when the user already wants the destination and understands how the input improves it. Teams should measure both downstream conversion and abandonment across the whole flow; higher conversion among quiz survivors can conceal a worse start-to-finish rate.
When onboarding is a tax, not a feature
The video closes with a useful counter-pattern. For products whose value is obvious on first use — Mobbin (a design library), AI chat apps (the first prompt is the value) — onboarding is friction without payoff. Those products are better off letting users get in fast and not adding screens that delay the first interaction.
Diagnostic question: does this product reveal value quickly on its own? If yes, the onboarding flow should be near-zero. If no, the patterns above are the playbook.
Cultural variation
Users in eastern markets tend to be more comfortable with information-heavy interfaces.
Best-in-class onboarding is locale-dependent. What feels cluttered to one audience can feel efficient to another, which is why blindly copying a Western “minimalist” flow can underperform.
Related
- Aha moment — the destination an onboarding flow exists to deliver
- Social proof — frequently used in pre-paywall pages and outcome screens
- Scarcity principle — the one-time-offer mechanic stacked onto post-personalization paywalls
- Cognitive load — the budget all onboarding-flow decisions are paid out of
- Zeigarnik effect — the open-loop mechanism behind multi-step welcome flows
- Goal-gradient effect — progress bars; the boundary condition where revealing step count backfires
- Peak-end rule — how the remembered shape of onboarding shapes return behavior
- Familiarity principle — light personalization as a familiarity device
- Barnum effect — perceived personalization from a broad but personally framed result
- Effort justification — the investment-versus-tax boundary for quiz effort
- I Studied 1,460 Onboarding Flows (video) — descriptive survey
- The Hidden App Growth Killer (video) — prescriptive five-pattern playbook
- Celia Hodent — the onboarding-plan spreadsheet methodology; the FTUE framing
Sources
- I Studied 1,460 Onboarding Flows (video) — descriptive survey of ~1,460 flows
- The Hidden App Growth Killer (video) — prescriptive five-pattern playbook against the 77%-in-3-days dropoff
- The Gamer’s Brain (book) — the onboarding-plan spreadsheet methodology; SteamSpy FTUE churn stat; Fortnite worked example
- The Twisted Psychology Behind Top 1% Apps (video) — long quiz onboarding as a possible effort-justification amplifier; Starcrossed case and investment-versus-tax boundary