---
title: "App onboarding that converts: what Cal AI, QUITTR and Glowly do before the paywall"
canonical: "https://asoitis.com/blog/app-onboarding"
description: "App onboarding that converts in 2026: Cal AI, QUITTR and Glowly compared screen by screen, Apple's rules on permission screens, and day-0 conversion data."
kind: "blog-post"
category: "Growth"
datePublished: "2026-10-10"
dateModified: "2026-10-10"
keywords:
  - "app onboarding"
  - "app onboarding examples"
  - "app onboarding flow"
  - "mobile app onboarding best practices"
  - "onboarding quiz app"
  - "subscription app onboarding"
  - "permission priming ios"
---

# App onboarding that converts: what Cal AI, QUITTR and Glowly do before the paywall

> Onboarding that converts asks questions with an obvious purpose, gives people a personal result, primes each permission before iOS asks, and ends on the paywall while motivation is highest. In our recordings, Cal AI ran 29 screens and QUITTR more than 35 before the paywall, both delivering a result first; Glowly took three screens and withheld the result until payment. Apple's rules limit how you ask for permissions.

*Published 2026-10-10 · last updated 2026-10-10 · by Ahmed Gagan for ASOitis.*

Onboarding is where a subscription app earns most of its money: RevenueCat's 2026 benchmark found nearly all trial starts happen on the day of install. Apple's design guidelines say onboarding should be "fast, fun, and optional". The category leaders run 30-screen quizzes. We recorded three of them, screen by screen, to see what actually happens between "Get Started" and the price — and which of their tricks Apple's rules allow.

## Three onboarding flows, side by side

We installed each app and recorded every screen from first launch to the paywall. Cal AI and QUITTR were recorded on October 6, 2026; Glowly for our earlier teardown in 2026.

**Onboarding before the paywall, from our recordings**

|  | Cal AI | QUITTR | Glowly |
| --- | --- | --- | --- |
| Screens before the paywall | 29 | 35+ | 3 |
| Time for our tester | About 2¼ minutes | About 2 minutes 50 seconds | About 90 seconds |
| Questions | 16, mostly inputs to a calorie formula | A numbered 10-question quiz with “Skip test” | None — take or upload a selfie |
| Result before the price | Yes: a 1,938-calorie plan with macros | Yes: a dependence score and a quit date | No: the selfie appears on the paywall, unscored |
| Permission prompts | 3 (Apple Health, notifications, tracking) | 2 (tracking, notifications) plus Apple's rating prompt | Not noted in our teardown |
| Account before the paywall | Yes — “Save your progress” | No | No |
| Paywall | Hard, 3-day trial on annual | Hard, no trial | Hard, weekly pre-selected |

Full screen-by-screen breakdowns are in our [Cal AI](https://asoitis.com/case-studies/cal-ai), [QUITTR](https://asoitis.com/case-studies/quittr) and [Glowly](https://asoitis.com/case-studies/glowly-ai) teardowns.

## 1. Ask questions people can see the point of

Cal AI's first questions — sex, workouts per week, birth date, height, weight — are all inputs to the calorie maths, so each has an obvious reason to exist. Motivation questions (“What's stopping you from reaching your goals?”) come only after the maths ones. QUITTR numbers its questions (“#2 of 10”) so the length is visible, and puts a “Skip test” link on every screen so it stays voluntary.

Both also slip in “Where did you hear about us?”. For apps that grow through creators, self-reported attribution is often the only way to see which videos drive installs. Adapty reports one fitness client whose conversion rose “by around 10%” after adding that question, which it says made the app feel more expert-led — a single anecdote, not a benchmark.

## 2. Deliver a result — or deliberately withhold it

The two longer flows end with something personal before any price: Cal AI shows a full plan (1,938 calories, 121 g protein, 242 g carbs, 54 g fat, all editable), and QUITTR shows a dependence score and “You will quit porn by: Jan 4, 2027”. The user leaves onboarding already holding a piece of the product. Apple's HIG points the same way: “Prefer letting people experience your app or game before prompting them for ratings or purchases.”

Glowly does the opposite. After a selfie and an “Analyzing Your Face” progress sequence, the paywall opens with your own photo at the top — and the score behind the price. It's a curiosity gap: refusing to pay means not seeing your result. It's fast and it works for a scan-shaped product, but it gives the user nothing to judge the app by, and it leans on a weekly plan pre-selected by default.

> According to ASOitis's recordings of Cal AI, QUITTR and Glowly, the biggest difference between top subscription onboardings isn't their length — it's whether the user leaves with a result before the price: Cal AI and QUITTR hand over a personal plan and a date after 29 and 35+ screens, while Glowly withholds the score its whole three-screen flow was built to produce.

## 3. Prime permissions — within Apple's rules

Cal AI primes every permission before iOS asks: a mock-up of the Apple Health sheet showing exactly where to tap “Turn On All”, and an illustration of the notification alert with a hand pointing at Allow. Priming is allowed. How you do it is regulated by Apple's [Human Interface Guidelines on privacy](https://developer.apple.com/design/human-interface-guidelines/privacy):

**Apple's rules for a custom screen before a permission alert**

| Rule | Apple's words |
| --- | --- |
| One button, clearly opening the alert | “Include only one button and make it clear that it opens the system alert.” |
| Don't label it “Allow” | Use “a term like “Continue” or “Next” to title the single button” |
| No way out before the alert | “Don't provide a way for people to leave the screen or window without viewing the system alert — like offering an option to close or cancel.” |
| Tracking (ATT) is stricter | No incentives, no “screen or window that looks like a request”, no “image of the alert” — such screens “will lead to rejection by App Store review” |
| Timing | “Avoid requesting permission at launch unless the data or resource is required for your app to function.” |

Apple enforces the spirit of this in review. In May 2025 an app was rejected under Guideline 5.1.1 because “a custom message asks users to reconsider their decision after they deny the permission request” — a re-ask that came from a third-party SDK. Prime once, with a Continue button, and accept the answer. If someone declines, Guideline 5.1.1(iv) asks you to “provide alternative solutions” where possible, such as typing an address instead of sharing location.

## 4. Accounts and rating prompts

- **Account creation.** Cal AI asks people to sign up right before the paywall, framed as “Save your progress” for the plan they've just seen. That works when the account protects something real. Guideline 5.1.1(v) says that if your app “doesn't include significant account-based features, let people use it without a login.”
- **Rating prompts.** QUITTR shows Apple's rating prompt at step 12, before the user has used the app — the timing Apple's HIG advises against. Our [SKStoreReviewController guide](https://asoitis.com/blog/skstorereviewcontroller) covers when to ask instead.

## Why it matters: day 0 decides

- **Trials start on day 0.** RevenueCat's [State of Subscription Apps 2026](https://www.revenuecat.com/state-of-subscription-apps): “Nearly all trial starts happen on Day 0 across categories — users who don't try immediately rarely try at all.” For Health & Fitness the figure is 82.1%. Adapty puts it at 89.4% across its data.
- **So do cancellations.** RevenueCat: “55% of all 3-day trial cancellations happen on Day 0.” The first session decides both the trial and whether it survives.
- **Context before the price.** RevenueCat: “If a paywall appears before context is established, it feels jarring.”
- **Multi-page paywalls.** Superwall reported in May 2026 that across more than 40 million onboarding paywall opens, multi-page paywalls converted at 12.41% versus 9.07% for single-page — “a 37% gap” — though they're still only 24% of onboarding paywall opens.

The tension with Apple's “fast, fun, and optional” is real but smaller than it looks. Apple warns against teaching too much and against nonessential setup. The leaders' long flows aren't tutorials or settings — they're questions that build a result. A 29-screen quiz that ends in your plan is closer to the HIG's spirit than a five-screen feature tour that ends in a price.

## An onboarding checklist for indie subscription apps

1. Every question feeds the result, attribution or personalisation — cut the rest.
2. Show progress, and let people skip optional parts.
3. End with something personal: a plan, a score, a date, a first output.
4. Prime each permission once, right before it's needed, with a single “Continue” button.
5. Only ask for an account if it protects something the user just created.
6. Put the paywall straight after the result, with the billed price most prominent — see our [hard vs soft paywall guide](https://asoitis.com/blog/hard-paywall-vs-soft-paywall).
7. Keep the rating prompt for after real use.
8. Measure screen-by-screen drop-off and trial starts on day 0; nobody publishes per-screen benchmarks, so your own data is the only reference.

## Frequently asked questions

### How long should app onboarding be?

As long as each screen earns its place. Apple's HIG recommends a flow that's fast, fun and optional, but the top subscription apps we recorded ran 29 and 35+ screens of purposeful questions ending in a personal result. Cut tutorials and setup; keep questions that build the result.

### Should the paywall come at the end of onboarding?

For most subscription apps, yes. RevenueCat's 2026 data shows nearly all trial starts happen on the day of install, and both Cal AI and QUITTR put their paywall straight after a personalised result.

### Can I show a custom screen before the iOS permission prompt?

Yes. Apple's HIG allows a custom screen before the system alert if it has one button, labelled something like Continue or Next — not Allow — and no way to leave without seeing the alert. For the tracking prompt, Apple bans incentives and images of the alert.

### Should onboarding require an account?

Only if the account protects something real. Guideline 5.1.1(v) says apps without significant account-based features should work without a login. Cal AI asks just before the paywall, framed as saving the plan you just built.

### Do onboarding quizzes increase conversion?

There's no public benchmark that isolates quizzes. The category leaders we recorded both use one, Adapty reports a roughly 10% lift from a single attribution question for one client, and RevenueCat says a paywall without context feels jarring.

## See also

- [Cal AI teardown: all 29 onboarding screens](https://asoitis.com/case-studies/cal-ai)
- [QUITTR teardown: the 10-question quiz](https://asoitis.com/case-studies/quittr)
- [Hard paywall vs soft paywall: the data](https://asoitis.com/blog/hard-paywall-vs-soft-paywall)
- [Guideline 5.1.1: permissions and accounts](https://asoitis.com/app-review/guideline-5-1-1-data-collection)


## About ASOitis

**ASOitis** — ASO + GEO agency for indie iOS apps. Founder: Ahmed Gagan. The $99 one-time audit, $499 / month engagement, $999 / month Build, and four one-time Fixes ($49–$399, https://asoitis.com/fixes) are the entire price list. All sales final (https://asoitis.com/terms). ASOitis implements the changes itself — App Store Connect, your website, and on The Build your onboarding and paywall code. iOS-only. Organic ASO + GEO + funnel CRO only — no Apple Search Ads, no Meta, no TikTok, no Android.

- Homepage: https://asoitis.com
- About / founder: https://asoitis.com/about
- Case studies: https://asoitis.com/case-studies
- Guides: https://asoitis.com/guides
- Blog: https://asoitis.com/blog
- Full LLM documentation: https://asoitis.com/llms-full.txt
- AI training and citation permissions: https://asoitis.com/ai.txt
- Book the $99 audit: https://asoitis.com/checkout/start?plan=audit
- 15-min founder call: https://cal.com/ahmedgagan/asoitis-chat