How Many Onboarding Screens Should an App Have?

Onboarding length gets argued as a number when it is really a question about what each screen does. A four-screen flow full of marketing carousels will lose more users than a seven-screen flow where every step collects an answer the app visibly uses.

The published data supports a range rather than a target. Produktly’s 2026 onboarding benchmarks, drawn from 464 SaaS products between January and June 2026, put median completion at 73% for flows of one to two steps, 38% at three to five, 25% at six to eight, and 8% at nine or more.

Meanwhile Superwall’s analysis of more than 40 million onboarding paywall opens has longer paywall flows converting better than short ones.

How many onboarding screens is too many?

Nine or more steps is where onboarding stops working for most apps. According to Produktly’s 2026 benchmark data, median completion falls to 8% once a flow passes nine steps, against 38% in the three to five step band, which means the last few screens are being seen by a small minority of the cohort you paid to acquire.

Onboarding length	Median completion	What usually fits
1 to 2 steps	73%	one framing screen, nothing else
3 to 5 steps	38%	framing, one or two questions, a permission
6 to 8 steps	25%	the above plus a paywall sequence
9 or more steps	8%	more than most cohorts will finish

Completion figures in that table are the medians from Produktly’s 2026 study, and the right hand column is what those step counts have to accommodate in a subscription app.

The drop is not linear, and the shape matters more than the headline. Going from two steps to five costs roughly half the completion in that dataset. Going from five to eight costs a third of what is left, and past nine the flow is effectively optional.

Read those figures as the cost of length rather than as a verdict on it. Produktly measured product tours inside SaaS products, which is not the same format as a mobile onboarding flow, and a survey that personalises the app is a different proposition from a carousel explaining features. What transfers is the rate of attrition per screen, which is steep in every published dataset.

Three to six screens is the range where most subscription apps can still do the jobs onboarding has to do: establish what the app is for, collect the one or two answers that change the experience, prime a permission, and arrive at a paywall with the user still present.

Past six, ask what the extra screen is buying. If the honest answer is that a stakeholder wanted a feature mentioned, delete it. The feature will be better explained by the user encountering it in context.

Which screens actually earn their place?

A screen earns its place when it either collects something that changes what the user sees next, or delivers something the user can feel. Every screen that does neither is a tax on completion paid by the whole cohort.

Three categories clear that bar in most apps. A framing screen that states what the app does in the user’s own terms, a personalisation question whose answer visibly changes the next screen, and a first-value moment where the app does something for the user rather than describing itself.

Feature carousels usually fail the test, because nobody remembers a feature list. The Nielsen Norman Group position, in Onboarding: Skip It When Possible, is that in-context guidance at the moment of use beats a tour nobody asked for. Sign-up walls placed before any value has been delivered fail for a different reason: they ask for effort before giving one.

Permission prompts sit in between, and they are covered separately below. The test is the same one: a permission screen earns its place only if the screen before it made the permission make sense.

One practical way to apply this is to write the list of screens down and mark each one as collect, deliver or explain. Explain screens are the ones to cut first, because explanation is what the product itself is supposed to do.

How many screens should come before the paywall?

Enough screens to produce one real value moment, and no more. Adapty’s State of In-App Subscriptions 2026 found paywalls triggered after a measurable value moment starting 2.1 times as many trials as paywalls shown immediately on open, which is the largest single effect in the published onboarding data.

The window is tighter than it sounds. According to the same Adapty study, 89.4% of all trial starts happen on install day, so the value moment has to fit inside the first session. A plan that shows the paywall on day three is a plan to show the paywall to almost nobody.

The paywall itself is allowed to be long, which surprises most teams. Superwall’s study of more than 40 million onboarding paywall opens between February and May 2026 measured multi-page paywall flows converting at 12.41% against 9.07% for single-page, a 37% relative difference in favour of the longer format. Multi-page flows were only 24% of onboarding paywall opens in that dataset, so the better-converting format is still the minority choice.

Read those two findings together and the rule is not “shorter”. Screens that build context for the purchase earn their keep, and screens that delay the first useful thing the app does are the ones costing you the cohort. Benchmarks for what the paywall should then convert at are in our piece on app paywall conversion rates.

Do survey questions justify the extra screens?

Survey questions justify their screens when the answers visibly change the app, and not otherwise. Adapty’s 2026 onboarding write-up reports personalised onboarding flows lifting trial starts by 8.5%, paying conversions by 17% and ARPU by 22% against a non-personalised control, which is a real return on two or three extra taps.

The mechanism is not the data you collect. A question signals that the app is about to be built around the answer, and the next screen either honours that or breaks the implied promise. An onboarding survey whose answers never resurface is drop-off with a research benefit.

Adapty also reports a survey-plus-lesson structure, where the questions are followed by a short piece of teaching rather than by a feature tour, producing a 25% increase in trial starts and a 78% lift in ARPU in its testing. Both numbers come from the vendor’s own platform data rather than an independent study, so treat the direction as reliable and the magnitude as optimistic.

The practical limit is three to four questions. Past that, completion falls faster than the personalisation gain, and the answers start to be given carelessly, which makes the segmentation worse rather than better.

Where do permission prompts belong in the flow?

Permission prompts belong behind an explainer screen, and after the user has seen something the permission would improve. The operating system gives you one native prompt, and a denial is expensive to reverse, so the screen before it is doing more work than the prompt itself.

Tracking permission is the one with the most money attached. App Tracking Transparency opt-in sits at roughly 27% globally in 2026 according to Business of Apps, with the United States higher than the European average, which means the majority of your users will not be individually attributable whatever you do. Asking well moves that number, and asking cold on screen one wastes it.

Notification permission is the cheaper prompt to get right, and you will lose about half of iOS on it whatever you do. Airship’s Mobile App Push Notification Benchmarks 2026 puts median opt-in at 51% on iOS against 81% on Android, so the explainer screen in front of the system prompt is the only part of that number you control.

Sequence both prompts after the first value moment rather than before it. A user who has seen the app do something useful has a reason to say yes, and a user on screen two has only a reason to say no.

Does the answer change when your installs come from creator ads?

Creator-led traffic changes the answer, because the ad has already done part of the onboarding. A user who watched a 20 second video showing the product in use arrives knowing what the app is for, which makes a framing screen repeating that claim a wasted step.

The mismatch is the real risk. When the ad promises one outcome and onboarding starts asking about a different one, the drop happens on the first screen and looks like an onboarding problem in your dashboard. The creative team reads its own metrics as healthy, the product team reads a first-screen drop, and neither one is looking at the seam.

Segment onboarding completion by acquisition source before you change any screens. If completion is fine on organic installs and poor on one campaign, the flow is not the problem and the targeting or the hook is.

The stronger version of this is to design the first screen against the ad that brought the user in, so the promise carries through rather than restarting. Running the creative and the onboarding as one experiment is the part of an app launch most teams have split across two vendors.

How do you tell whether your onboarding is too long?

Read completion screen by screen instead of end to end, because the total tells you nothing about which step is costing you. Chart the share of installs reaching each screen and find the single biggest drop, then fix that screen before touching the count.

What share of installs ever reaches the paywall tells you whether onboarding is the bottleneck. A healthy paywall conversion rate over a tiny number of paywall views means the screen is fine and the flow in front of it is not. Comparing completion by acquisition source answers the other half of the question, which is whether the problem is the flow or the traffic.

Test by removal rather than by addition. Cutting a screen is a cleaner experiment than redesigning one, and it usually produces a bigger effect, because the question is whether the screen was worth its attrition in the first place.

Hold each test for a full install cohort rather than a rolling window. Given that 89.4% of trial starts happen on install day per Adapty’s 2026 study, a cohort read by install date resolves in about 72 hours, which is faster than most teams expect and much faster than the 30 day window the advice usually recommends.

The short version

Three to six onboarding screens is the working range for most subscription apps. Produktly’s 2026 benchmark study found median completion at 38% for three to five step flows, 25% at six to eight and 8% at nine or more, so every extra screen is paid for out of the cohort you bought.

Judge screens by function rather than by count. Keep the ones that collect an answer the app visibly uses or deliver something the user can feel, cut feature carousels, and put permission prompts behind an explainer and after the first value moment.

Then place the paywall after one real value moment inside the first session, since per Adapty that timing starts 2.1 times as many trials, with 89.4% of trial starts happening on install day. The paywall itself can run long: Superwall measured multi-page onboarding paywalls at 12.41% against 9.07% for single-page.

Read completion by screen and by acquisition source, because a flow that works for organic installs and fails for one campaign is not an onboarding problem.