Onboarding that keeps users: the first 60 seconds of an app

You download a new app because a friend recommended it. It opens with five swipeable cards about features you have not used yet, then asks you to create an account, then asks for notifications, then shows you an empty screen. Somewhere around the account form, you close it and forget it exists.

That first minute decides a lot. This article covers what to cut from it, what to keep, and how to get a new person from "opened the app" to "this is useful" before they lose interest.

Why the first minute matters so much

In 2015 Andrew Chen published retention data from Quettra, a mobile analytics company, based on anonymised data from more than 125 million Android devices. The average app in that data lost 77% of its daily active users within the first three days after install, 90% within 30 days, and more than 95% within 90 days.

That data is a decade old and covers Android only, so treat the exact numbers with care. The shape is the useful part: most of the loss happens right at the start. A small improvement in the first session affects everything after it.

Onboarding, by the way, is the set of screens and hints that introduce a new person to a product. It can include setup, a tutorial and the first task.

The paradox that explains most onboarding mistakes

In 1987 John Carroll and Mary Beth Rosson described what they called the paradox of the active user. People skip the manual and start using software straight away to get their task done, even though, as Nielsen Norman Group puts it, they "would save time in the long term" by learning it first.

So people will not read your tour. That is a design fact, not a user failing. Nielsen Norman Group's analysis of mobile-app onboarding (Alita Kendrick, 2020) goes further and recommends avoiding onboarding whenever possible. It notes that their research on deck-of-cards tutorials (the swipeable feature slides) found that "tutorials didn't improve task performance", and that instructional onboarding "should not be used to supplement poor design."

What to cut from the first minute

The feature tour

If your first screen is a set of slides that explain features, test the app without them. Kendrick's article advises against promoting features at first launch and suggests explaining them in context, when they become relevant. If you must keep a tour, keep a visible Skip, use as few cards as possible and put one idea on each.

The account wall

A login wall is a screen that blocks everything until you sign up. Raluca Budiu's article Login Walls Stop Users in Their Tracks (Nielsen Norman Group, 2014) explains why they hurt: they have a high interaction cost and ask for commitment before people have seen any value. Her example is oDesk, which let visitors browse jobs and contractors and only asked for an account when they wanted to hire or post. She makes an exception for highly personal services such as banking or email, where an account is the point.

Questions you do not need yet

Every question in setup costs time. Kendrick's article separates useful customisation, such as choosing a language or a fitness level, from cosmetic choices like a colour scheme, which "doesn't belong in onboarding." It also recommends keeping requests brief and explaining why you need each piece of information. It points out that people tend to stick with defaults, so good defaults often beat asking.

What to put in the first minute instead

One clear first task

Decide the single action that shows your app's value, sometimes called the "aha moment". For a notes app it is writing the first note. For an invoicing app it is seeing a finished invoice. Design the first minute to get there, and move everything else out of the way.

Help in context, not up front

Page Laubheimer's Onboarding Tutorials vs. Contextual Help (Nielsen Norman Group, 2023) contrasts "push" help, like tours that interrupt people before they need anything, with "pull" help that appears when there is a signal the person needs it, such as a tooltip on hover or a hint when they start a related task. Pull help is easier to use because it arrives when it is relevant, so people do not have to memorise anything. The same article notes one exception from the author's research: walkthroughs did help for genuinely new kinds of interaction, such as augmented reality.

Empty screens that teach

A brand-new account is often empty: no files, no projects, no favourites. Kate Kaplan's guidelines for empty states (Nielsen Norman Group, 2021) say "do not default to totally empty states." Use the space to explain what will appear there and give a direct way to start. Her example from DataDog reads "Star your favorites to list them here." Another, from Loggly, offers to add log sources or explore demo data.

Demo or sample data is a strong move for any app that only makes sense once it holds content. People can see what "good" looks like before they put in any effort.

A worked example: a small invoicing app

Here is one app's first minute, before and after.

Before (about 8 screens before any value)

1. Splash screen with logo
2. Four feature cards: "Send invoices", "Track payments", "Reports", "Go paperless"
3. Create account: name, email, password, company name, phone
4. Verify email
5. Allow notifications?
6. Choose a theme colour
7. Empty dashboard with "No invoices yet"

After (value in about a minute)

Nothing was taken away from the product. The steps were reordered so the person gets something real before being asked for anything.

A short checklist for your first 60 seconds

• Can a new person reach your main value without creating an account? If not, is there a good reason, like banking or private messages?
• How many taps from opening the app to the first useful result? Count them.
• Does every setup question change what the person sees? If not, remove it or use a default.
• Are permission requests (notifications, location, contacts) asked at the moment they are needed, with a reason?
• Does every empty screen say what goes there and offer a button to start?
• Is there a visible Skip on anything that is not essential?