Congratulations! You've just landed a new customer.
Between the three weekends you spent editing a new promotional reel, and the exhorbitant amount of cash you've burning on Google Ads - it's all finally paid off. Job's done. Like any good founder, you can go ahead, kick back and post on Instagram about how good life in a Balinese villa is.
Except, the job isn't done. It's not even close.
On the other side of the screen, nothing has happened yet. Sure, your latest 'customer' has handed over an email address and started a 14 day trial, but neither you, or them, have any idea whether they're going to stick around.
And the odds are, they won't. The average app loses 77% of its users within three days - because a users first session never gives them a reason to come back.
So - how can you improve your retention? In this article, we're going to look at what the research says about why users leave so early, how people actually learn new products, and what you can change to convince them to stick around.
Want to skip the clickbait (and world-class storytelling)? Here's the TL;DR:
- Most users decide whether to come back during their first session. The best apps don't 'keep users for longer' but they win them on day one.
- New users want to do things, not learn things. Product tours rarely help, and can make a product feel harder than it actually is.
- Every form, permission prompt, verification step and empty dashboard costs you users.
- Find the one action that turns one-offs into regulars and shorten the path to it. Make sure the first thing people do actually works.
- Track activation rate, time to activation, where people drop off and early retention, then fix it one step at a time.
A quick confession about 77%
I'll be honest, that headline number comes from a dataset that's now over a decade old. Andrew Chen and Quettra analysed data from over 125 million Android devices and found the average app lost 77% of its daily active users within three days of install, and 90% within a month.[1] Apps, phones and SaaS has all changed since then - so treat these exact stats with caution. Honestly, I just led with it because it's catchy.
What hasn't changed though is the shape of the curve, and that's the part that matters. People make their judgements early - at every timescale we can measure - and the front of the experience always carries the most weight.
At the smallest scale, people form an opinion of how a page looks at around 50 milliseconds. Lindgaard and colleagues showed participants website homepages for a twentieth of a second, then asked them to rate on visual appeal. These ratings lined up extremely closely with ratings given after a longer look.[2] That's not even enough time to read a word of copy! But, for our brains, it's plenty to decide whether something looks trustworthy and legitimate.
On a larger scale, researchers at Microsoft Research analysed 'dwell times' across more than 205,000 web pages and found that the chance of someone leaving is highest in the first few seconds, then drops sharply for those who stay.[3] They called this "negative ageing": the longer a page survives, the less likely it is to be abandoned. Jakob Nielsen's summary of that data puts the critical window at about ~10 to 20 seconds.[4] Users screen first and read later, and they almost never make it to the reading part.
Chen's data shows the same curve playing out over days. The steepest drop happens right at the start, and the users who make it past the first few days are far more likely to stay.
So when I talk about the first session in this article - it's not always a minute, or an hour. It's simply the amount of time between someone arriving and someone deciding your product is for them. Unfortunantely for you, that timeframe is often a lot shorter than we'd like it to be.
The best apps win users earlier
However the most useful detail in Chen's data isn't the average but how top apps compare. The top ten apps surveyed retained around 75% of users on day one, compared to roughly 29% for the average app. But from day two to day thirty, the top apps lost users at about the same rate as everyone else.
This is an interesting point - because it goes against traditional thinking about user retention. The best products weren't better at keeping people around over time - they were better at convincing people on day one. The gap was opened almost entirely in the first session and never closed.
This is known as 'The Technology Acceptance Model' - one of the most cited frameworks in information systems research. This model suggests that people's decision to use a system is driven mainly by two beliefs: that it's useful, and that it's easy to use.[5] Of the two, perceived usefulness was the stronger predictor. People will put up with a clunky tool if it clearly solves their problem. They won't put up with a slick tool that doesn't seem to do anything.
Bhattacherjee's expectation-confirmation model takes this a step further by looking at why people keep using something.[6] His answer, is that continued use depends on satisfaction, and satisfaction depends on whether the first experience confirmed what the user expected. If your landing page and ads say one thing, and your first session says another - that expectation gets broken, and ultimately your product fails the test it sets for itself.
People want to do things, not learn things
If the first session matters so much, the natural instinct is to use it to teach. It makes sense to want to make sure your users understand the product before they start using it - but nearly every piece of evidence says it doesn't work.
In 1987, John Carroll and Mary Beth Rosson studied people learning to use new computer systems and noticed something.[7] Learners skipped training material and jumped straight into real tasks, even when reading first would have saved them time. Carroll and Rosson called this the paradox of the active user. People are motivated by getting something done (what the authors called a "production bias"), and they'd rather make mistakes on a real task than study a system in the abstract. Nearly forty years later, the paradox is alive and well in every "Skip tour" button ever shipped.
More recent research backs this up. In a 2020 study, Nielsen Norman Group split 70 participants into two groups across four mobile apps. One group were forced to go through the apps' inbuilt onboarding and tutorials, the other skipped them.[8] Task success was about the same (91% for tutorial readers, 94% for skippers), and there was no significant difference in task time. However, people percieved the product differently. Of those that read the tutorial, nearly all of them rated the tasks as harder. Evidently, a tutorial tells users that "this system is complicated enough to need explaining".
Games offer an even clearer case study, because game designers have been obsessing over the first few minutes for decades. Andersen and colleagues ran large-scale online experiments across three games of varying complexity, testing tutorials against no tutorials, and up-front instruction against instruction delivered at the moment it was needed.[9] Tutorials were only worth it in the most complex game. In the simpler games, they made no meaningful difference to how long people played. And, when a tutorial WAS needed, users much prefered it when it was contextual and relevant - rather than a big info-dump all at once.
None of this means you shouldn't ever explain anything. It means the explanation needs to arrive at the moment it's useful, attached to the thing it's explaining. Nielsen Norman Group describes this as the difference between "push" and "pull" revelations.[10] A five-step tour on first launch is a push: the product decides it's time for you to learn. A hint that appears the first time you hover over a feature is a pull: it shows up because you've shown interest. Pulls get read. Pushes get dismissed.
Where the first session goes wrong
When I audit an onboarding flow, I'm looking for anything that sits between the user and their first USEFUL result. Any screen in that stretch is a cost, and most aren't worth paying. These are some of the things I see most often :
Asking for everything up front
Signup forms have a bad habit of growing. Someone in sales wants company size, someone in marketing wants a phone number, someone in product wants to know the user's role. Each field seems harmless on its own, but together they make a form that asks for commitment before the product has earned any.
Form design is one of the better-studied corners of interaction design. Seckler and colleagues took forms from real company websites, applied 20 established usability guidelines, and tested both versions with eye tracking.[11] The improved forms were faster to complete, needed fewer submission attempts and left people more satisfied. The quickest win is usually simpler than rewriting a form, though. Ask whether you need the information now. Anything you don't, move it to a settings page or ask for it later down the track.
Permission prompts with no context
Mobile apps love to open with a wall of system prompts (enabling notifications, allowing the camera, privacy, etc.). At that point the user has no idea why an app would need any of those things, and the prompt gives them no reason to say yes.
Bonné and colleagues at Google studied how 157 Android users responded to real permission requests over six weeks, surveying them right after each decision.[12] One of the main factors in whether someone granted a permission was whether they expected the app to need it. It's all about asking in context. A camera request right after someone taps "Scan receipt" makes sense. The same request on a splash screen looks like overreach and, once it's been denied, requires the user to take a trip to their settings to get it back.
Verification walls
Requiring email confirmation before someone can see anything is another common way to lose people in the gap between tabs. The user leaves your product to open their inbox, where they're greeted by 40 other things competing for their attention.
Don't get me wrong, there are plenty of good reasons to verify an email address, but it rarely needs to happen before the user sees any value. Let them in, and ask them to confirm before it actually matters, such as when they invite a teammate or export their work. Mailgun (not sponsored) does this really well. When I set up a fresh account recently, I was able to get in and test everything I needed straight away. Verification only came up when I went to add a DNS record for my own domain, which is exactly the point where it's reasonable to ask. By then I'd already seen the product work, so confirming my email wasn't an issue.
The empty dashboard
This is the one I see alllll the time, and it's the one that does the most damage. A new user lands in a product built for someone with months of data, and every chart, table and list is empty. There's nothing to react to and no obvious first move.
Hutchins, Hollan and Norman described two gaps every interface has to bridge: the gulf of execution (working out how to do what you want) and the gulf of evaluation (working out whether it worked).[13] An empty dashboard makes both gulfs as wide as they can be. The user doesn't know what to do first, and even if they do something, there's no context to tell them whether it's useful.
A tour instead of a task
See above. If your onboarding opens with a carousel explaining the product, you're asking people to study before they've had a chance to do anything. Most of them will skip it, and the ones who don't will come away thinking your product is harder than it is.
What good looks like
The goal of a first session isn't to teach people your product. It's to get them to one moment where the product proves its worth.
Find your activation moment
Every product has an action that separates people who stick from people who don't. For Slack, it was a team exchanging 2,000 messages. Stewart Butterfield said that of the teams that crossed that line, 93% were still using Slack.[14] Obviously this differs greatly from product to product - find the problem you're solving (the reason people use your platform) and put it front and center.
Shorten the path, or pre-fill it
Once you know what that action is, count the steps between signup and getting there. Then remove as many as you can.
There's a well-established finding from teaching research that novices learn faster from worked examples than from solving problems from scratch.[15] The same applies to software. A template, sample project or prefilled example gives people something to look at and play around with. This is far more easier than facing a blank canvas - and it narrows both of Norman's gulfs at once. Users can see what the product is for, and they can see what "done" looks like.
Defaults do similar work. Johnson and Goldstein famously showed that organ donor consent rates were dramatically higher in countries where donation was the default.[16] People stick with defaults, so the defaults you set in onboarding shape what most users will end up doing. Pick defaults that lead to your activation moment and let people change them later.
Give people a head start
Nunes and Drèze ran an experiment at a car wash.[17] Customers got one of two loyalty cards. One needed eight stamps and started empty. The other needed ten stamps but came with two already stamped. Both required the same eight purchases. The customers with the head start were far more likely to finish: 34% redeemed their card, compared to 19% of those who started from zero.
This is the endowed progress effect and it's why a good onboarding checklist starts with a box already ticked. "Create your account" counts because the user wants to feel like they've made progress. When they see they have, finishing feels more attainable than it did a second ago.
Keep it fast
Every loading state in onboarding is a small invitation to leave. Jakob Nielsen's long-standing guidance on response times (which builds on Robert Miller's work from the 1960s)[18] is that around 0.1 seconds feels instant, 1 second keeps the user's train of thought intact, and 10 seconds is roughly the limit before attention wanders.[19] If your first session includes a 15-second setup spinner, your user is already clicking off.
Measuring onboarding sessions
You can't improve a first session you can't see. At a minimum, I'd want to be tracking four things:
- Activation rate. What percentage of new signups reach your activation moment?
- Time to activation. For those who get there, how long does it take?
- Step drop-off. Where in the path between signup and activation do people stop?
- Early retention by cohort. Of the people who signed up this week, how many came back on day one, day seven and day thirty?
Then pair the numbers with something qualitative. Session recordings, a handful of usability tests with people who've never seen your product, or even just watching a friend sign up will show you why people stop.
Where I'd start
Once you've got the data, you don't need to entirely rebuild onboarding to improve it. If I was given a product right now, this is the order I'd work in:
- Sign up as a new user yourself, on a phone, and time how long it takes to get to something useful.
- Remove one question from the signup form.
- Replace the empty state on your main screen with a template or sample data.
- Move one permission request or verification step to the moment it's actually needed.
- Delete the product tour and add contextual hints to the two features people use most.
Make one change at a time and measure the effect on activation. It's slower than a redesign, but you'll know what actually worked.
The first session is where your product makes its case. The research is fairly clear on what wins it - get people doing something real, get them to a result quickly, and let the product prove itself before you ask for anything else. Everything between signup and that first result is a cost, and your users are counting it even if you aren't.