16 App Onboarding Examples, and the Psychology Behind Each One

The best app onboarding flows give new users something worth keeping before they ask for much in return. Duolingo teaches you a word before it asks for an email address. Shopify builds your store before you sign up, and Wise shows the exact price of a transfer before any form. When a flow has to ask for more, the good ones explain why, show how far you've come, and keep each screen to one easy decision.
We went through 16 onboarding flows screen by screen. For each one we show the key screens and name the behavioral principle that makes the design work, or backfire. Where the company has published numbers, we add them.
The psychology behind good onboarding
A useful way to think about onboarding comes from Stanford researcher BJ Fogg. His behavior model is a design framework rather than a tested law. It says a behavior happens when three things meet at the same moment: enough motivation, enough ability (the action is easy), and a prompt.
Every onboarding screen is a request for a small behavior, like typing a name or allowing notifications. So every screen either raises motivation, makes the action easier, or picks a better moment to ask.
The flows in this article do those three jobs in different ways. Duolingo, ChatGPT and Shopify raise motivation by handing you something valuable first. Instagram and Uber make each step easier by asking for one thing at a time. SoundCloud and Tinder time their asks so they arrive right after a reason to say yes.
The table below names the effects you'll meet most often, with the research behind each one.
| Effect | What it means for onboarding | Where you'll see it | Research |
|---|---|---|---|
| Reciprocity | A favor creates a pull to give something back, so value before the ask makes the ask easier | ChatGPT, Duolingo | Regan, 1971 |
| Endowment effect | People value what they already have more than what they could get, so "save your progress" beats "create an account" | Duolingo, ChatGPT | Kahneman, Knetsch and Thaler, 1990 |
| IKEA effect | People value things more when they helped make them, as long as they finish | Shopify, Tinder | Norton, Mochon and Ariely, 2012 |
| Labor illusion | Seeing the work behind a result makes it feel more valuable | Shopify | Buell and Norton, 2011 |
| Worked-example effect | Beginners learn faster from a solved example than from a blank page | Linear | Atkinson et al., 2000 |
| Ambiguity aversion | People prefer a known cost to an unknown one | Wise | Ellsberg, 1961 |
| Reason-giving | A request with a reason gets more yeses, and for big requests only a real reason helps | MyFitnessPal, Noom, Tinder | Langer, Blank and Chanowitz, 1978 |
| Defaults | People tend to keep whatever option is already selected | MyFitnessPal | Jachimowicz et al., 2019 |
| Commitment and consistency | A small yes makes a bigger yes more likely | Noom, Tinder | Freedman and Fraser, 1966 |
| Implementation intentions | Deciding when you'll act makes you more likely to act | SoundCloud and Calm | Gollwitzer and Sheeran, 2006 |
| Goal gradient and endowed progress | People speed up as a goal gets closer, and a head start makes it feel closer | LinkedIn, Tinder | Kivetz, Urminsky and Zheng, 2006 |
| Cognitive load | People can hold only a few things in mind at once, so one decision per screen is easier | Instagram, Uber | Cowan, 2001 |
One caution before the examples. Most of these effects were measured in labs, loyalty programs and surveys, not in your app.
Some are smaller and more conditional than their popular summaries suggest. Choice overload, the idea that too many options stop people choosing, averaged out to roughly zero across 50 experiments in one meta-analysis. It shows up mainly when options are hard to compare and people don't know what they want (Chernev, Böckenholt and Goodman, 2015).
Treat each principle as a hypothesis about your users, then check it against what they actually do. A section near the end of this article shows how.
Duolingo: a whole lesson before the signup wall
Principle: the endowment effect. Account: after the first lesson.
Duolingo shows up on almost every list of great onboarding, and the web version shows why. We pressed "Get started" on duolingo.com, picked Spanish and went through about a dozen short screens. There was a welcome from Duo, five quick questions, an overview of what we could achieve and a couple of encouraging messages.
The questions asked how we'd heard about Duolingo, why we're learning, how much we already know, how many minutes a day we want to practice and where to start. Each one is a single choice from a short list.
Then came our first real question, and when we got it right, Duolingo told us "Good job!"
No email field appeared before or during the lesson. Duolingo's own beginner's guide frames the account step as "Create a profile to save your progress". That phrase turns signup into a way to keep something you already have, which is where the endowment effect comes in. In a classic experiment, students handed a Cornell coffee mug asked more than twice as much to sell it as other students would pay for one (Kahneman et al., 1990).
Economists still argue about how strong the effect is, but the product logic is simple. Once you've earned some progress, an account stops being a hurdle and becomes a way to keep what you have.
Duolingo has numbers on the timing. Gina Gotthilf, Duolingo's former VP of Growth, told First Round Review in 2017 that moving the signup screen back a few steps increased daily active users by about 20%. The test can't say which effect did the work: less friction, reciprocity and the endowment effect all fit.
The daily goal screen does quieter work too. Choosing "10 min / day" is a small commitment you made yourself, and Duolingo can hold you to it later with reminders and streaks.
Duolingo also shows the limits of the idea. In its Q4 2025 shareholder letter, the company said it had been adding "small amounts of friction" to push people towards a subscription. It added that "this extra friction is part of the reason our DAU growth has slowed." Value before the ask works in both directions: take value away and people notice.
ChatGPT: the answer comes before the account
Principle: reciprocity. Account: when you want to keep your chats.
The best onboarding for a tool is using the tool. Since April 2024, OpenAI has let people use ChatGPT without signing up. On the phone, the very first screen offers "Try it first" next to "Log in" and "Sign up for free".
For people who do create an account, the flow asks for a password, an emailed code, a name and a birthday. Then comes one optional question about how they'll use the product.
The empty chat that follows isn't really empty: it offers suggestions like "Create image", "Summarize text" and "Analyze images". That matters more than it looks. A blank box asks the user to invent a first task, and a suggestion turns that invention into a tap.
The logic of letting people in first is reciprocity. In a 1971 experiment, a stranger handed some students a free Coke (Regan, 1971). Later, those students bought nearly twice as many of his raffle tickets as students who got nothing.
A product that has already answered your question has done you a favor, and an account feels like a fair trade for keeping it. OpenAI's help page on logged-out use spells out the trade. Without an account you work in one temporary conversation, and once you create one, "the thread to your latest conversation will appear on the side".
Shopify: your store exists before your account does
Principle: the IKEA effect and the labor illusion. Account: after you pick a design, for the free trial.
Shopify has the hardest version of this problem. Its product is an online store, and nobody can see what their store would look like before they've built one. Its AI store builder answers with a prompt box: describe your business in a few words and press "Generate".
We typed "Handmade ceramic mugs and plant pots". Instead of a spinner, Shopify narrated its work: "Refining the prompt, 2 minutes left", then "Defining palettes and styles, 1 minute left". Then it showed three finished store designs with our products in them, under the headline "Pick your favorite design included with your free trial". No email address had been asked for yet.
Two effects stack up here. The first is the IKEA effect. In experiments, people who assembled an IKEA storage box would pay more for it than people handed the same box ready-made (Norton et al., 2012). The catch is that they had to finish building it.
One sentence of input is a small contribution, but the result is unmistakably yours.
The second is the labor illusion, tested with a simulated travel site that showed the work behind its search (Buell and Norton, 2011). With the work visible, 62% of people chose it over an instant site with identical results, against 42% when the work was hidden. Shopify's narrated waiting screens turn a delay into proof that something is being made for you. The authors add a caution that applies here too: "the fact that firms can induce the labor illusion does not mean that they therefore should induce it."
The headline makes the next step clear: the designs come "included with your free trial". By then the trial isn't an abstract offer. It's the way to keep a store you've already seen, which is a much easier yes than "create an account to start building".
Linear: a demo workspace instead of a blank page
Principle: the worked-example effect. Account: when you want your own workspace.
For software that teams adopt together, the question before signup is "what will this look like once we're using it?" Linear answers it with a public demo workspace that opens without an account. Its own start guide suggests it: "To learn Linear's layout before signing up, explore our demo workspace."
The demo is a fully furnished product: a welcome document, issues with comments and history, projects and cycles. You learn the layout by clicking around a workspace that already works, rather than reading about one. When you're ready, "Sign up" leads to "Create your workspace".
This is the worked-example effect from learning research. Beginners learn faster from studying a solved example than from being dropped into a blank problem. The finding comes from years of studies in algebra and geometry classes (Atkinson et al., 2000).
A finished workspace is a worked example of how Linear expects a team to use it. Later research found the benefit fades as people gain experience (Kalyuga et al., 2003), which is a good reason to keep the demo optional.
Notion does a version of this after signup. Its help center says new users see "a collection of templates in your sidebar… selected for you based on what you tell us during onboarding." Either way, the first thing a new user sees is a page that already works, not an empty one.
Airbnb: browse everything, sign in to book
Principle: ask when motivation peaks. Account: when you book.
Airbnb lets you do almost everything before it asks who you are. You can search, open listings, check dates and see the full price without an account. On the web, "Log in or sign up" appeared only after we pressed "Reserve". By then we'd checked a summary of our dates, guests and total price, with a note that the price was below the 60-day average.
Fogg's model explains the timing. Motivation isn't fixed; it rises and falls. A visitor idly browsing has little reason to fill in a form, while someone who has just picked a flat for their trip has every reason.
Airbnb puts its biggest ask at the peak. By then you've chosen a place, so the account is the last step to something you want, not a wall between you and browsing.
Identity checks wait too. Airbnb's help center says guest ID details are collected "when you book". Uber learned the same lesson on the web in 2019. When it let new riders book a first trip straight from the homepage, first-trip conversions rose 10% globally and about 60% in India and Brazil, according to Uber's engineering blog.
Wise: the price before the paperwork
Principle: ambiguity aversion. Account: when you press "Send money".
Money apps can't let you skip identity checks, but they can choose what comes first. Wise leads with the thing people came to find out: what a transfer costs. Its send money page works logged out.
On the day we checked, sending 1,000 US dollars would get the recipient 884.93 euros. The fees were 5.47 dollars, and the money would arrive "Today - in seconds".
Only after "Send money" did Wise ask whether this was a personal or business account. Its registration page then starts with your email address.
This order answers a basic preference: people would rather face a known cost than an unknown one. Daniel Ellsberg described it with a bet (Ellsberg, 1961).
Picture two urns of red and black balls, one with a known 50-50 mix and one with an unknown mix. Asked to bet on drawing red, people tend to prefer the known urn, even though nothing says the second is worse. Later lab studies found that preference in many people, though not in everyone (Trautmann and van de Kuilen, 2015).
A money transfer is full of unknowns: the rate, the fees, the timing. Wise removes them before asking you to commit to anything.
The paperwork that follows is honest about itself. Wise's verification guide says the checks depend on how much you send and where, and that review usually takes one to two working days. It also says that "in some cases, we can automatically verify your identity with your first bank transfer."
If your onboarding has to include a heavy step, telling people the price and the reason first is the cheapest way to make it feel worth it. We've written more about measuring fintech onboarding if you're working on one of these flows.
MyFitnessPal: every question says what it's for
Principle: reason-giving. Account: after nine screens.
MyFitnessPal asks a lot before it asks for an email address. On the web we went through nine screens under the made-up name "Alex". Seven were questions: name, goals, past barriers, activity level, sex and birth date, height and weight, and a weekly target. It gets away with it because every answer goes into something you're about to receive, your calorie goal.
Three details make the length bearable. The second screen uses your name ("Thanks Alex!"). When we said a lack of time had stopped us before, the next screen answered that exact point: "We get it. A busy lifestyle can easily get in the way of reaching your goals."
And the sensitive questions come with reasons: "We use this information to calculate an accurate calorie goal for you."
That last habit has a famous experiment behind it, run at a copy machine in 1978 (Langer et al., 1978). Asking to cut in line "because I'm in a rush" got 94% of people to agree for a short job, against 60% with no reason. For a bigger job, only the real reason helped, and a later study couldn't reproduce the effect of empty reasons (Folkes, 1985).
That's the useful part for onboarding. When you ask for something personal, the reason has to be real, and MyFitnessPal's is: your weight really does go into the calorie goal.
The weekly goal screen adds a gentle default, "Lose 1 pound per week (Recommended)". Defaults are among the sturdier effects in this article. In a meta-analysis of 58 studies, most found that defaults shifted what people chose, though by very different amounts (Jachimowicz et al., 2019).
The risk is in what happens next. You've put seven answers into a plan you haven't seen yet when the "Almost there! Create your account" screen appears. If those answers then lead to the same home screen everyone else gets, people notice, and every later question feels like a waste of their time.
Noom: a long quiz that answers back
Principle: commitment and consistency. Account: after the quiz.
If MyFitnessPal's nine screens sound like a lot, meet Noom. Its web funnel is a questionnaire that runs to "up to 113 screens" and 10 to 15 minutes, according to a walkthrough on RevenueCat's blog. The price comes after roughly 100 of them. We went through the start of it.
What stands out is how much effort goes into explaining each question. Before weight: "We don't mean to pry, we just need to know so we can build a plan that's right for you." Before diabetes: "We ask this question to tailor your program with the right support."
The answers also pay off quickly. When you enter a goal weight, Noom answers with a recommended weight range, and in our walk an interstitial broke the rhythm with a chart and a study result.
A long quiz like this builds commitment one small answer at a time. A classic study shows how strong this can be (Freedman and Fraser, 1966). People who first answered a few survey questions by phone were more than twice as likely to agree later to a long home visit, 52.8% against 22.2%. Later reviews found this foot-in-the-door effect is real but far less reliable than those first numbers (Burger, 1999).
By the time Noom shows a price, you've told it a lot about yourself, and walking away feels inconsistent with all those answers.
That's also the danger. The same pull is the sunk cost effect, the urge to keep going because of what you've already put in. Theatergoers who paid full price for season tickets went to more plays early in the season than those who got a discount (Arkes and Blumer, 1985). A quiz that's long because it needs the information is one thing; a quiz that's long to make leaving feel expensive will cost you trust and refunds.
There's a twist on personal quizzes from Headspace. The behavioral design firm Irrational Labs tested five onboarding versions for the app and published the results in a case study. In the original flow, 31.25% of new users started a course. In the winning version, the app asked a few questions first, and 62.97% started a course.
Yet everyone still got the same Basics course: the quiz changed how the recommendation felt, not what it was. None of the versions changed how many days people meditated, so measure week two before you celebrate week one.
SoundCloud: three artists, then the notification ask
Principle: prompt timing. Account: first, then the picks.
Music apps have a cold-start problem: on day one they know nothing about your taste. SoundCloud solves it the direct way. After a quick account step with three fields, it shows a grid of artists and asks you to pick three.
The artist grid is the first moment the app feels like yours. Only after it does SoundCloud ask to send notifications about "new tracks, exclusive releases, and more from your favorite artists", with "Maybe later" right below. Asked at that point, a notification has an obvious purpose: you'll hear when the artists you just picked release something. Asked on screen one, it's just a system dialog.
This is Fogg's prompt at the right moment, and it has real numbers behind it. In a 2016 case study, Calm found that fewer than 1% of users discovered its daily reminder in settings. When it asked right after the first meditation, 40% set one, and users who set a reminder showed almost three times the retention, according to Amplitude.
Choosing a time to meditate is an implementation intention, a plan that says when you'll act. A meta-analysis of 94 tests found these plans have a medium-to-large effect on reaching goals (Gollwitzer and Sheeran, 2006). Outside the lab the effect is smaller. Asking people to write down a date and time for their flu shot raised vaccination by about 4 percentage points (Milkman et al., 2011).
Spotify has published research on the picker itself. Its 2025 research post says onboarding picks like artists and genres "enable instant personalization from day one". Its new model of users, which uses those picks, made predictions 5% more accurate in a new user's first four hours.
LinkedIn: a progress bar and "Skip for now"
Principle: the goal-gradient effect. Account: first.
LinkedIn needs much more from a new member than Instagram does: where you live, what you do, whether you're looking for a job. It splits all of that into small steps, shows a progress bar on the signup form, and gives most of the optional steps a "Skip for now".
The idea behind progress bars is the goal-gradient effect: people speed up as a goal gets closer. Members of a café reward program bought coffee more often as they neared their free one (Kivetz et al., 2006).
A head start helps too. In one study, car-wash customers got either a ten-stamp card with two stamps already filled in or an empty eight-stamp card (Nunes and Drèze, 2006). Both needed eight washes, yet 34% finished the first card against 19% for the second.
LinkedIn's bar jumps quickly, from 30% to 60% to 90% over three screens, which makes the finish line feel close.
Then six more screens follow: a security puzzle, location and job, a photo, an email code, a job-search question and notifications. A bar that says "almost done" when you aren't is a small broken promise.
Survey researchers have tested progress bars a lot. In a review of 32 web-survey experiments, steady, accurate bars didn't reduce drop-off (Villar et al., 2013). Bars that crawled early on made people more likely to quit.
Bars that moved fast early and slowed later did a little better, but not reliably. LinkedIn's bar behaves that way, and it costs some trust when people notice. The better version is to put the quick, easy steps first, so the bar moves fast for real.
The ending is well judged. The first home screen isn't empty: a "Welcome to LinkedIn" card with a three-step checklist suggests adding your experience so recruiters can find you.
Checklists aren't magic, though. The email app Superhuman learned this when it moved from one-on-one onboarding calls to self-serve onboarding. Its first attempt, a setup checklist with tooltips, "yielded zero impact", according to Gaurav Vohra's account in First Round Review.
Full-screen setup steps that couldn't be skipped, each with sensible defaults, lifted setup completion from 30% to over 98%. We took a closer look at LinkedIn's flow in our LinkedIn onboarding teardown.
Tinder: a 19-screen profile worth building
Principle: commitment and consistency. Account: first.
Tinder breaks every rule about short onboarding. Our capture runs to 19 screens. After a phone number, a code, an email and an optional Apple sign-in come house rules, first name, birthday, gender, distance and what you're looking for. Then come school, lifestyle habits, personality questions, interests and photos, and finally location, a screen for blocking people from your contacts, and a promo for a new feature.
It can afford the length because here the profile is the product. Every answer makes your profile better, and a better profile means better matches.
After the sign-in steps comes a pledge. "Welcome to Tinder. Please follow these House Rules": be yourself, stay safe, play it cool, be proactive.
You tap "I agree". It's the foot-in-the-door effect from the Noom example: a small yes makes the bigger yeses that follow easier. Agreeing to "be yourself" also sets the norm for the profile you're about to build.
The optional steps are handled with care. Lifestyle and personality screens show a counter on the button, "Next 0/4", and a "Skip" in the corner. The photo screen asks for two photos to start, with six slots to fill. Counters like these make progress visible inside a single screen, and the slots invite you to complete the set.
Because you write the profile yourself, the IKEA effect applies too: people value what they've made.
Tinder is also careful about what other people will see. The birthday screen says "Your profile shows your age, not your birth date", and showing your gender on your profile is an unticked option. The name screen warns that it "Can't change it later."
The relationship goal screen offers six answers, including "Still figuring it out", and adds "All good if it changes." Tinder says more than half of members in its test markets added a relationship goal to their profile (Tinder, 2022).
The location request is a good model for permission screens. Before iOS shows its system dialog, Tinder explains in one sentence what you lose without it: "You won't be able to match with people otherwise." Under the "Allow" button, "How is my location used?" opens the details. The one thing we'd change is the missing "Not now": people who hesitate have nowhere to go but the system dialog.
One thing has changed since our capture. Tinder now requires new users in several countries, and in California, to verify themselves with a short video selfie it calls Face Check, according to Tinder's newsroom. It's a heavier step than anything in our capture, and a good test of whether the flow's habit of explaining itself carries over.
Instagram: one question per screen
Principle: cognitive load. Account: first.
Instagram doesn't explain itself, and it doesn't need to. Almost everyone who installs it already knows what it's for. So its signup does one thing: it gets you an account with the least possible thinking per screen.
Eight screens sounds long, but each holds a single field and a big blue button. That keeps the mental load low. Working memory holds only about four things at once (Cowan, 2001), and a long form asks people to track several fields, rules and error messages together. A screen with one field has one decision on it.
The username screen comes pre-filled with a suggestion and a green tick, and the birthday screen links to "Why do I need to provide my birthday?" The two steps Instagram would like but doesn't need, a profile photo and a phone number, both have a "Skip" under the main button.
Don't copy the missing product tour; that only works for a brand everyone already knows. Copy the way the form is cut into questions people can answer without thinking.
Meta's newer app Threads went further at launch by letting people import their Instagram follows. In November 2024, Adam Mosseri said Threads was removing the follow import from onboarding because people "prefer their experience if they build a different graph on Threads." Even a shortcut that drove a launch can be the wrong shortcut a year later.
Uber: one identifier instead of "sign in or sign up?"
Principle: fewer decisions. Account: first.
Every "Sign in / Create account" fork is a small quiz: do I already have an account here? Uber removed the question. Its first screen asks for a phone number, with Apple, Google and email as alternatives, and works out from there whether you're new.
Uber's engineers have written about this twice. A 2017 post describes removing the Sign In / Register choice by asking first for "a unique piece of personally identifiable information: a mobile number". The same redesign moved to one request per screen. A 2022 post explains the system behind it: you give one identifier, and Uber routes you to login or signup automatically.
The passkey screen is the newest part. Uber offers it right after you've proven who you are, with "Not now" in plain sight. In 2025 the company told Google's Android Developers Blog that "90+% of passkey enrollments come from promoting passkey creation at key moments" like signup.
In our capture, "Where to?" was the seventh screen. The phone number, an email address, a four-digit code, your name, the passkey offer and the terms came first, one per screen. None of them asked for a payment method.
Amazon: eight screens, a puzzle and two codes
Principle: friction adds up. Account: first, in eight screens.
Amazon is the counterexample. Like most big shops, it lets you browse without an account, so its signup doesn't have to sell anything. Still, it's worth counting what a new customer goes through once they tap "Create account" in the app.
We counted eight screens. After name, email and password comes an image puzzle ("Solve this puzzle to protect your account") and a one-time code by email. Then Amazon asks you to add a mobile number "to enhance your account security", and sends a second code by text message. Each step has a reason, and a shop people already trust can get away with more of them than most products can.
Most products can't. Every extra screen is another moment to leave, and the losses multiply. If each of eight steps keeps 95% of the people who reach it, only about two-thirds make it through all eight.
Monzo shows how much a team can win back by finding the step that costs the most. In October 2023, only 9% of people who downloaded the Monzo US app finished signup; the flow had 39 screens and an average of 210 taps.
The team found that one check caused 40% of the drop-off and let low-risk customers skip it. Device autofill, address autocomplete and folding the terms into the phone number screen shrank the flow to 17 screens. Completion rose to 40% after the risk-based change, according to Monzo's engineering blog. Monzo has since closed its US accounts, but the method holds.
If you need a puzzle or a second code, run it only for the signups that look risky, as Monzo did, instead of for everyone.
Temu: a marketing ask on top of a security step
Principle: trust is a budget. Account: first.
Temu starts well. Its sign-in screen shows two reassurances before any button, "No import charges" and "Free returns", plus "All data is encrypted". Then the asks start to pile up.
While we were verifying our email, a pop-up appeared on top of the security step. "Your Mystery Offers Are Waiting", it said, asking for a phone number to receive marketing text messages. The first account screen after that asked us to "Turn on Notifications for updates on your latest order status". Right below sat a credit balance and a "Buy to get $50 credits" offer.
Every one of these asks spends some of the trust a new user brings. A marketing opt-in dressed up as part of a security step spends a lot of it, because it borrows the urgency of a security step for something that isn't one.
In 2019, researchers crawled 11,000 shopping websites and found 1,818 of these "dark patterns" on about one site in nine (Mathur et al., 2019). They included tricks that push people into creating accounts or receiving marketing, and Temu's pop-up resembles them. They may lift opt-ins in the short term. They also teach new customers to read your screens with suspicion.
YouTube: seven screens of Google account setup
Principle: context matters. Account: when you sign in.
Signing in to the YouTube iPhone app without a Google account sends you somewhere else. Our capture shows an in-app browser on accounts.google.com with Google's general account form. It asks whether the account is personal or for business, then for your name, birthday and gender, a new Gmail address, a password and a phone number for a code. A long privacy page comes last, before "I agree".
Some of it is well made. The screens are short, the last name is marked "(optional)", and the birthday screen points to an explanation of why Google asks.
But that explanation lives outside the flow: you're told to visit g.co/age-gender in your browser. The phone step is titled "Confirm you're not a robot" with no visible way past it, although Google's help page lists adding a phone number as optional. And nothing in the seven screens mentions videos, channels or anything you came to YouTube for.
Google can live with this because most people already have an account. The same help page points out that anyone who has signed in to Gmail, Maps or YouTube before has one. YouTube's help pages also describe recommendations built from what you watch and search, with nothing about picking interests at signup. Its help page on recommendations says that if your watch history is off and you have little history, the home page shows only the search bar and the menu.
The lesson is about context. The same request feels different depending on what surrounds it. Airbnb asks for an account inside a booking you want to finish; Google's form could be for any app.
If your signup has to be shared across products, add one line of context at the start, such as "Create an account to subscribe to channels and save videos". Explain sensitive questions on the screen itself, and bring people back to where they were the moment they finish.
What the best onboarding flows have in common
Put the 16 flows side by side and five habits stand out. None of them needs a big budget. They need a clear idea of what the user came for, and the discipline to ask for things only when the user can see why.
They give before they take
Wherever the product allows it, the best flows put value first. Duolingo gives you a lesson, ChatGPT an answer, Shopify a store, Linear a workspace, Airbnb a place to stay and Wise a price, all before any account. Duolingo's roughly 20% lift in daily active users from moving signup back is the best-known number. The logic is simple: people sign up more willingly to keep something than to get permission to look.
Every question changes something
MyFitnessPal turns your answers into a calorie goal, and Noom into a weight range. SoundCloud uses your three artists to make its notification ask make sense.
A good test for any onboarding question is whether its answer changes the very next screen. If it doesn't, cut the question or show what it's for. Headspace's study is the honest footnote: a quiz can work by making people feel understood, but it won't keep them if the product doesn't deliver.
Long flows show the finish line
LinkedIn and Tinder both ask a lot, and both show progress: a bar along the top, and on Tinder a "0/4" counter on the button. Progress bars are less reliable than their reputation, though: what helps is making the early steps genuinely quick. Optional steps get a visible "Skip", so the flow never holds people hostage.
Each screen asks for one thing
Instagram, Uber and Duolingo keep each screen to one field or one tap. That makes a flow longer in screens and shorter in effort. Uber goes further by removing the "Do I have an account?" question altogether, and Instagram pre-fills a username that's already available.
Permission asks come after a reason
SoundCloud asks for notifications after you've picked artists, and Tinder explains why it needs your location. Calm got 40% of the people who saw its prompt to set a reminder by asking right after the first meditation. Airship's 2026 push benchmarks found that apps running onboarding campaigns see opt-in rates as much as 40% above their category average.
How to design your own onboarding flow
You can apply all of this without copying anyone's screens. Start with your own users and work through these steps in order.
Name the moment of value. Pick the first action that shows a new user got what they came for, like a first lesson or a first transfer. Everything in onboarding should lead there.
Move value in front of the account wherever you can. List every screen before that moment and ask which ones truly must come first. Accounts, permissions and payment details often can wait until the user has a reason.
Make every question pay off. For each question, write down what changes on the next screen because of the answer. If nothing changes, cut it or explain what it's for.
Explain every hard step in one sentence. Identity checks, phone numbers and location requests all go down easier with a plain reason next to them.
Keep each screen to one decision. Split long forms into single questions, preselect sensible defaults, and give optional steps a visible "Skip".
Time your prompts. Ask for notifications, reminders and reviews right after the moment that gives them a purpose.
Show honest progress. If the flow is long, put the quick steps first, show where people are, and let them save and come back.
Then test it. Every principle in this article is a hypothesis until your own users confirm it.
How to find where your onboarding loses people
Studying other apps' flows gives you ideas. It can't tell you which screen in your own flow is losing people, or why. A funnel shows the step where people leave, never what went wrong on it. For that you need to see what new users actually do.
This is where UXCam comes in. It's a product analytics platform for mobile apps and websites, and it captures every tap, scroll and screen without manual event tagging. Its AI analyst, Tara, goes through new users' sessions every day, flags the problems behind drop-offs and attaches the session replays that show them.
Heatmaps show where people tap on each onboarding screen, including taps on things that don't respond. Funnels show the exact step people leave, and retention analytics shows whether the people who finish onboarding come back.
Costa Coffee is a good example of how this plays out. Its team knew that 30% of new app users dropped off between downloading the app and finishing loyalty registration, but not why. With funnels and custom events in UXCam they found the biggest single cause: about 15% of new users dropped off after entering a password the app rejected.
A rejected password isn't a flaw you'd spot in a design review, because the screen looks correct every time. The team proposed a redesign of that step, and Costa reports a 15% increase in registrations.
A simple way to start on your own flow:
Pick your activation event and build a funnel from first open to that event, with every onboarding screen as a step.
Find the step with the biggest drop, and watch 20 or so replays of people who reached it and left. You'll usually see the problem within the first ten.
Fix the most common cause, then compare the next cohort's completion and day-7 retention with the last one.
If you'd like to learn this workflow properly, UXCam Academy's free Fundamentals course covers it in short video lessons, with a certificate at the end. And if you want to see it on your own app, you can start a free trial or book a demo.
Also worth a look
A few more flows that didn't make our main list but teach something specific:
Notion asks what you'll use it for during signup. Then it fills your sidebar with templates "selected for you based on what you tell us during onboarding", according to its help center.
Canva asked new users what kind of design they wanted and tailored the welcome message to the answer. Activation among people making posters rose 10%, according to an Appcues case study.
Hinge builds profiles around written prompts. It reports that likes on prompts are 47% more likely to lead to a date than likes on photos (Hinge, 2025).
Bluesky starter packs bundle recommended accounts and feeds into one invite link. They account for about 20% of all follows on the network, according to a 2025 study.
For more on the in-product side, see our guides to SaaS onboarding best practices, mobile onboarding best practices and user onboarding software.
How we picked and captured these flows
We started from the apps people most often cite for onboarding and kept the ones that teach one specific, transferable habit, plus three flows that show what to avoid.
Our team captured nine flows on iPhone and Android in July 2025 by creating new accounts: Amazon, ChatGPT (in Safari), Instagram, LinkedIn, SoundCloud, Temu, Tinder, Uber and YouTube. We walked the web flows (Airbnb, Duolingo, Linear, MyFitnessPal, Noom, Shopify and Wise) on 17 October 2026, logged out, in a fresh browser in Germany. We used a phone-sized window, except for Linear, which we walked at desktop size. On the web we used made-up answers and stopped at the first email or password field, so no accounts were created.
Onboarding changes all the time, so every image carries its capture date. We've hidden email addresses, phone numbers, usernames and names in the screenshots.
Numbers come from the companies' own blogs, help pages and filings, from named case studies and, for Noom's screen count, from a walkthrough on RevenueCat's blog. Each one is linked where it appears. Research findings link to the original paper or publisher. Where an effect is weaker or more disputed than its popular summary, we've said so.
Frequently asked questions
Which app has the best onboarding?
Duolingo is the most cited example. It gives you a real lesson before asking you to sign up, and its former VP of Growth said moving signup back increased daily active users by about 20%. But the best onboarding depends on the product. ChatGPT and Shopify show value before an account exists, Wise shows the price first, and Tinder turns 19 screens into building your profile.
What is app onboarding?
App onboarding is everything a new user goes through between opening an app for the first time and getting real value from it. It usually includes creating an account, a few questions about goals or preferences, permission requests and a first task, like Duolingo's first lesson or SoundCloud's artist picker.
What is user onboarding?
User onboarding is the process of getting a new user from signup to their first success with a product, on mobile apps or the web. It covers the first-run screens and also what happens after them: empty states, templates, checklists and the first emails. Good user onboarding is measured by activation and retention, not by how many people finish the tour.
How do you design a good app onboarding flow?
Start with the moment a new user first gets value, then remove or delay everything that doesn't lead there. Show value before asking for an account where you can, and keep each screen to one decision. Explain every hard step in one sentence, and ask for permissions right after a reason to say yes. Then measure where people drop out.
How many screens should an onboarding flow have?
There's no right number. Duolingo's web flow has about a dozen short screens before the first lesson, mostly single-choice questions, and Tinder's 19 screens each add something to your profile. Monzo's biggest US gain came from letting low-risk users skip one costly check, not from the screen count. Judge each screen by what it costs and what it gives back.
Should users sign up before or after trying the app?
After, whenever your product allows it. Letting people try first, as Duolingo, ChatGPT, Shopify, Linear and Airbnb do, turns signup into a way to save something they already have. Some products need an account first, like banks, wallets and email clients. In those cases, explain why each step is needed and show progress.
What psychology principles are used in app onboarding?
The most common is the endowment effect: people value what they already have, which is why "save your progress" is a popular way to ask for an account. The IKEA effect says people value what they helped make, and the goal-gradient effect says they speed up as a goal gets closer. Commitment and consistency, reason-giving and cognitive load (one decision per screen) round out the list.
What is the endowed progress effect?
The endowed progress effect is the finding that people are more likely to finish a goal when they're given a head start toward it. In the original study, car-wash customers given a ten-stamp card with two stamps filled in completed it 34% of the time. An empty eight-stamp card got 19%. In onboarding, it's the logic behind setup checklists that start with a step already ticked, like "account created".
What is the IKEA effect in onboarding?
The IKEA effect is the tendency to value things more when you helped make them. In onboarding, it's why Shopify builds a store from your own description before you sign up. It also keeps people going through Tinder's long profile builder: the more you put in, the more the result feels like yours.
What is an aha moment in app onboarding?
An aha moment is the point where a new user first experiences the product's core value. Examples are the first right answer in Duolingo, the first useful reply in ChatGPT and the first generated store in Shopify. Good onboarding gets people there quickly. Find yours by comparing what retained and churned users did in their first sessions.
When should an app ask for push notification permission?
After the user has seen a reason to say yes. SoundCloud asks after you've picked artists to follow. Calm asked right after the first meditation, and 40% of the people who saw that prompt set a daily reminder. Airship's 2026 benchmarks found that apps running onboarding campaigns see opt-in rates as much as 40% above their category average.
How do you measure onboarding success?
Choose one activation event that shows a new user got value. Then track the share of new users who reach it and how many come back on day 1, day 7 and day 30. A funnel with each onboarding screen as a step shows where people drop out, and session replays of those users show why.
AUTHOR
Silvanus Alt, PhD
Founder & CEO | UXCam
Silvanus Alt, PhD, is the Co-Founder & CEO of UXCam and a expert in AI-powered product intelligence. Trained at the Max Planck Institute for the Physics of Complex Systems, he built Tara, the AI Product Analyst that not only analyzes user behavior but recommends clear next steps for better products.

Related articles
Product best practices
16 Login Screen Examples, and the Psychology Behind Each One
16 login screen examples from Netflix, Slack, GitHub, Discord, Uber, Instagram and more, on iPhone, Android and the web, with the psychology that makes each design work or...
Product best practices
5 feature adoption metrics and KPIs to measure (with formulas)
The user research has been done, you’ve built a cool new feature for your app and finally released it. But, how do you know if your feature is a success? Find 5 KPIs to measure to improve your feature adoption...

Product Analytics Expert
Product best practices
5 Important Product Health Metrics and How to Track Them
Find out the 5 key product health metrics critical for success and learn how to track them effectively to optimize your mobile app product...

Product Analytics Expert

