Onboarding Psychology: Fix Week One So Users Don't Churn in Month Three
The churn that shows up in month three was mostly decided in week one. Onboarding is a behaviour change problem, and the psychology research on cognitive load, defaults, endowed progress, implementation intentions and habit formation tells you what to change. Here's what applies at each stage, what the studies actually found, and an audit to score your own flow.
TLDR: The churn that shows up in month three was mostly decided in week one. Onboarding is a behaviour change problem, and psychology has decent evidence about what changes behaviour. Here's what applies at each stage, and what the studies actually found.
Setup: cut the load. People hold about four things in working memory, most of them keep whatever you pre-select, and a checklist that starts with honest credit gets finished far more often.
Aha: get to value in the first session, let them build something of their own and make sure they finish it, and end on a win.
Habit: ask when they'll next use you, prompt on their natural rhythm, and keep onboarding going well past the trial. Habits took a median of 66 days to form in the best real-world study.
Some famous findings don't hold up as well as their reputation. Choice overload is the obvious one. Design from the robust ones and test the rest.
Why does onboarding decide churn months later?
Because there are two ways a new customer fails, and onboarding causes both.
The first is that they never get value at all. They sign up, poke around, get stuck on setup and leave. You see this as a steep drop between period 0 and period 1 on your retention curve. If you plot your cohorts as separate lines and the gap between good and bad cohorts opens in that first period, the problem is activation. I walked through reading that in how to plot a retention curve.
The second is slower. They get value once, maybe twice, and never build a routine. Nothing goes wrong. They just stop coming back, and in month two or three the card gets cancelled during a budget review. That churn gets logged as an engagement problem, but its cause was an onboarding that ended after setup, as if getting the account configured were the goal.
Most onboarding is designed to get people set up. Retention depends on getting them into a routine, which is a different job and takes much longer.
Reforge splits activation into three moments, each with its own question in the user's head. I find it the most useful frame for applying psychology, because different research applies at each one.
What does psychology actually say about changing behaviour?
The simplest model I know is BJ Fogg's behaviour model: a behaviour happens when motivation, ability and a prompt show up at the same moment. Onboarding can work on all three, but the cheapest lever is nearly always ability. You can't do much about how motivated somebody is when they sign up. You can make the next step much easier, and you can make sure something prompts them at the right time.
One warning before the useful part. Behavioural science went through a replication crisis, and some of the findings that turn up in every onboarding blog post didn't survive it well. I've linked the original paper for every claim below, and I've said so where the evidence is weaker than its reputation. Treat all of it as a reason to run a test, not a guarantee.
How do you get new users through setup?
At the setup moment the user is asking "what do I do now?" They don't understand your product yet, so everything takes effort. The job is to cut that effort.
Cut cognitive load: one decision per screen
John Sweller's cognitive load research starts from a simple fact: working memory is small. Nelson Cowan's review puts it at about four chunks at once, not the seven that people usually quote. A new user is already spending most of that on working out what your product is. Every extra field, option and tooltip competes for the rest.
In practice:
Ask only what you use in the first session. Company size, role and "how did you hear about us" can wait until they're relevant, or until the user is invested enough not to mind.
One decision per screen. A screen that asks for a workspace name, a template choice and a team invite is three decisions competing for the same four slots.
Show advanced options later.Progressive disclosure keeps the first screen simple without taking power away from people who want it.
Recommend a default path instead of a menu
Hick's law says decision time grows with the number of options. The more famous claim, that too many options stop people choosing altogether, comes from Iyengar and Lepper's jam study: 30% of shoppers who stopped at a table of 6 jams bought one, against 3% at a table of 24.
So I wouldn't redesign onboarding around choice overload. I'd redesign it around defaults, which have much stronger evidence. In Johnson and Goldstein's "Do Defaults Save Lives?", countries where organ donation was the default had effective consent rates of 86% to almost 100%. In countries where people had to opt in, the rates ran from 4% to 28%. Same decision, different default, wildly different outcome.
For onboarding, that means pre-selecting the template most customers end up using, filling in sample data so the product isn't empty, and making the recommended path the obvious one. Keep "start from scratch" available as a smaller link. Most people will take the path you set, so set the one that gets them to value fastest.
Give honest credit for progress already made
This is my favourite finding in the area, and it has held up well.
Nunes and Drèze gave car wash customers one of two loyalty cards. One needed 8 stamps for a free wash. The other needed 10, but came with 2 already stamped. Both needed exactly eight more washes. The card with the head start was completed 34% of the time. The plain card, 19%.
The onboarding version is a setup checklist that starts partly complete, with real credit: "Create your account" and "Verify your email" are genuinely done, so tick them. Keep the remaining list short, so the finish line looks close. What you shouldn't do is invent fake steps to tick, or show a progress bar that doesn't match reality. People notice, and then they stop trusting the rest of your interface.
How do you get users to the aha moment faster?
At the aha moment the user is asking "why should I use this?" The answer has to come from experience, not from your marketing. If you can't name your aha moment in one sentence, start with finding your aha moment.
Put value in the first session
People weight rewards they get now much more heavily than rewards they're promised later. A product that pays off after a week of setup is competing with everything else that's paying off today. That's why the single highest-impact onboarding change is usually to move the first real value event into session one. Seed the account with sample data, import their real data on day one, or do the first run for them. The activation milestones playbook walks through finding and shortening that path.
Let them build something, and make sure they finish it
The IKEA effect study by Norton, Mochon and Ariely found people valued things they'd assembled themselves more highly than identical ready-made ones. The part most people skip: the effect went away when participants built something and then had it taken apart, or didn't finish it.
Get weekly churn reduction tips.
No spam, unsubscribe anytime.
You're in.
Check your inbox shortly.
That's a useful pair of instructions. Have new users make one real thing of their own during onboarding (a first report, a workflow, a page, an automation), because they'll value the product more for it. And keep that first thing small enough to finish in one sitting. A half-built dashboard abandoned on day one isn't an investment. It's a reminder of something they didn't finish.
End the first session on a win
People don't remember an experience as an average. They remember its best moment and how it ended. That's the peak-end rule, and it applies to a first session as much as to a support call. A first session that ends on a settings page or a half-finished import gets remembered as admin. One that ends with "here's your first report, and it found three accounts that went quiet" gets remembered as the product working.
The related move is to show the outcome in numbers inside the product: time saved, money recovered, a problem caught. Value people can't see doesn't get credit when the renewal comes round.
How do you turn a first win into a habit?
This is where most onboarding stops, and where most later churn is decided. The user has seen the value. Now they're asking "when should I use this?" and if nothing answers, their old routine wins by default.
The onboarding version takes one screen. At the end of the first win, ask "when will you next do this?" and offer a calendar invite: "Every Monday at 9, review last week's churn." It costs almost nothing, and it gives the habit a slot in a real week instead of hoping one appears.
Prompt on their rhythm, triggered by what they did
A prompt is the third leg of Fogg's model, and the most common mistake is timing. A fixed drip that sends day 1, day 3 and day 7 emails ignores what the user actually did. Trigger prompts from behaviour instead (they created a report but never shared it, they imported data but never came back), and time them to the product's natural usage frequency. I wrote about finding that frequency in trial length should match your usage frequency, and the behavioural retention emails playbook covers the emails themselves. Nir Eyal's hook model describes the same loop, from prompt to action to reward to investment, from the product side.
Keep onboarding going past the trial
The number most onboarding ignores: in Phillippa Lally's 2009 study, the median time for a new daily behaviour to become automatic was 66 days, with a range from 18 to 254.
Lally's participants were building personal habits like drinking water with lunch, not adopting software, so don't treat 66 days as a precise target. The direction is what matters. If your onboarding ends when the trial ends, it ends before the habit has had a chance to form, and the 52 days in between are exactly when quiet month-two and month-three churn gets decided. Keep behaviour-triggered help running for the first 60 to 90 days. The onboarding email playbook and what good first-month retention looks like are useful next reads.
Make what they'd lose visible, honestly
Kahneman and Tversky's prospect theory showed that losses weigh more than equal gains, and later estimates put loss aversion at roughly twice. The legitimate version for onboarding is to make the value a customer has stored with you visible: their history, their setup, the reports they've built, the results you've tracked. Someone who can see three months of work in your product thinks harder before leaving than someone looking at an empty dashboard.
The illegitimate version is making cancellation hard or guilt-tripping people on the way out. Those are deceptive patterns, and they don't reduce churn so much as delay it and make it angrier.
Time re-engagement for fresh starts
One smaller finding worth using. Dai, Milkman and Riis's fresh start research found people are more likely to start aspirational behaviour after temporal landmarks: the start of a week, a month, a year, or a birthday. If you're nudging a user who has gone quiet, a Monday or the first of the month is a better moment than a random Thursday. It's a small edge, and it costs nothing to schedule.
Score your onboarding
Twelve questions, four for each moment. Answer honestly for what a new user experiences today, not what's on the roadmap. You'll get a score per moment and the three changes most likely to matter.
How well does your onboarding use what we know about behaviour?
Pick No, Partly or Yes for each. Results update as you go.
Answer a few questions to see your scores.
Weights reflect how directly each item affects early and later churn, not how easy it is to fix. Nothing you click leaves your browser.
Where these scores come from: each answer scores 0 for No, 1 for Partly and 2 for Yes, multiplied by the item's weight. Items weighted 3 are the ones closest to the churn mechanism: whether people can reach the first useful screen without friction, whether a default path exists, whether value arrives in session one, and whether you ask for a specific next-use plan. Everything else is weighted 2. A moment's score is its weighted points divided by the most it could have scored, counting only the questions you've answered. The three fixes are the answered items losing the most weighted points, so a No on a weight-3 item ranks above a Partly on a weight-2 one. The weights are my judgement from the evidence above, not a fitted model. Use the result to decide what to test first, not as a grade.
Where does onboarding psychology go wrong?
Three ways, and all three get recommended in good faith.
Treating a single study as a law. Choice overload is the obvious example, but it's not the only one. Results from a lab or a car wash are a reason to try something in your product, not proof it will work there. Run it as an experiment, measure activation and period-1 retention, and keep what moves them.
Using it to trap people. Fake progress, manufactured urgency, cancellation mazes. They can lift a short-term number while making the customer trust you less, and a customer who feels tricked leaves at the first opportunity and tells people why. The test I use: would the customer be fine with it if they saw exactly why you designed it that way? Getting them to value faster passes. Making it hard to leave doesn't.
Optimising setup and calling it done. The setup moment is the easiest to measure, so it gets the attention. But a customer who breezes through setup and never builds a habit still churns, just later and more quietly. The habit moment is where most of the remaining retention is.
What to do this week
Run the audit above on your own onboarding, honestly, from a fresh account. Then pick the lowest-scoring moment and make one change there.
If you're not sure which moment to start with, look at your retention curve. A steep drop in the first period points at setup or aha. A curve that holds early and then slides through months two and three points at habit. If you'd like a quicker read on where your biggest leak is, the free 60-second churn health check asks a few questions and tells you.
Setup gets people in. The aha moment gives them a reason. The habit is what keeps them. Most onboarding only does the first one.
Answers to the questions I get most often about this topic.
How does onboarding affect churn later on?
Onboarding decides whether a new customer ever gets value, and whether using you becomes a habit. Customers who never reach the core value event leave early, which shows up as a steep drop in the first period of your retention curve. Customers who get value once but never build a routine leave later, often in month two or three, and that churn usually gets blamed on engagement when the cause was an onboarding that stopped too soon.
What psychological principles apply to SaaS onboarding?
The ones with the best evidence are cognitive load (people can hold about four things in working memory at once), the default effect (most people keep whatever is pre-selected), endowed progress (people work harder toward a goal they think they have already started), the IKEA effect (people value what they built, if they finished it), implementation intentions (a specific when-and-where plan makes follow-through much more likely) and habit formation research (habits took a median of 66 days to form in one real-world study).
What is the endowed progress effect in onboarding?
It is the finding that people are more likely to finish a goal when they are given a head start on it. In Nunes and Drèze's 2006 car wash study, a 10-stamp loyalty card with two stamps already filled was completed 34% of the time, against 19% for a plain 8-stamp card, even though both needed eight washes. In onboarding, it means a checklist should give honest credit for steps the user has already done, such as signing up and verifying their email.
How many steps should SaaS onboarding have?
As few as it takes to reach the first real value event, and each screen should ask for one decision. There is no universal number. The better test is whether every field and step is needed in the first session. Anything that can be asked later, when it is relevant, should be, because each extra item adds to the load a new user is carrying while they are still working out what your product does.
How long should onboarding last?
Longer than most teams run it. Setup can be over in minutes, but in Lally's 2009 study habits took a median of 66 days to become automatic, with a range from 18 to 254 days. Many SaaS onboarding sequences stop around day 14, when the trial ends. Keep behaviour-triggered help, prompts and progress reports going for the first 60 to 90 days, sized to how often your product is naturally used.
Does choice overload really hurt onboarding?
Sometimes, and less reliably than the famous jam study suggests. Iyengar and Lepper found shoppers offered 6 jams were far more likely to buy than shoppers offered 24, but a 2010 meta-analysis of 50 experiments by Scheibehenne and colleagues found the average effect was close to zero. It depends on the situation. The safer onboarding advice is to recommend one default path, because a good default helps whether or not choice overload is present.
Is using psychology in onboarding manipulative?
It is when it works against the user. Fake progress bars, guilt-tripping copy and hidden cancel paths are deceptive patterns, and they tend to move churn later rather than prevent it, because a customer who feels tricked leaves as soon as they notice. The test I use is whether the customer would be happy if they saw exactly why you designed it that way. Helping someone reach value faster passes. Making it hard to leave does not.
What is the first onboarding change most SaaS products should make?
Move the first real value event into the first session. That often means seeding the account with sample or imported data, choosing a default template for the user, and cutting every signup field that is not used straight away. It is the change that most directly reduces the early drop in the retention curve, and the other principles in this post work better once it is in place.
Founder of ChurnTools. I spend my time studying how SaaS companies lose customers and building tools to help them stop. Previously worked in SaaS growth and retention across multiple B2B products. I also write about growth and answer-engine optimization (AEO) at growthpigeon.com.
One tap, no signup. It builds your retention map as you read.
You have flagged enough to be worth costing out. The Health Check scores the same
dimensions properly and tells you which gap is losing you the most money.
Knowing the number is not the same as knowing the cause
A calculator tells you how much is leaving. It cannot tell you which part is failed payments,
which part is people who never activated, and which of the two is cheaper to fix.
I work that out on your actual data and send back three fixes ranked by what each one is worth,
within 10 working days.
✓ Voluntary and involuntary churn, split and costed