They sign up and never come back: how to build onboarding and activation

Most teams answer this with "let's add an onboarding screen" and the number does not move. Users who sign up and disappear usually do so because nothing useful happened in the first session, not because a welcome screen was missing. Onboarding is not explaining the product, it is getting a user to a real result with their own data by the shortest route available. The order matters: define what that result is, remove the obstacles on the way to it, and only then consider tooltips and product tours.

Activation is not signup

The definition of activation rate is simple enough: the share of people who completed signup and then reached a defined milestone. Choosing the milestone is the hard part. Lenny Rachitsky's write-up on activation rates describes it as the earliest moment that demonstrates value and predicts long-term retention, and lists the usual mistakes: setting the bar too early, setting it too late, or picking something most users cannot realistically reach.

An early bar feels good and teaches nothing. If "filled in the signup form" counts as activation, your rate will sit around ninety percent and guide no decision. A late bar is equally useless: "issued their third invoice" tells you nothing you can act on during a user's first week. The useful point sits between the two, at the first moment a user sees with their own eyes what the product does for them. Created their first record. Pulled their first report. Sent their first invoice.

One caution about the industry averages that circulate in benchmark posts. Every company defines activation around a different behaviour, so those numbers are not comparable to each other. Where your own rate sits against last quarter says far more than somebody else's percentage.

Find the activation moment in your data, not in a meeting

There is a practical way to derive it instead of voting on it:

1. List candidate behaviours: creating the first record, inviting a teammate, connecting an integration, downloading the first report, installing the mobile app.
2. Group users into cohorts by signup week.
3. For each candidate, compare week-four and week-eight return rates between users who did that thing in their first seven days and users who did not.
4. Pick the behaviour where the gap is wide and a meaningful share of users actually gets there.

The classic trap here is reading correlation as causation. "Accounts that invite five colleagues stick around" does not mean that forcing five invitations will make accounts stick; the accounts that were already serious are the ones inviting people. The way to check is to make that step noticeably easier and watch whether retention in later cohorts actually moves.

There is outside evidence that the first week is worth the attention. Amplitude's 2025 benchmark work, built on product data from more than 2,600 companies, presents first-week activation as the strongest predictor of three-month retention, reporting that 69% of products in the top quartile on day-seven performance were still in the top quartile at three months. That is vendor data and a correlation, but it captures why investment in the first week shapes everything that follows.

In B2B, activation is an account-level measure

Unless you sell a single-player tool, per-user activation misleads you. The person who made the buying decision explores everything on day one. They saw the demo and know what to look for. Activation actually happens when the team starts using the thing.

An account-level definition looks more like this: at least three users signed in, real data was imported, one integration was connected, and at least two people from the team used the product during a normal working day in the first week. Compare renewal rates between accounts that clear those bars and accounts where only the manager ever logs in, and the gap shows up in your own data. The definition also gives sales and support a shared vocabulary, which ends the recurring argument about what counts as an active customer and turns at-risk accounts into a list.

The real blocker before the first session is setup work

In most products what stops a user is not the interface, it is that first value sits behind setup. Importing data, inviting the team, connecting the accounting system, configuring templates. All of it falls on the user and none of it shows what the product does.

Three approaches help. The first is moving first value ahead of setup: it should be possible to create an order with a single product before the whole catalogue is imported, to produce a timesheet for one person before all staff are loaded. The sequence of "configure everything, then use it" kills the first session.

The second is making import genuinely easy: a ready template, a column-mapping screen, a flow that sets bad rows aside and carries on with the rest. We covered how to design that in the piece on bulk import from Excel and CSV, and if people are coming off an older system, data migration is part of the same problem.

The third is doing the setup yourself for a while. "Send us the file and we will load it" does not scale, but it teaches: by the third customer you know which step is genuinely hard, and you stop guessing about what to automate.

A product tour does not fix an interface that cannot explain itself

The product tour is most teams' first instinct because it can be bolted on without changing anything in the product. The evidence does not support the instinct.

In a study Nielsen Norman Group published in 2020, 70 users across four mobile apps were split into a group that read the card-deck tutorial on launch and a group that skipped it. Both groups were then given a task the tutorial had covered. Task success was 91% among readers and 94% among skippers, and the difference in completion time was not statistically significant either. The one significant difference went the other way: readers rated the task as harder (4.92 versus 5.49 on a seven-point scale). A tour can make a simple app look more complicated than it is.

The rule that follows is to give help when the user gets stuck, where they get stuck. A five-step tour is usually a sign that the interface cannot explain itself, and the fix is the screen, not a prettier tour.

The empty state is your cheapest teacher

The first thing a new user sees is almost always an empty list. In most products it says "No records found" and nothing else.

Kate Kaplan's guidance on empty states at NN/g gives that space three jobs: communicate system status (nothing here yet, nothing has gone wrong), teach what belongs in this area, and open a direct path to the key task.

In practice that means a "Create your first order" button on the empty order list, one sentence of explanation beside it, and where possible an option to fill the screen with sample data. Sample data sounds like a small detail, but seeing a product full is much easier than imagining it full. Templates do the same job: changing a ready-made structure takes far less effort than building one from nothing.

Ask for less at signup, ask for the rest in context

Every required field in a signup form is a chance to quit before activation. Company size, industry, phone number: those make life easier for the sales team, not for the user trying to reach a first result. Ask for what you really need later, after the user has seen something work, on the screen where the answer is actually used.

Verification email is a quiet activation killer too. If the message is slow or lands in spam, the user is lost in the middle of the first session and most of them do not come back. Why transactional email fails to arrive is its own subject, but since it sits at the top of your funnel, measure the loss at that step separately.

The message that brings a stalled user back is specific, not generic

In most products the post-signup email sequence goes out to everyone in the same order. Meanwhile you already know where each user stopped.

The difference is between "three great features of our product" and "you invited your team but no records have been created yet, here is the screen where the first one takes two minutes". Behaviour triggers it, the content is specific to that behaviour, and the timing sits close to the moment the user walked away.

The same holds for in-app messages. Channel choice and frequency caps matter, so the rules in notification system design apply here as well. Send the same reminder three days running and what goes up is your unsubscribe rate, not your activation rate.

For internal software, delivery is the start of activation

The same problem wears a different costume in custom software projects. The system is delivered, training is run, and three months later the team is still updating the old spreadsheet.

Here activation means adoption, and adoption is measurable: how many people from each department sign in on how many days per week, how many process steps get completed in the system, which work still happens outside it. Without those numbers you are not managing adoption, you are hoping for it.

A few things work. One internal owner per department, someone closer at hand than you are when a question comes up. A written date for switching the old method off, because as long as two systems run in parallel the habitual one wins. And developer capacity reserved for the first two weeks: what blocks adoption is often a small piece of friction that takes three days to fix, but left in a queue it hardens into an opinion about the system. In the user acceptance testing and go-live process, the four weeks after the sign-off decide whether the project is considered a success. We wrote about the resistance to expect when a team moves from spreadsheets to software separately.

You cannot fix a funnel you do not measure

Three things are enough to manage activation: a funnel whose steps are marked by events, cohorts grouped by signup week, and a weekly look at the drop-off at each step. Instrumenting events properly is a discipline of its own, covered in the piece on product analytics and event tracking.

Numbers tell you where people fall out, never why. The why lives in support tickets, session recordings and fifteen-minute calls with five users. The two work together and neither is enough alone.

There is a traffic question when you measure changes, as well. Most B2B products do not have the volume for a classic A/B test inside the activation flow. A before-and-after comparison across cohorts, combined with the alternatives in the piece on how many users an A/B test needs, is a more honest route.

What you can do this month

• Write your activation definition in one sentence and date it: which behaviour, within what window, measured per user or per account.
• Test two candidate behaviours against the last three months of cohorts: is week-eight return meaningfully higher for users who did it in week one?
• Count the required fields in your signup flow and make every one that is not needed for first value optional or deferred.
• Open the three empty screens new users see most, and give each a one-line explanation, a single primary button and, where it fits, a sample-data option.
• List the last 20 users who stalled at the biggest drop-off in your funnel and talk to five of them for fifteen minutes each.
• If you are rolling out internal software, put the shutdown date for the old method and the first two weeks of fix capacity into the plan now.