How to improve user activation: find the behavior that predicts a paying customer

Search "how to improve user activation" and you get the same list every time. Shorten the sign-up form. Add a checklist. Send a welcome email. Build a product tour.

None of it is wrong. Most of it won't move your number, because it skips the question that decides everything else: activated to do what?

At Deviniti, trial-to-paid conversion went from 26% to 59% in three months, and monthly recurring revenue rose 33%. Nobody redesigned the onboarding. We found one thing paying customers did in week three of their trial that nobody had noticed, and made it easier to do. I'll come back to what it was.

Here's the method in five steps:

1. Check it's an activation problem. Rule out retention before you touch onboarding.
2. Find the behavior that predicts paying. Compare the users who paid with the ones who didn't.
3. Turn it into a number a team can move. One behavior, one target, tied to revenue.
4. Ask why users stop short of it. Interview the people who stopped, about the moment they stopped.
5. Test the smallest fix. Write the success bar and the guardrails before anything ships.

Why bother with a method? Because activation is the cheapest revenue you have. You already paid to acquire every user who signs up, and every number after activation, from conversion to expansion, only works on the users who get through.

1. Check it's an activation problem at all

Start with the question most teams skip.

Is the money leaking at the start, or later on? If users activate fine and then drift away, you have a retention problem. A better onboarding will just pour more people into a leaking bucket.

At Deviniti, the team's own Tableau data settled this before anyone touched the product. Retention came back healthy: new revenue consistently exceeded churn, and every customer segment stayed beyond twelve months. Activation was where the money leaked. Around thirty trials arrived each month, and fewer than three in ten converted, against historical peaks above fifty percent.

Not sure which side you're on? Here's what good retention looks like for B2B products.

2. Find the behavior that predicts paying

This is the step that changes everything after it.

Most teams define activation as a milestone the product chose: finished onboarding, created a first project, invited a teammate. The useful definition is one your paying customers chose, by what they did in their first days or weeks.

To find it, split your users into two cohorts: the ones who became paying customers, and the ones who never did. Then put every hunch the team holds through the same three steps. Say what you think is happening. Go and look. Keep it or kill it.

At Deviniti we ran six hypotheses through that in PostHog. Five died. The sixth came back at a hundred percent: every trial customer who made actions available to end users on their customer portal during week three went on to pay. Not most of them. All of them.

A split that clean is rare, and you don't need one. At UX Pilot, users who tried Deep Design, a higher-fidelity generation mode, in their first seven days converted to paid at 12.3%. Users who didn't: 1.3%. Not a perfect predictor. A gap too large to ignore, which is all you need to start.

3. Turn it into a number a team can move

Nobody can "increase conversion" on a Monday morning.

Conversion is a lagging number. By the time it moves, the quarter is over. The behavior from step 2 is a leading number, and a team can act on it today: make one setting easier to find, call the customers who haven't found it yet.

That's your activation rate: the share of new users who do the behavior inside the window where it predicts paying. Then tie it to the business result it predicts. Deviniti wrote it as three layers:

• Business outcome: increase trial-to-paid conversion by 60%.
• Product objective: increase the availability of customer actions on end-user portals.
• Product outcome: increase the share of week-three trial users who make actions available, from 32% to 45%.

The team owns the last layer. Leadership watches the first. The middle one explains why moving the last should move the first. I've written more about why a revenue target backfires as a team goal.

Then do the arithmetic once: what is one point of that behavior worth in monthly revenue? From then on, every experiment gets read in money, not only in percentages.

4. Ask why users stop short of it

Analytics shows where people stop. It can't show why.

For that you need story-based interviews with the users who stopped, about the moment they stopped. Not "would you use a feature that..." but "tell me about the last time you...". At Deviniti we interviewed trial users in their third week, to find what stopped them making actions available. The experiments that followed were about one thing: how easy that setting was to find.

Sometimes the interviews overturn the obvious read. At elyps, a neobank where I led product, roughly eight in ten new sign-ups left at the screen before identity verification. Everyone blamed the ID check. Churned users told a different story: they had opened the app on the subway or at work, and didn't want to pull out their ID in public. The problem was the moment, not the step. A "Do it later" option with an 8pm reminder cut drop-off there from roughly 80% to 60%. The 3W Loop post walks through that loop end to end.

Talk to the users who got through, too. The contrast between the two groups is usually where the explanation sits. And don't skip the ones who left: churned users have no reason to be polite.

5. Test the smallest fix, and decide what must not get worse

Write the success bar before the test, not after.

At UX Pilot, the fix was switching Deep Design on by default for every first-time user. We ran it as a Bayesian A/B test on 120,000 users over 11 days, and set the success bar at +10% before it launched. Free-to-paid conversion rose 44.67%. The correlation from step 2 held up as causation, and the change went to 100% of users.

A success bar protects you from the result you want. Guardrails protect you from the one you didn't see coming. On Taxfix's Activation team, where I was embedded as a fractional Senior PM, we set three before any concept went live: activation conversion, booking conversion, and product regret (users who chose the expert service, then backed out of it). One concept lifted the number it was built for and quietly dragged the others down. The guardrails turned a plausible story into a clear no.

And keep the fix small. Deviniti made one setting easier to find, and had customer success reach out to trials in week three rather than at the end. UX Pilot flipped a default. elyps added a button and a reminder. None of them was an onboarding redesign, which is why each could be live and measured within weeks.

If your experiments tend to end in a slide rather than a number, design the experiment before the fix, and make sure it can fail. A test you cannot fail is a demo.

The best onboarding doesn't simplify

Here's what most activation advice gets backwards.

The standard playbook is subtraction: fewer fields, fewer steps, less to think about. Look at the three fixes above. UX Pilot gave first-time users more, not less: the fuller, higher-fidelity mode, on from the very first generation. elyps kept every verification step and only changed when users could take it. Deviniti didn't shorten anything. It pointed trial users at the one setting that made the product worth paying for.

The best onboarding doesn't simplify. It impresses. It gets a new user to the thing your paying customers already do, at a moment they're ready to do it.

The lasting change at Deviniti wasn't the 59%. It was measuring outcomes weekly instead of quarterly, and judging experiments on the product outcome, not on whether they shipped. That's the part that keeps paying.

If you own an activation metric and want to run this on it, that's what the User Activation Sprint is for: one metric, one quarter, a new experiment every week.