User activation

Activation is the one stage where you win twice. Fix it and more of the people you have already paid for become customers, and every channel you run afterwards starts to look better than it did last month, on the same spend.

It is also the stage with the least ownership. Marketing owns the page. Product owns the first session. Sales owns the demo. The stretch between them belongs to nobody, which is why it leaks in the same three places in almost every company I have looked at. The page promises something the product does not show in the first five minutes. The signup asks for things nobody will read this week. The product waits for the new user to do work before it gives them anything worth having.

Before you read any further, build one report. Four numbers, in order:

1. Visit to signup, as a percentage.

2. Signup to the activating action, as a percentage.

3. Median time from signup to the activating action, in minutes or days.

4. The share of new signups who come back for a second session within 7 days.

You can build that in whatever analytics you already have, in an afternoon. The worst of those four numbers is your week's work. Everything below is about how to read each one and what to do about it.

What does the page promise, and does the product keep it?

Here is the exercise, and it takes 20 minutes. Write down, in one sentence, the promise your landing page makes. Then open a brand new account and write down, in one sentence, what a new user actually sees in their first five minutes. Put the two sentences next to each other.

If they are not the same sentence, you have found the leak, and no amount of button testing will close it. The page is selling a destination the product does not reach quickly. That gap shows up as a good signup rate and a terrible activation rate, which is the pattern that gets misread as a product problem when it started as a copy problem.

The promise has to be settled early. Nielsen Norman Group's 2018 eyetracking study, across 120 participants and more than 130,000 eye fixations, found that 57% of page viewing time happened above the fold and 74% within the first two screenfuls, with more than 42% of viewing time in the top 20% of the page (Nielsen Norman Group, April 15, 2018). People do scroll. They just decide before they do.

Speed is part of the promise too, and it is the cheapest part to fix. Deloitte and Google studied 37 European and American brand sites and more than 30 million user sessions, and found that a 0.1 second improvement in mobile site speed lifted retail conversion rates by 8.4% and lead generation form submissions by 21.6% (web.dev, June 24, 2020). The current thresholds to aim at are Largest Contentful Paint under 2.5 seconds, Interaction to Next Paint under 200 milliseconds and Cumulative Layout Shift under 0.1, measured at the 75th percentile of real users (web.dev, updated October 31, 2024).

The mistake that costs six weeks: redesigning the page before you have read what brings people to it. The ad headlines, the search queries, the referring pages. A page that answers a different question than the one people arrived with converts badly no matter how good it looks, and a redesign hides that rather than fixing it. Read the traffic first, in an afternoon, then change the page.

The two pieces that go deep on this are the landing page teardown checklist, which is the pass I run before I change anything, and Making Websites Win, which is where the research method comes from.

Which action predicts a customer?

One action, early, countable. Not a score, not an index, not a list of five.

The method, concretely. Take everyone who signed up 60 to 90 days ago, far enough back that you know how their story ended. Split them into two groups: still here, and gone. Then list every countable thing a new user can do in their first week, which is usually 10 to 15 events. For each event, work out what share of the stayers did it in week one, and what share of the leavers did it in week one. Put those two columns side by side and look for the biggest gap.

Then check the lift from the other direction. Among everyone who did the candidate action in week one, what share was still here at day 30? And among everyone who did not? The bar worth holding to is that your activators retain at least twice as well as your non-activators (Lenny's Newsletter, October 25, 2022). If the gap is smaller than that, you have found something true about your users rather than something you can act on.

Two ways this goes wrong. The first is picking a metric that only proves the user is alive: logged in twice, visited three pages. It correlates with everything and tells you nothing. The second is picking one you cannot influence, like "invited four colleagues" in a product where invites depend on the customer's own org chart. A good activating action is one your next release can move.

On benchmarks: in a survey of more than 500 companies, the average activation rate was 34% and the median 25%, and for SaaS alone the average was 36% and the median 30% (Lenny's Newsletter, October 25, 2022). The spread inside those numbers is very wide, and it has to be, because every company draws the line in a different place. Use the range to know whether you are in a different order of magnitude from everyone else. Use your own trend, week by week, for everything else.

One more honest point. The comparison above gives you correlation. The only way to know whether the action causes retention is to nudge a random half of new users toward it and watch what happens to the other half. That test is worth running once, early, because it settles the question that otherwise gets argued about for a year. Hacking Growth is where the loop for running that kind of test comes from.

How long should the first session take?

Measure it as a median and a 75th percentile, never an average. Averages in onboarding are decided by the people who left the tab open overnight.

What actually shortens it, in the order I try things:

Take steps out. Count the screens between the signup button and the activating action. Then count the fields. Baymard Institute's 2024 benchmark of ecommerce checkouts found an average of 5.1 steps and 11.3 form fields, where most sites need only 8 fields in total (Baymard Institute, June 26, 2024). B2B signup forms carry the same tax, usually worse, and usually because someone in a meeting two years ago asked for a field.

The rule I hold to: every field has to change what somebody does this week. If the answer sits in the CRM unread, it belongs in enrichment, or in the second session, or nowhere.

Fill the empty state. A product that greets a new account with zero rows has no first value to show. Seed it: sample data, a template, the first record imported from whatever they used before. The empty state is the single most common reason a first session ends without anything happening.

Defer, do not demand. Anything the user can do later should be asked for later. Verification, team invites, billing details, preferences. Each one is a place to lose someone who was about to succeed.

Then watch the second session. The first session is the one you designed. The second one tells you whether the design worked, which is why the 7-day second-session rate is the fourth number in that report at the top. A product can have a beautiful onboarding completion rate and a dead second session, and that combination means the first session taught people how to use something they did not need.

What should a trial be for?

A trial exists to produce one piece of evidence: that this works on the buyer's own case, with their own data. Not that the product has features. A trial spent touring features ends with a user who understands the product and has no reason to pay for it.

Which decides the length. Match the trial to one full cycle of the customer's own work, not to a round number. A tool somebody uses once a week needs more than 7 days to prove anything. A tool used every day proves itself in less than 14, and a 30-day trial on a daily-use product mostly buys the user permission to forget about it. If you cannot get one real cycle done inside the trial, extend the trial or change what the trial asks them to do.

On trial-to-paid benchmarks, I would rather be useful than confident. The published numbers vary so widely that they cannot steer a decision, because they mix opt-in trials with opt-out ones, self-serve products with sales-assisted ones, 7-day with 30-day, and free tiers with real trials. Any single figure you read has quietly picked one of those worlds. Your own conversion, measured the same way for three months in a row, is the number worth acting on.

Demos deserve the same question. A demo is an activation only if it ends with the prospect inside the product, looking at their own data. A demo that ends with a recording and a follow-up email is an acquisition event that has been filed in the wrong column, and it is why pipeline reports and activation reports so often disagree. The company in 2.4x the qualified demos on the same budget had exactly that problem, and fixing the definition came before fixing the page.