पाठशाला Pathshala · उत्पाद Utpād, The product · Lesson 06 · Build

Activation: the first session that decides everything

Most people who sign up never see the thing the product is for. Define the moment they do, measure the share who reach it and redesign the first session until that share doubles.

Pathshala, The Founder Library · 11 October 2026 · 8 min read

Quettra’s data from 125 million Android devices found that the average app lost 77 per cent of its daily users within three days of install. Its chief executive’s reading was that users decide within the first three to seven days which apps they will stop using. The first session is not the beginning of the customer relationship. For most sign-ups it is the whole of it.

This lesson defines activation so it can be measured, gives the method for finding the aha moment in your own data, sets the benchmarks, and then spends most of its length on the only thing that moves the number: the redesign of the first session. The figure in the middle shows what doubling the rate does to active users three months later, with nothing else about the product changed.

What activation is, precisely

Lenny Rachitsky and Yuriy Timen’s benchmark study gives the working definition: the activation rate is the percentage of new users who hit the activation milestone, and the milestone is the earliest point in onboarding that predicts long-term retention. Their test for whether you have chosen the right milestone is quantitative: users who reach it should retain at least twice as well as those who do not. The formula is users who reached the milestone divided by users who completed sign-up. Not visitors, not downloads; the denominator is people who got through the door.

The word to avoid is onboarding completion. Finishing a tour, filling a profile and granting permissions are steps you asked for, and a user can do all of them without once experiencing the product. The [glossary](/library/glossary) puts it as the moment a new user first gets the value the product promises. The aha moment is the first time the product does for the user the thing they came for: the first invoice that reaches a customer, the first ride that arrives, the first report that shows something the user did not know.

Two famous milestones, and why the numbers do not matter

Facebook’s growth team is remembered for seven friends in ten days. Slack’s is remembered for two thousand messages. Stewart Butterfield told First Round in 2015 that any team which had exchanged two thousand messages had really tried Slack, that 93 per cent of those teams were still using it, and that the number was chosen by looking at which teams stuck and which did not: for a fifty-person team it is about ten hours of messages, for a ten-person team about a week. The same article notes that more than 90 per cent of people who signed up never invited anyone or started using the software, which is the gap the milestone was built to close.

Andrew Chen’s note on the Facebook figure makes the point founders miss: the exact numbers are interchangeable, and ten friends in twelve days would have done as well. The value of the milestone is that it is dead simple to talk about and that the whole team can drive towards it. A milestone nobody can repeat in a sentence will not be worked on.

Finding yours: correlate, then test

Chen’s method is six steps and a startup with three months of data can run it in a day. Define the success metric: days active in the first twenty-eight, or retained at month three, whichever the [retention lesson](/library/retention-curves-and-flattening-test) has already made you measure. Build a table of recent users with that metric beside a dozen candidate first-week behaviours: invited a colleague, added a product, sent a first invoice, connected a bank account, used it on two separate days. Look for the behaviours that move with the success metric, remembering his warning that fire engines correlate with house fires without causing them. Then test the strongest candidate by pushing more new users through it and watching whether their retention follows. If it does, you have a milestone. Then make it simple: one action, one time limit, one sentence.

Two checks on the result. It must be early, inside the first session or the first day for most products, because a milestone at week two is a retention metric wearing a different name. And it must pass the two-times test: if users who reach it retain only a little better than those who do not, you have found a correlate of curiosity rather than of value.

What good looks like

Rachitsky and Timen’s survey of more than five hundred companies’ self-reported figures found an average activation rate of 34 per cent and a median of 25. For software alone, excluding marketplaces, e-commerce and direct-to-consumer, the average was 36 and the median 30. Marketplaces and e-commerce had the lowest rates, consumer freemium the highest. They define good as the 60th percentile and great as the 80th, so a company at the median has a lot of room before it is good. The numbers are self-reported and each company’s milestone is its own, so treat them as the scale of the opportunity rather than a target: a product at 20 per cent activation is losing four in five of the people who wanted it enough to sign up, before the product has had a chance to fail on its own merits.

The figure uses two retention rates on purpose. Activated users retain at one rate and the rest at another, and the second is never zero, because some people find their way without onboarding. If your own two numbers are close together the milestone is wrong, and the readout says so.

A worked example: the first invoice

A Hindi-first invoicing app for small traders in Indore gets 10,000 sign-ups a month from app-store listings and WhatsApp referrals. The team’s analysis finds one behaviour that predicts month-three retention far better than any other: sending a first invoice to a real customer on WhatsApp within the first session. 22 per cent of sign-ups do it. Of those, 55 per cent are still invoicing in month three. Of the rest, 8 per cent are. That is a seven-times gap, so the milestone is right, and the month-three cohort is 10,000 × (0.22 × 0.55 + 0.78 × 0.08), about 1,830 traders.

The first-session recording shows why 78 per cent never send one. The app asks for business name, address, GST number, logo and bank details before it shows an invoice screen; the trader is at a counter with a customer waiting; the logo upload fails on the shop’s phone; the invoice screen, when reached, is empty and the trader does not know what a complete one looks like. Four changes over six weeks: ask only for a phone number and a GST number, and pull the business name and address from the GST record; show a filled sample invoice as the first screen, editable; let the trader send the first invoice to themselves to see it arrive; move logo and bank details to after the first send, with a reminder. Activation goes from 22 to 41 per cent. Month-three actives rise to about 10,000 × (0.41 × 0.55 + 0.59 × 0.08), roughly 2,730, a 49 per cent increase in retained traders with no change to anything the product does after the first invoice.

Onboarding is not the part before the product. It is the part of the product most of your users will ever see.

Redesigning the first session

Count the steps before the aha and remove them. Every screen between sign-up and the first result is a place to lose a third of the people who reach it. Profile, preferences, permissions, tours and invitations go after the first result or go away. Do the setup for the user. Import their data, pull it from a public record, pre-fill from what you already know, start with a working template rather than a blank page. The Collison installation in Paul Graham’s essay, the Stripe founders taking the customer’s laptop and setting it up on the spot, is the manual form of this, and the lesson is that the setup is your job, not the user’s. Deliver a real result inside the session. A sample is better than an empty screen; a real result, the first invoice actually sent, the first report on their own numbers, is better than a sample. Measure time to the aha in minutes, as a median by weekly cohort, and treat it as a product metric. Make the milestone visible to the team: one sentence, one number, on a wall.

For India add four specifics. Phone number and OTP, not email, because the email account is often unused and the OTP arrives on the device in hand. WhatsApp as the first-session channel where it fits, since the user already knows how it works and the result can arrive there. The first session happens on a ₹8,000 phone over a patchy connection, often at a counter with someone waiting, and a step that takes forty seconds to load is a step most users will not finish. Where a KYC step is legally unavoidable, do it inside the session with the e-KYC route rather than as a gate before the product is shown, and show the value first wherever the rules allow.

Five mistakes make activation look fine when it is not. Measuring tour completion rather than value received. A milestone too late, which is retention, or too early, which is a login, and either way fails the two-times test. Optimising the sign-up page when the loss is after it; a cheaper sign-up that feeds the same leaking first session buys more leakage. Asking for data you want before giving the result the user wants; every field before the aha is a fee you are charging in attention. Treating activation as a one-time project: every new acquisition channel brings a different user to the first session and the rate moves with the mix, so measure it by cohort and by channel, forever.

A weekly ritual, in twenty minutes

Every Monday, for the sign-ups of the previous week: the activation rate, by channel; the median minutes from sign-up to the aha; and the share of the week before last’s activated and non-activated users still active, to confirm the two-times gap holds. Watch one new user’s first session, recorded or in person, and write down the step where they hesitated. Then choose one change to the first session to ship this week, with the activation rate it is expected to move and by how much. Fifty weeks of this is fifty experiments on the part of the product most of your users will ever see, and the company that runs them will have twice the retained users of the one that built features instead, from the same sign-ups.

The benchmarks are self-reported figures gathered by others and your milestone is your own; the sources are below and the Slack and Facebook stories are worth reading in full for how the numbers were chosen.