The two numbers that predict whether a company works. Acquisition can be bought; these cannot.
12 min read · Level 2
Acquisition can be bought and these two cannot, which is why they predict everything else. Activation is where most of the loss happens and where founders look least, because it sits between two things that are easier to measure. And the definition matters more than the measurement: if your activation metric does not separate people who stay from people who leave, you have chosen the wrong action and every conclusion drawn from it will be wrong.
Activation is the moment a new user first experiences the core value. It must be a specific, observable action.
Retention is the proportion still active after a period, measured by cohort.
Activation predicts retention; retention predicts everything else.
Because acquisition is purchasable and these two are not. You can buy a million signups; you cannot buy people coming back.
And because they are where most of the loss happens. A company losing 90% at activation and 85% by week 8 does not have a marketing problem, however much it feels like one.
Defining activation properly matters more than it sounds.
Weak: 'created an account'. Everyone does this and it predicts nothing. Weak: 'logged in twice'. Correlates with retention without causing it. Strong: 'published their first rota'. This is the moment the product delivered its promise.
Once defined, the numbers become actionable:
Signups 320 Activated 26 (8%) Of activated, week-8 retained: 71% Of non-activated, week-8 retained: 2%
That last pair is the whole argument. Activated users retain at 71%; everyone else effectively vanishes. So the single highest-value thing this company can do is get more people to that first published rota — not acquire more signups, not add features.
Find the equivalent number for your product and it will usually reorganise your roadmap.
First-time founders often define activation as something easy to measure rather than something that means anything. If your activation metric does not separate retained users from churned ones, it is the wrong metric.
The second mistake: measuring retention in aggregate rather than by cohort. Aggregate retention rises simply because new signups are counted as active, hiding the fact that older users are all leaving.
The third: optimising onboarding without watching anyone use it. Five recorded sessions will tell you more than a month of funnel analysis.
The specific action where the product first delivers what it promised.
Compare retention of activated versus non-activated users. A big gap means you picked the right action.
Group by join week and follow each. Aggregate numbers hide churn behind new signups.
The specific step where they stop. Session recordings beat speculation every time.
How long from signup to value? Shortening it usually raises activation more than any redesign.
Every acquisition dollar is multiplied by your activation rate. Fix the multiplier first.
From your first fifty users onward, and continuously after.
Do not measure cohort retention with two weeks of data. Curves need six to eight weeks before they say anything.
Facebook's growth team is widely reported to have identified a specific early-usage threshold — connecting with a number of friends within a short window — that strongly predicted long-term retention, and then oriented onboarding around reaching it. The exact figure matters less than the method: find the behaviour that predicts staying, then engineer toward it.
Look for the action that separates retained users from churned ones, then redesign onboarding around reaching it faster.
Measured in aggregate, retention can rise while every individual cohort collapses — because new signups are counted as active and mask the decay of everyone who came before. It is one of the easiest ways to feel confident about a product that is quietly failing.
Always measure retention by cohort. The aggregate number flatters you exactly when you most need the truth.
Create a free account to read the rest and apply it to your own startup.