Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Metric

Activation Rate

The share of new users or sign-ups who complete the activation event, the action marking first real value, within a set window.
AnalyticsmetricUpdated

Formula

New users who complete the activation event within the window ÷ New users in the cohort × 100

Unit: % · higher is better

Definition

The activation event is a single, written action chosen as the early sign that a new user has experienced the product's core value. It is not picked because it is easy to count. The tutorial on finding the app activation event selects it from the product's own data, as the early action that most clearly separates users who stay from those who leave, and then counts it identically in analytics, the measurement partner and the ad networks. In the AARRR framework, activation is the stage between acquisition and retention, and this event is what makes that stage countable.

Activation rate is the share of a group of new users who complete that event within a set window.

New users who complete the activation event within the window ÷ New users in the cohort × 100(%)

  • New users in the cohort are the people who installed, signed up or started a trial in the same period, such as a week. They form a fixed group, read together; later arrivals belong to a later cohort.
  • New users who complete the activation event within the window are members of that cohort who performed the qualifying action, counted once at its first occurrence, inside the window measured from each person's own start.
  • The window is the number of days after arrival in which the event counts. It is part of the definition: a rate read over a few days and a rate read over a fortnight are different numbers for the same product.

How to interpret it

Higher is better, provided the event still predicts that people stay. The figure is read in three cuts.

  • By cohort. Each week's or month's arrivals are followed separately, so a change in onboarding or in the mix of users shows up in the cohorts after the change rather than being averaged away.
  • By source. Users from different channels, networks and campaigns activate at very different rates. Cost per activated user, which is spend divided by activated users, can rank channels in the opposite order to cost per install.
  • By window, once closed. A cohort that started yesterday has had no time to activate. Rates are compared only for cohorts whose window has fully passed, or the newest cohorts always look worst.

The rate is also read against what follows it. Activation earns its place only if activated users go on to retain and, where there is a paid step, to pay. A rate that rises while retention of activated users falls usually means the event has become easier to reach without becoming more meaningful.

Illustrative example — not a real result

A habit-tracking app compares activation by ad network

Plinthgrove Habit Tracker defines activation as logging any habit on three separate days within seven days of install. In one month, Network A delivers 20,000 installs at £1.20 each and Network B delivers 8,000 at £2.00. After every cohort's seven-day window has closed, 18% of Network A's installs have activated against 41% of Network B's. That gives 3,600 activated users from A at £6.67 each and 3,280 from B at £4.88 each. Network A had looked like the better buy on cost per install; on cost per activated user it is the dearer one. The team moves part of A's budget to B and asks A for placement reports on the cohorts that activated least.

What moves it

  • Onboarding. The first screens decide how many people reach the event. Removing steps that stand between a new user and the first useful action usually matters more than adding explanation, and behaviour-triggered messages can help, as the onboarding email cadence test sets out to check by reading activation rather than opens.
  • Permission requests. Asking for notifications, location or tracking before the user has seen any value can end the first session early; asking at the moment the permission is useful costs less.
  • Acquisition source and promise. Users who arrive from an ad that described the product accurately are more likely to find what they expected. The guide to app install campaigns moves campaign optimisation from installs to the activation event once enough data exists, which changes who the networks find.
  • The event and window chosen. Redefining the event or lengthening the window moves the rate without changing anything users do. Any change to the definition starts a new series.
  • Tracking. An event that does not ship in the release cannot be counted. The guide to launching a new app specifies the activation event before launch for this reason and reads the first month by cohort and source.

What good looks like

There is no reliable benchmark, and the site does not cite one. Activation rates depend on the event chosen, the window, the category and how much effort the first useful action takes, so a rate that would be poor for one app is strong for another, and two apps quoting the same number rarely mean the same thing by it. The useful comparisons are internal: each new cohort against earlier cohorts with the same definition and window, each source against the others, and the cost per activated user against what an activated user is worth to the business once retention and revenue are taken into account.

Activation is read by Cohort, the group of users who started together, because blended figures hide whether recent arrivals do better or worse. Churn Rate measures the loss that follows; users who never activated tend to make up much of the early churn, so improvements in activation often show up as lower churn a cycle or two later. CPI is the acquisition price that activation rate corrects: cost per install divided by activation rate gives the cost per activated user. In products with a free trial, Trial-to-Paid Conversion Rate is the paid step that comes after activation, and the two are read together because a trial user who never activates rarely pays.

Common mistakes

  • Choosing an event because it is easy to count, such as sign-up complete, which nearly everyone reaches and which separates nothing.
  • Comparing rates with different windows, or comparing a cohort whose window has not closed with one whose window has.
  • Using different definitions in analytics, the measurement partner and the networks, so campaigns optimise towards one behaviour while the report counts another.
  • Reading a blended rate instead of rates by cohort and source, which hides a channel that brings cheap installs that never activate.
  • Treating the event as causal. The analysis that picks it shows association; onboarding experiments test whether bringing more users to it actually improves retention.
  • Leaving the definition unchanged as the product changes, so the event stops marking first value.

More terms to look up

Where this term is used

Related