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

Mobile & App Marketing › Mobile Measurement

Finding Your App's Activation Event

How to choose an app's activation event from its own data: candidate actions, retention by cohort, timing, volume for campaigns, and one shared definition.
Tutorialadvanced11 min readUpdated

What an activation event is for

An activation event does several jobs at once, which is why it has to be one event and not one per team.

UseWhat the event does thereCovered in
North star and OKRsThe value moment the north star or its inputs countNorth star and OKRs for app installs
KPI treeThe leaf that turns installs into activated usersKPI tree for app installs
Campaign goalThe optimisation event once installs have gathered enough dataApp install campaigns
iOS conversion valueA signal the conversion value can map to within its windowsApp attribution
Onboarding targetThe action onboarding is designed to bring new users toThe app's own onboarding work
Retention reportingA cohort start or a definition of "retained"Cohort analysis

In the AARRR model, activation is the stage between acquisition and retention, and the event is what makes that stage countable. If marketing, product and the networks each use a different action, the install campaigns optimise towards one behaviour, onboarding pushes another and the board report counts a third. The app attribution guide's advice on postbacks makes the same point from the other side: the event a campaign optimises to must be defined identically in the app, the partner and the network.

The analysis below is correlational, and the page says so plainly. Users who set up a profile and stay may have stayed anyway, because the same commitment produced both. The output is the best available working definition, to be confirmed by experiments that help more users reach the event and check whether retention follows.

Before you start

You need event-level data for new users, with the app version and install source, from the analytics SDK. The deep links guide describes how Google Analytics collects app data through the Firebase SDK into app data streams, with automatically collected, recommended and custom events and event parameters that describe them; Google Analytics Help and the Firebase documentation are the references for the setup. You need a tool that can compare groups of users and cohorts: Explorations in Google Analytics, which the explorations guide describes for segment comparisons and cohort retention, or a warehouse export for analysis the built-in reports cannot do.

Check the volume. The analysis compares retention between groups of new users, and small cohorts give unstable answers; the guide to reading small numbers explains how to show ranges rather than single figures and when to pool periods. If one month of new users is too few, pool several months from a period when the product did not change much.

Agree who will sign off the definition: usually the heads of product, growth and analytics together, because each will use it.

The steps

  1. Write down the question the event must answer

    State it in one sentence: which early action best indicates that a new user will still be using the app later? Then write down what "later" means for this app and what "using" means. A daily habit app and a monthly bill-paying app need different answers. Note who will use the event and for what (the table above), because those uses add constraints later.

  2. List candidate actions from the event data

    From the app's event list, pick the actions a new user can take early that plausibly reflect the core value: finishing a first lesson, adding a first item, connecting an account, inviting someone, completing a first transaction. Include the one everyone assumes (often finishing sign-up) so it can be tested rather than assumed. Leave out automatically collected events such as first opens and screen views, which every user produces. Aim for a short list that a reviewer can discuss, each with its exact event name and any parameter that defines it.

  3. Choose an early window and a later retention measure

    Pick an early window in which the action has to happen to count, short enough to be "early" for this app's rhythm. Pick a retention measure later on: active in a chosen later week, or performing a core-value event then, as the deep links guide suggests defining retention precisely. Set both from the product's natural rhythm, write them down before looking at results, and keep them fixed for every candidate so the comparison is fair. The reverse of the measure, users who are gone by then, is the early form of Churn Rate for new users.

  4. Compare retention for users who did and did not take each action

    For each candidate, split one or more install cohorts into users who took the action within the early window and users who did not, and read the retention measure for both groups. In Explorations this is a segment comparison or a cohort exploration with the action as the condition; in a warehouse it is two counts and two rates per candidate. The candidates worth keeping show a wide gap between the groups. Record the counts behind every rate, not only the rates.

  5. Check how soon and how often each happens

    A strong separator that only a few users reach, or that most users reach late, is a poor activation event. For each candidate on the shortlist, read what share of new users take it within the window and how long after install they typically do. The best events combine a clear retention gap with a meaningful share of new users and an early typical time. Look also at whether the action can happen more than once; the activation event should be counted as the first occurrence.

  6. Check volume against what campaigns need

    The install campaigns guide explains that networks learn from a steady flow of the optimisation event, and that the usual sequence is installs first, then a shallow event most useful users perform early, then a deeper one only if it occurs often enough. Estimate how many times a day or week the candidate would fire for paid users at the planned budget. If it would be too rare for the networks to learn, it can still be the activation event for reporting, but campaigns will need a shallower stepping-stone event for now; write that down rather than quietly using two definitions.

  7. Check timing against what iOS postbacks can carry

    On iOS, for users who have not allowed tracking, the app attribution guide describes post-install behaviour reaching networks only as a conversion value that the app updates within limited windows after install, with postbacks that are delayed and coarser for small campaigns, as Apple's SKAdNetwork documentation describes. An event that typically happens late may fall outside what the conversion value can report. Check the current SKAdNetwork documentation and your measurement partner's configuration for the windows that apply, and prefer a candidate that most activating users reach early enough to be captured.

  8. Pick one event and write its definition

    Choose the candidate that best balances the retention gap, reach, timing and volume, and write its definition in full: the event name, any parameter values that qualify it, the window after install in which it counts, whether only the first occurrence counts, and the platforms it applies to. Note the evidence (the comparison, the counts, the period) and the limitation that it shows association, not cause. Where the analytics property marks important events, mark it as a Key Event so it appears as a measure of success in reports.

  9. Put it in the tracking plan, campaigns and reports

    Add the definition to the tracking plan with its owner, so that releases do not rename or move it. Send it to the measurement partner and, through the partner, to the networks as the event campaigns will optimise to when volume allows. Map it into the iOS conversion value. Make it the activation leaf of the KPI tree for app installs and an input to the north star and key results the north star metric and OKRs framework sets. Give onboarding the target of bringing more new users to it within the window, and run that as an experiment so the causal question gets answered. The deferred account creation test is one such design, reading the share of each install cohort that reaches the event when the account form moves later.

  10. Revisit it when the product changes

    A redesigned onboarding, a new core feature or a new audience can change which early action matters. Rerun the comparison when the product changes in a way that affects the first sessions, and on a review date the team sets for itself. If a new event wins, change it everywhere at once, record the change and its date, and expect a break in every trend that uses it.

A worksheet for the comparison step keeps every candidate on the same terms.

Candidate actionUsers who took it in the windowRetained (took it)Users who did notRetained (did not)Typical time to actionNotes
Each action, with event nameCountCount and rateCountCount and rateFrom installVolume, iOS timing, repeatability

A savings app finds its activation event (illustrative)

Illustrative example — not a real result

Clinkerbee Savings compares five early actions

Clinkerbee Savings, an invented savings app, has used "sign-up complete" as its activation event since launch, and its campaigns optimise to it. Pooling three months of new users (about 48,000 installs), the analyst compares five actions within the first three days after install against being active in week five. Users who completed sign-up retain at 21% against 4% for those who did not, but 83% of installs complete sign-up, so it says little about who will stay. Connecting a bank account (41% of installs) shows 30% against 9%. Setting up a first savings goal (34% of installs, typically on day one) shows 44% against 8%. Turning on round-ups (12%) shows 47% against 15% but usually happens after day five, which is late for the iOS windows the partner has configured, and is too rare for campaigns. Inviting a friend (3%) is far too rare. The team defines activation as "first savings goal created within three days of install", maps it into the iOS conversion value, moves campaign optimisation from sign-up to the new event and starts an onboarding test that prompts goal creation on the first screen, to see whether helping people set a goal changes their week-five activity or merely identifies the keen ones.

An activation rate is only as reliable as the install count beneath it, and the tutorial on reconciling app install numbers lines up store, analytics, partner and network installs each month. The measurement hub shows where both routines sit beside attribution and deep links.

When installs keep arriving but the share reaching the event falls, the diagnosis of app installs not turning into active users separates a broken or mis-chosen event from campaign, fraud, listing and first-session causes.

Verify it worked

  • Open the worksheet. Each candidate should show counts and retention for users who took the action and users who did not, with its typical time to action.
  • Look at the event everyone assumed, often finishing sign-up. If most installs reach it, expect a weak separator, as Clinkerbee found.
  • Read the written definition. It should give the event name, qualifying parameters, the window, the first-occurrence rule and the platforms.
  • Compare the tracking plan, the measurement partner, the campaign settings, the iOS conversion value and the KPI tree. All should count the same event in the same way.
  • Check that an onboarding test exists. It should aim to bring more new users to the event, so the causal question gets answered.

If it goes wrong

  • Every candidate shows a similar retention gap. The likely cause is an early window or retention measure that does not fit the app's rhythm. Write down a new window and measure before reading results, and keep them fixed.
  • The rates change sharply between runs. The likely cause is cohorts too small to be stable. Show the counts behind every rate, show ranges and pool months from a period when the product changed little.
  • The best separator is rare or arrives late. The likely cause is that only the most committed users reach it. It can still serve reporting. Give campaigns a shallower stepping-stone event for now and write down both.
  • Tools disagree on activation counts. The likely cause is an event not defined identically in the app, the partner and the network. Compare names, parameters and windows against the tracking plan, and check that no release renamed or moved the event.
  • Retention does not move in the onboarding test. The likely cause is that the action was a symptom of commitment, not a cause. Return to the shortlist and test the next candidate.

Next steps

Common mistakes

  • Choosing the event by opinion. Finishing sign-up is everyone's first guess and often the weakest separator. Compare candidates on data.
  • Reading association as cause. Keen users do everything. Confirm the event with onboarding experiments.
  • Picking a strong separator that few users reach. An event only the most committed reach cannot drive campaigns or onboarding.
  • Ignoring timing. A late event starves networks and may fall outside what iOS conversion values can report.
  • Comparing candidates on different windows. Fix the early window and the retention measure before reading results.
  • Small cohorts read as precise. Show counts and ranges and pool periods where needed.
  • One event per team. The partner, the networks, product and finance must all count the same action.
  • Never revisiting it. Product changes move the value moment; rerun the analysis when the first sessions change.

Frequently asked questions

What is an app activation event?

The action that marks a new user's first experience of the app's core value, defined precisely enough that every tool counts it the same way. The deep links and app analytics guide describes it as one event per app, agreed across teams and sent to the measurement partner and the networks. In practice it is chosen as the earliest action that clearly separates users who stay from users who leave, as the north star page for app installs puts it, while still happening soon and often enough to be used by campaigns.

Does finding the activation event prove that it causes retention?

No. The analysis compares users who took an action with users who did not, and the two groups differ in more than the action: people who were already keen are more likely to do almost everything. The chosen event is a working definition. Confirm it by changing onboarding to help more people reach it and checking, in a controlled test, whether retention moves for the group that was helped. If it does not, the action was a symptom of commitment rather than a cause of it.

Why does the activation event need to happen early?

For two practical reasons as well as an analytical one. Ad networks learn from a steady flow of the optimisation event, so an event that happens rarely or late starves campaigns, as the install campaigns guide explains. On iOS, the attribution guide describes post-install behaviour reaching networks only through a conversion value set within limited windows after install, so an event that tends to happen late may never reach them. And an early event gives onboarding a target it can influence while the new user is still paying attention.

Terms used on this page

Frameworks

Where this page is used

Also related

Sources

  1. Firebase documentation — Google
  2. Google Analytics Help — Google Analytics Help
  3. [GA4] Get started with Explorations — Google Analytics Help
  4. SKAdNetwork — Apple Developer

Related