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

Mobile & App Marketing › Troubleshooting

App Installs Not Turning Into Active Users

Why app installs do not become active users and how to fix it: campaigns bought on installs, fraud, a mismatched listing, a slow first session and lost links.
Troubleshootingintermediate9 min readUpdated

First, confirm activation is measured

Start with the event, not the campaigns. The tutorial on finding the app's activation event asks for one written definition (the event, its qualifying parameters, the window after install, the platforms) held in the tracking plan with an owner, and warns that a redesign, a new feature or a new audience can change which early action matters. Check three things: that the event still fires in the current app version on both platforms, that the analytics, the measurement partner and every network count the same action, and that nobody has quietly moved the window. A renamed screen or a refactored flow can stop an event firing without anyone noticing, and the fall then looks like a behaviour change.

Next, read activation as the share of each install Cohort that reached the event within the agreed window, split by source and campaign. Keep iOS and Android apart. The app attribution guide explains that for iOS users who decline tracking, Apple's SKAdNetwork sends delayed, aggregated postbacks in which post-install behaviour is compressed into a conversion value, so paid iOS activation by campaign arrives late and coarse; read it over weeks, and use in-app analytics for the behaviour itself. Firebase's documentation, cited by the deep links and app analytics guide, offers cohort and retention views for exactly this reading.

Stay inside the symptom. Fewer store visitors installing is a separate problem with its own page in the app marketing troubleshooting section, as are installs that cost more each month. This page starts once installs have happened.

Diagnosis table

Diagnosis

SymptomCheckLikely causeFix
Installs rise with a new campaign while activated users stay flatActivation by source and campaign, and each campaign's optimisation goalCampaigns optimised to installs on networks that find people who never open the appMove the goal to the activation event once volume allows, and exclude placements whose cohorts never open
One sub-publisher's installs show no sessions at allClick-to-install times, device models, install hours and activation for that source in the measurement partnerInstall fraud inflating installs that never produce a real sessionReject invalid installs through the partner's rules, exclude the source and withhold payment
People open the app once and leave from the first screenFirst-session drop by campaign theme, against what the ad and listing showedAn ad and store listing promising something the first session does not deliverMatch ad, listing and first screen to one promise, with a listing per campaign theme
Most new users stop at a sign-up form, tour or promptScreen-by-screen first-session funnel from first open to the activation eventAn account form, a tour and permission prompts standing before first valueDefer the account, the tour and permissions until the moment each is needed
Users from shared or ad links activate no better than organic installsWhether link-driven first opens land on the linked item or a generic startNew installs from a link landing on generic onboarding because no deferred deep link existsCarry the destination through the install with a deferred deep link and a designed default
Activation dropped on the day of a release, everywhere at onceEvent counts per app version, and the written definition against what is reportedAn activation event chosen by opinion, or no longer firing, so activation is misreadRestore the event, rerun the selection analysis and write one definition everywhere

Fix, verify and prevent each cause

Campaigns bought on installs alone

The app install campaigns guide describes install optimisation plainly: the network finds people likely to install, which it does cheaply, and they include many who open once and leave. A cheap CPI is therefore no evidence of a good source.

Fix: once the activation event has accumulated enough volume, move the campaign's goal to it, as the guide's sequence recommends, and exclude placements whose cohorts show no activation.

Verify: activation within the window rises for that campaign's cohorts, and cost per activated user falls even though each install costs more.

Prevent: report every campaign on cost per activated user by week rather than on install cost, so a source that buys empty installs is visible before it is scaled.

Installs with nobody behind them

The same guide lists the forms install fraud takes, from click spamming and click injection to device farms and spoofed install signals, and the signature they leave: implausible click-to-install times, odd hours and device models, and sub-publisher cohorts with no activation and no retention.

Fix: measure through the measurement partner rather than the network's own count, switch on its rules for rejecting installs that fail its checks, exclude the offending sub-publishers and enforce the quality clause in any agreement with a smaller network.

Verify: the rejected share appears in the partner's reports, the source disappears from billing, and blended activation recovers without any product change.

Prevent: write what counts as an invalid install into every agreement before spend starts, and review new sources on activation in their first weeks rather than at the end of the month.

A promise the first screens do not keep

An ad built around one feature, landing on a listing that leads with another and an app that opens on a third, loses people at every step. The install campaigns guide warns that what the ad promises the listing and the first session have to deliver, and the custom store listings tutorial gives each campaign theme its own listing and asks the team to follow installs through to activation, because a page can win installs by promising more than the first session delivers.

Fix: align each theme's ad, listing screenshots and first screen around a single promise, and drop themes the app cannot honour in its first minutes.

Verify: first-session drop by theme narrows, and activation for each theme's cohorts approaches the app's organic installs.

Prevent: add the first screen to the creative brief, so nobody signs off an ad without seeing where the install will land.

A wall before the first useful moment

The guide to app onboarding and permission requests treats the first session as a funnel from install to the activation moment, where every account form, tour screen and permission prompt is a cost the person pays on the way. Its table of usually deferred items, from account creation to the notification permission, is the checklist for this cause.

Fix: map every screen between first open and the activation event, then defer the account until the person has something to save, replace the tour with tips where each feature lives, and move each permission to the moment its feature is first used.

Verify: run the change as a randomised test and read activation, permission grants and early retention for the cohorts that received it, side by side and platform by platform.

Prevent: keep the first-session screen map in the release checklist, so any new screen before first value needs an explicit decision.

A person who tapped an ad or a friend's link for a specific item, then installed, should open the app on that item. Without a deferred Deep Link, the store does not pass the destination on, and the person meets generic onboarding with no memory of why they came. The deep links guide explains how a link service records the destination and returns it on first open, and why a sensible default is needed where matching is limited by platform rules.

Fix: add deferred deep links for every campaign and sharing link, shorten onboarding for people whose context is known, and design the default first run for the rest.

Verify: among link-driven installs, the share whose first screen is the linked item rises, and their activation pulls ahead of comparable installs without a link; the tutorial on testing app deep links gives the fresh-device tests.

Prevent: pair every new campaign listing with a tested deferred link before it spends.

An activation number that cannot be trusted

When the event was picked in a meeting, it may not separate users who stay from users who leave; when it stopped firing after a release, the fall is an artefact. The activation tutorial chooses the event from the app's own data by comparing later retention for new users who did and did not take each candidate action, checks timing and volume against what networks and iOS postbacks need, and asks for one definition in the tracking plan, the partner, the networks and the reports.

Fix: restore the event if it broke, rerun the comparison if it was chosen by opinion, and change the definition everywhere at once with the date recorded.

Verify: event counts per app version are stable, every system reports the same activation for the same cohort, and the trend shows a marked break only on the recorded date.

Prevent: give the event an owner in the tracking plan and add an automated check that it still fires on each new build.

A prevention routine for activation

Treat activation as the paid team's result and the product team's target at the same time, so neither can improve their own number at the other's expense. Each week, read activation and cost per activated user by source, campaign and install cohort, iOS and Android apart, with in-app analytics as the behaviour record and the partner's postbacks as the campaign record. Flag any source whose cohorts activate well below organic installs before its budget grows.

Before each release, confirm the activation event fires on the new build and walk the first-session screen map. Before each campaign, check that the ad, the listing and the first screen make the same promise, that a deferred link routes link installs to the item, and that the campaign's goal is the agreed event wherever volume allows. Before each new network, agree the invalid-install terms. Every time the product changes how new users start, rerun the activation analysis rather than assuming the old event still holds, and record any change in definition so trends stay readable.

The app install campaign launch checklist holds the campaign half of this routine, and the mobile measurement hub explains why install counts from the network, the store and the partner never agree exactly, which is a reconciliation question rather than an activation fault.

Illustrative example — not a real result

Mumblecrisp Budgets traces a fall in bank-linking

Mumblecrisp Budgets, an invented budgeting app, counts activation as linking a bank account within seven days of install. Over six weeks installs rise from 14,000 to 23,000 a month after a new in-app network campaign launches, but activated users fall from 4,200 to 3,900. The event still fires on both platforms. By source, organic and search cohorts activate at 31% and 28%; the new network's cohorts activate at 4%, and two of its sub-publishers produce 2,600 installs with a median click-to-install time under ten seconds and no sessions after the first open. The partner's rules reject those installs, the sub-publishers are excluded, and the campaign moves its goal from installs to bank linking. The funnel shows a second problem: since a release the month before, every new user meets an account form, a four-screen tour and the notification prompt before the bank-linking screen, and 46% leave there. Mumblecrisp tests a flow that opens on a spending preview, asks for the account at bank linking and drops the tour. In the test cohorts, activation reaches 36% against 27% for the old flow, and cost per activated user on the network falls from £38 to £19.

Common mistakes

Judging sources on install cost when the business is buying active users. Pooling iOS and Android activation into one rate. Treating a release that broke the event as a change in behaviour. Fixing onboarding while a fraudulent source is still inflating the denominator. Changing the activation definition without recording the date.

Frequently asked questions

What share of installs should become active users?

There is no borrowed figure worth using, because activation depends on which action the app counts, the window it allows and where its installs come from. Compare the app with itself instead: activation by install cohort and by source, iOS and Android apart, against the weeks before the fall. A source whose cohorts activate far below the app's organic installs is the finding, whatever the absolute level.

Should we stop buying installs until activation recovers?

Not across the board. Read activation and cost per activated user by source first, because the fall usually concentrates in one network, sub-publisher or campaign. Pause or narrow the sources whose installs never open the app, keep the ones whose cohorts still activate, and fix any first-session wall before scaling again, since a slow first session wastes good installs as well as bad ones.

Why do our iOS activation numbers look worse than Android's?

Often because they are measured differently, not because iOS users behave worse. The attribution guide explains that where people decline tracking, Apple's SKAdNetwork reports installs and post-install signals in delayed, aggregated form through a conversion value, so paid iOS activation by campaign arrives late, coarser and sometimes not at all for small campaigns. Read in-app analytics for the behaviour, read the postbacks over weeks rather than days, and never pool the two platforms into one rate.

Terms used on this page

Where this page is used

Also related

Sources

  1. SKAdNetwork — Apple Developer
  2. Firebase documentation — Google

Related