What are uninstalls?

How they are measured

Indirectly, in most cases. An app cannot observe its own removal, so uninstalls are inferred — typically from silent push notifications that fail to deliver, which indicates the app is no longer present.

That inference is imperfect. A user who disabled notifications, or whose device has been offline for a long period, can appear uninstalled. Rates should therefore be read as directional rather than exact, and compared against your own history rather than a benchmark.

Timing is the informative part

The rate matters less than when the uninstalls happen.

Within minutes points at a mismatch between the ad and the product — the creative promised something the app does not do, which is the classic high-IPM, low-retention pattern.

Within the first day usually points at onboarding: friction, a forced signup, or an unclear first experience.

After a week or two points at value: the product worked and did not earn a place.

Spikes after a release point at a specific regression, and are the most actionable signal the metric produces.

Using it properly

• Segment by acquisition source. Uninstall rate by source is one of the fastest quality signals available.
• Watch for release-correlated spikes, which is where the metric earns its place operationally.
• Do not chase the rate itself. Uninstalls are a symptom; retention and activation are where the causes live.
• Never message an uninstalled user through a channel that still reaches them — it is the most reliable way to convert indifference into hostility.