What is an install?

Why first open, not download

The measurable event is the first launch, not the download, and the distinction is practical rather than pedantic.

A download that is never opened cannot report anything — the measurement SDK has not run. It also has no value to anybody: the user acquired nothing and the business acquired nobody.

The gap between downloads and first opens is real, particularly on preloaded and incentivised traffic, which is why OEM install counts can look strong while activated user counts do not.

Why installs are a weak success metric

An install is the beginning of a relationship, not evidence of one. It says nothing about whether the user found value, returned, or paid.

The pattern is consistent across every acquisition channel: sources that produce the cheapest installs produce the least valuable users, and install-stage metrics cannot distinguish them. A source at half the CPI with a third of the retention is a worse buy, and only downstream measurement reveals it.

The useful working metric is cost per activated user — someone who reached the point the product becomes useful — with installs as the intermediate step rather than the goal.

What measures installs now

• The [install referrer](/resources/glossary/install-referrer) on Android, which carries the campaign that drove it.
• [SKAdNetwork](/resources/glossary/skadnetwork-skan) and [AdAttributionKit](/resources/glossary/adattributionkit) on iOS, aggregated and delayed.
• [MMP](/resources/glossary/mobile-measurement-partner-mmp) SDKs, which de-duplicate across sources.
• [Reinstalls](/resources/glossary/reinstall) and [reattributions](/resources/glossary/reattribution), which must be separated from genuinely new installs or acquisition counts inflate.