Desktop analytics inherited a comfortable fiction: a session is a stretch of attention with a timeout. Mobile App Analytics copied the timeout and then wondered why engagement looked both frantic and empty.
People background an app to check a message, to use a camera, to authenticate with a bank. The OS wakes your process for a push. If your session definition treats each wake as a new visit, you will report a loyal audience of lock screens.
Time in foreground is not a job
Duration metrics are catnip in a leadership pack. They go up when a screen is confusing and when a screen is beloved. Mobile Funnel Practice asks students to retire “average session length” from any argument about quality unless they can pair it with a finished job. Minutes spent on a map can mean lost; minutes spent on a lesson can mean stuck.
Notification loops disguise themselves as love
A well-timed push can look like retention. Count returns that begin from a notification separately from returns that begin from the home screen. If the first set dwarfs the second, you do not have a habit. You have a tap. That is allowed — many products live on reminders — but it should be named, or someone will “improve engagement” by sending more mail.
Two platforms, two clocks
iOS and Android disagree about background life. If you merge them in one session chart, you are averaging climates. Keep platform as a load-bearing property, and do not celebrate a lift that exists on only one OS unless you understand the lifecycle difference. This is tedious. It is also why dual-platform pairs belong in the same sitting rather than in two Slack threads.
The programme that drills this is Mobile Funnel Practice. Bring both clients, or admit you only measure one.