Why does App Store Connect show fewer sessions than my own analytics?

Apple's engagement metrics (sessions, active devices, retention, crashes, deletions) come only from devices whose users opted in to share analytics with developers. Your in-app SDK sees every launch it's installed for. Neither is wrong; Apple's device-side numbers are a large sample, not a census, and should be read as trends rather than totals.

Which App Store Connect metrics are complete counts?

Store-side activity is complete: impressions, product page views, downloads and redownloads, and the money: units and proceeds in Sales and Trends. These are recorded by the store itself, no opt-in involved. The opt-in line starts where measurement moves onto the user's device: installations, sessions, retention, crashes and deletions.

Can I still use deletions and retention data if it's opt-in only?

Yes, as trends. Opt-in rates are fairly stable over weeks for a given audience, so changes in the deletions-to-installs ratio or a retention curve are real signal even though absolute counts undercount. Compare a metric with itself over time, keep ratios within one data family, and opt-in metrics become some of the most useful numbers you have.

FunnelHound / Guides / The opt-in data gap

The opt-in data gap: which App Store Connect numbers are complete (and which aren't)

By Emin, indie iOS developer building FunnelHound · Published 3 Oct 2026

Half your Apple dashboard is a census, half is a survey, and nothing on the screen tells you which is which. Store-side counts (impressions, page views, downloads, proceeds) are complete. Device-side metrics (sessions, retention, crashes, deletions) come only from users who opted in. Every confused ratio in indie dev crosses that line somewhere.

Where the line runs

Census (complete): impressions, product page views, downloads and redownloads, App Units, proceeds. Recorded by the store itself: the user's privacy choice can't hide a store page view from the store.
Survey (opt-in devices only): installations, sessions, active devices, retention, crashes, deletions. Measured on the device, shared only if the user agreed when setting up their iPhone.

The store funnel (the census side) is the ground your metric comparisons stand on. The survey side is where the post-install story lives. Both are essential; they're just different instruments, and the dashboard shelves them side by side without a warning label.

The classic misreadings

"Only a third of my downloads ever open the app!": computed as sessions (survey) over downloads (census). The real open rate is higher; the denominator counted everyone, the numerator only volunteers.
"Deletions look tiny, churn is fine": deletions undercount in absolute terms for the same reason. The ratio's trend is the signal, not the count.
"Apple and Firebase disagree, one must be broken": three instruments, three definitions, three denominators. Disagreement is the healthy baseline.

Reading survey metrics like a grown-up

Opt-in rates for a given audience are reasonably stable across weeks, which rescues everything: a survey metric compared with itself over time is real signal. Deletions-to-installs creeping up after a release, a retention curve bending after an onboarding change, crash counts stepping down after a fix: all trustworthy trends, even with undercounted absolutes. The two rules: trends over totals on the survey side, and never build a ratio across the census/survey line: Apple's own peer percentiles help here, since your competitors' data has the same gap yours does. The gap doesn't make the data bad; it makes the data typed, and types just want respecting.

Ratios that respect the line, by construction

FunnelHound builds its funnel from matching data families (census with census, trends where trends belong) from App Store Connect.

Get FunnelHound

Share this on X

Data notes: metric sourcing per Apple's App Store Connect documentation as of 2026; Apple has adjusted which metrics include modeled or expanded data before: re-check definitions after analytics redesigns. Opt-in rates vary by audience and iOS version; stability over weeks is an observation, not a guarantee.