Crash-free sessions

A session is one period of app use; a crash is an unhandled error that terminates it. Crash-free sessions divides the sessions that did not crash by all sessions over a window, usually per app version and per platform. Its companion, crash-free users, counts the share of users who experienced no crash, which weights a single bad crash affecting many people more heavily. Both come from crash reporting tools such as Firebase Crashlytics or Sentry.

The metric is useful because it is leading. A drop in crash-free sessions on a new release shows up hours before the app store reviews, uninstall spikes and support tickets that follow. Teams set a release threshold (for example, no rollout beyond a small percentage until crash-free sessions match the previous version), watch the metric during staged rollouts, and treat the top crash groups as the first items in the next sprint.

It does not capture hangs, freezes, slow screens or errors that are caught and shown as failures, so it belongs alongside ANR rate, cold start time and error rates. A high crash-free number on a small, engaged user base can also hide device- or OS-specific problems that only appear at scale.

• Product manager: app store ratings and retention track stability closely; this is the metric to watch on every release.
• CTO / Head of Engineering: crash groups with stack traces turn a vague quality complaint into a ranked engineering backlog.
• Support manager: a stability regression shows up here before the tickets arrive, giving time to prepare or halt the rollout.
• Operations head: for field apps, a crash mid-task means lost data and a driver or technician on the phone; stability is operational uptime.

The metric exists because mobile apps run on thousands of device and OS combinations the team never tested, and because users punish instability immediately and publicly. Crash-free sessions gives a single comparable number per release that predicts reviews and retention, and links directly to the crash reports needed to fix problems. The trade-off is that it measures only one failure type, so teams that optimise it alone can ship apps that do not crash but freeze or fail quietly. It is necessary, not sufficient.

• A consumer FinTech app gating rollout beyond ten percent of users on crash-free sessions matching the previous release.
• A logistics driver app tracking crash-free sessions per device model to find low-end Android problems.
• A patient app for a hospital network monitoring stability after each OS update.
• A learning app correlating crash groups with specific video and quiz components.
• A retail app watching crash-free sessions in real time during a peak-season sale.

MetricA number tracked through crash reporting tools and reviewed on every release, with thresholds tied to rollout decisions. Eazyware instruments and monitors it for the apps it builds under the React Native mobile service and the maintenance and support Care Plans.

Eazyware service that covers it: React Native Mobile Application Development. Starting prices are on the pricing page.