Knowing how it behaves in the field.
WHAT TO MONITOR
Crash-free session rate Application startup time Network error rates Key task completion rates Adoption of each released version
WHY THAT LAST ONE
It tells you how long an old version persists, which determines how long you must support it.
WHAT TO ALERT ON
A rise in crashes after a release A sharp fall in a key task completing Backend error rates affecting the application
WHY BUSINESS OUTCOMES SPECIFICALLY
Every technical measure can look healthy while a broken flow has stopped all sign-ups.
WHAT TO SEGMENT BY
Platform Application version Operating system version Device model Country
WHY
Problems frequently affect one segment only, and aggregate figures conceal them.
WHAT TO WATCH AFTER EVERY RELEASE
The first hours, closely.
WHAT TO PREPARE IN ADVANCE
The ability to halt a rollout The ability to disable a feature remotely
WHY IN ADVANCE
Both are useless if built after the problem appears.
