Completion rates that hide overnight retries
A programme can look successful while the same work quietly retries until morning — unless application analytics separate first-pass from eventual completion.
Completion percentages feel definitive. They are also easy to inflate when retries are counted as fresh success. If your application analytics only mark a job done when any attempt finishes, sponsors never see how often the first pass failed.
Start by asking whether “completed” means first-pass or eventual. Then check whether retry identity survives in the event stream. Gaps between programme claims and operator experience usually show up as overtime support without matching failure alerts.
A practical fix is to emit both first-pass and eventual completion indicators, and review them in the same weekly programme pack.