Predict Success in Production

The macro goal the ten Test Desiderata properties serve: when the suite is green, shipping is safe; when it’s red, it isn’t. Bache elevated Beck’s Predictive property to this status, and the field agrees: Fowler’s “when this suite goes green, we can release” is the same commitment.

Inspiring confidence is this goal’s outcome, not a property: every proposed “inspiring” metric turns out to proxy something else: deployment frequency proxies speed and determinism, rollback rate proxies prediction. There is no Inspiring KPI; a team that wants confidence tunes the constituents.

The honest limit (Testing vs. Checking): automated checks predict only what was specified. An 80% mutation score can still ship the UX bug no test described. Green means “nothing we thought to check broke”; exploration covers the rest.

Measuring: change-failure rate (DORA elite: 0–15%), escaped defects per release, and mutation score on critical paths.