Transcript
A release is not automatically a full retest
Weekly SaaS releases can make blanket regression impractical and low value. Start with authoritative release information, known defects, supplier test evidence, and the customer configuration inventory. Determine whether the changed component, shared service, interface, role, or data behavior enters the approved GxP boundary.
Consider consequence and detectability
For exposed functions, trace the credible failure and potential impact. Consider configuration, open records, integrations, browser or device variants, detectability, rollback, and production monitoring. Supplier regression may support standard behavior, while local routing, identity, exports, and interfaces may still need customer evidence.
No impact, targeted checks, or regression
Choose no impact when exposure is credibly absent, targeted confirmation when a bounded risk remains, or broader regression for material or uncertain change. Record the evidence, limitations, decision, monitoring period, and escalation criteria. Continuous delivery requires continuous assessment, not automatic repetition of every historical test.