Why context changes the answer
“How much regression testing is enough?” does not have one universal answer. The appropriate scope depends on what changed, what could break, how critical the system is, and what evidence already exists.
The five-part question model
| Input | What to provide |
|---|---|
| Context | Product, system, project, users, delivery model |
| Change | What is different in this release or situation |
| Risk | What would matter if it failed |
| Constraint | Time, people, data, environments, tools, governance |
| Decision | What you need to decide or recommend |
Weak vs useful question
Weak
How much regression should we do?
Better
We have a web application with about 400 regression cases. The release changes authentication and three shared APIs. We have two days before production. How should we prioritize regression and communicate residual risk?
Include what has already been tried
Tell AskAQA whether automation has run, which defects exist, which areas are blocked, or which stakeholders have already made decisions. This prevents repeating work you have already completed.
Avoid hiding the real decision
Instead of asking: “Is 70% coverage enough?”
Ask: “Can QA recommend release when 70% of regression is complete but all P1 flows passed and the missing 30% is low-risk?”
Ask: “Can QA recommend release when 70% of regression is complete but all P1 flows passed and the missing 30% is low-risk?”