Confirming nothing broke.
WHY IT GROWS
Every release adds behaviour that must keep working.
WHAT HAPPENS WITHOUT MANAGEMENT
The suite becomes too slow to run, and is run less often.
WHAT TO AUTOMATE
Everything run every release.
WHAT TO SELECT FOR A SMALLER RUN
Tests covering areas that changed Tests covering areas depending on what changed Tests covering high-risk areas A core set always run
HOW TO KNOW WHAT DEPENDS ON WHAT
Coverage data mapping tests to code.
WHAT TEST IMPACT ANALYSIS PROVIDES
Running only tests affected by a change.
WHAT IT REQUIRES
Reliable mapping, and trust in it.
WHAT TO KEEP RUNNING FULLY
The complete suite, on a schedule, regardless.
WHY
Selection can be wrong, and a full run catches what it missed.
WHAT TO REVIEW PERIODICALLY
Tests that have never failed Tests that fail frequently without finding defects Tests duplicating others
WHAT TO DO WITH EACH
- Never failed: consider whether it verifies anything
- Frequently failing: fix or remove
- Duplicates: consolidate
WHAT TO MEASURE
Suite duration over time, which grows silently.
