Regression Suite Pruning (Retention)

Regression Suite Pruning (Retention)

1. 01

   Tag every automated test with the client account and the feature it guards

   Without ownership tags, suites accumulate tests nobody can attribute to a retainer line, and pruning becomes a political argument instead of a maintenance decision.

2. 02

   Pull 90 days of run history and flag tests that have never failed on a real defect

   A test that passes through every release while contributing no caught bugs is a maintenance cost with no evidence of value. CueCast's step-level failure diagnostics make this audit tractable without reading raw logs.

3. 03

   Separate revenue-path coverage from cosmetic coverage before cutting anything

   Signup, email delivery, and checkout flows deserve permanent protection; a hover-state assertion on a footer rarely does. Mystra-style revenue-path checks are the floor you never prune below.

4. 04

   Set a maintenance budget per client in engineer hours and hold the suite to it

   Fixed-bid projects bleed margin when flaky selectors eat unbilled hours. Write the ceiling into the retainer scope so pruning is scheduled work, not emergency triage.

5. 05

   Quarantine flaky tests for two sprints, then delete or rewrite them

   Quarantining keeps the signal clean while you decide. Tests that stay flaky past two sprints are usually testing the wrong layer and should be replaced with a visual diff or a unit check.

6. 06

   Replace brittle selector chains with recorded or visual-diff coverage where the workflow is stable

   No-code recording tools replay real browser workflows and survive markup churn better than hand-coded XPath. Reserve coded frameworks for logic-heavy paths where visual diffs miss the failure mode.

7. 07

   Log every removal with the date and the reason in the client's QA register

   When a defect later escapes, the register answers whether coverage was deliberately reduced or quietly lost. That record protects the agency in scope disputes.

8. 08

   Re-baseline the suite against the current production build before the next release window

   Run the pruned suite against a disposable copy of production so schema or environment drift surfaces before the client sees a false failure.