Mutation Testing — Stryker Guide for QA Teams (2026)
Code coverage lies. Mutation testing measures the strength of your assertions, not the count of your lines. Full Stryker setup for JS/TS + Java, plus how to hit 80% mutation score without doubling your test time.
Last updated 2026-07-20 · 10 min read · By Avinash K
90% code coverage means nothing if the tests only call the code and never assert. Mutation testing tools like Stryker prove this by mutating your source and checking if any test fails. A surviving mutant = a missing assertion. This is the metric that actually correlates with production defect rate.
Key takeaways
- Why coverage lies and mutation score does not.
- Stryker setup for JS/TS and Java.
- Reading the mutation report.
- Hitting 80% mutation score without doubling test time.
1. Why coverage lies
function add(a, b) { return a + b; }
test('add works', () => { add(1, 2); }); // 100% coverage, zero assertionsStryker would change a + b to a - b and the test still passes — a "surviving mutant." That is your bug.
2. Stryker setup (JS/TS)
npm install --save-dev @stryker-mutator/core @stryker-mutator/jest-runner
npx stryker init
npx stryker runReport at reports/mutation/mutation.html. Java equivalent: PIT.
3. Reading the report
- Killed — mutant broke a test. Good.
- Survived — mutant lived. Missing assertion.
- No coverage — line not tested at all.
- Timeout — usually loop or recursion mutation.
4. Hitting 80% without doubling test time
Scope Stryker to changed files only in PR CI (Stryker supports git diff). Full run nightly. Focus on domain code — skip config, DTOs, and generated code. Related: flaky tests, test data management. Docs: stryker-mutator.io.
Mutation score only rises when the assertions were written with intent. Teams that already practise the workflow in our test-driven development (TDD) guide for QA engineers typically start 20–30 points higher on their first Stryker run.