📖 Business
Testing Strategy Pyramid
The test pyramid is a model for balancing test types by cost, speed, and confidence. Visualize a pyramid: a wide base of fast, cheap unit tests; a middle layer of integration and service tests; and a small top of slow, expensive end-to-end UI tests. The shape matters — invert it (lots of E2E tests, few unit tests) and you get a slow, brittle, expensive test suite that nobody trusts. Get the shape right and you get fast feedback, targeted coverage, and a pipeline that runs in minutes instead of hours. Humble and Farley place testing strategy early in the book because without reliable automated tests, a deployment pipeline is just a deployment script with extra steps.
2
Minutes
2
Concepts
+45
XP
1
How It Works
The pyramid layers:
Unit Tests (base — widest layer)
- Test individual functions, classes, or modules in isolation
- Run in seconds, catch logic errors immediately
- Should number in the hundreds or thousands
- No network, no database, no file system — pure logic
Service/Integration Tests (middle layer)
- Validate interactions between components
- Test API contracts, database queries, message passing
- Slower than unit tests but catch integration failures
- Use test doubles for external dependencies where possible
Acceptance Tests (upper-middle layer)
- Written in business language, verify business requirements
- "Given a user with a premium account, when they request a report, then it includes advanced analytics"
- Run against a deployed application in a staging environment
- Owned jointly by developers, testers, and business stakeholders
End-to-End / UI Tests (top — smallest layer)
- Test complete user journeys through the real UI
- Slowest, most brittle, most expensive to maintain
- Keep these minimal — only critical happy paths
Nonfunctional Requirement (NFR) Tests
- Performance, security, capacity, accessibility
- Run separately because they require production-like environments and load
Rules for all test types:
- Fast — the pipeline cannot wait hours for test results
- Reliable — no flaky tests. Fix them or delete them.
- Deterministic — same inputs produce same outputs, every time