Software testing time paradox

Exhaustive testing is impossible (the number of input combinations and paths is practically infinite), yet the product must not be delayed. In real projects this paradox is handled as follows:

1. Test early and continuously: testing starts with requirements and design reviews and continues through unit, integration and system tests, so that bugs are found when they are cheap and not left to a long final test.
2. Risk-based prioritisation: test the critical, frequently used and high-risk functions most thoroughly; give less time to the low-risk features.
3. Smart test design: use equivalence partitioning, boundary value analysis and coverage criteria to get maximum fault detection with a minimum number of test cases.
4. Automation and regression suites: automated tests run quickly and repeatedly, and tests run in parallel on many machines and in continuous integration.
5. Defined exit criteria: testing ends when coverage, pass rate and defect-rate targets are met, using reliability growth data (the bug discovery rate falls) and the cost-benefit of continuing.
6. Staged release: alpha/beta releases to real users, followed by patches and updates after release (maintenance).
7. Good development practice: reviews, code standards and static analysis reduce the number of bugs entering testing.
8. Parallel work: test planning and test-case writing during development, and separate test teams.