End-to-End Testing
End-to-end testing exercises a complete product workflow across integrated parts to check an externally visible outcome.
Download Markdown (.md)Soroush’s Product Wiki · https://wiki.soroush.uk/topic/end-to-end-testing/
01What it means
End-to-end testing follows a workflow through the system boundary the team intends to verify, such as placing an order and seeing its confirmation. It often uses a browser but can start at an API. The environment and any simulated external services must be stated so the claimed coverage matches what actually ran.[1][2]
02Why it matters
A journey can fail because of routing, authentication, wiring, or asynchronous processing even when individual components pass. A small set of complete-workflow checks helps detect these failures before a release reaches users.
03When to use it
Use it for critical journeys and integration risks that smaller tests cannot establish. Prefer focused checks for exhaustive business-rule combinations; browser-driven tests are expensive places to test every calculation. Do not treat automated completion as evidence of usability.
04How it works
- Choose a small set of important workflows and define successful outcomes.
- Provision isolated accounts and predictable data in a representative environment.
- Perform user-visible actions using stable labels and resilient waiting conditions.
- Verify the final business result rather than only a success message.
- Capture useful failure evidence and remove flaky causes instead of masking them with retries.
05A practical example
A marketplace test signs in, reserves a sandbox listing, completes a simulated payment, and verifies the order appears in history. It catches a callback-routing error. The team records that the payment provider’s live processing is outside this test’s coverage.
06Where it breaks
Failures can be slow to diagnose and depend on many systems. Simulated services reduce instability but narrow coverage. Passing a few journeys cannot prove all states, browsers, accessibility needs, or production conditions are correct.
Common mistakes
- Equating a mocked checkout response with a verified payment integration.
- Using fixed sleeps that fail under variable load.
- Expanding browser tests until feedback becomes too slow to act on.
07Sources
- Just Say No to More End-to-End TestsGoogle Testing Blog · Engineering guidanceJust Say No to More End-to-End Tests
- Best PracticesPlaywright · Official documentationBest Practices