Acceptance criteria come from requirements, early
The acceptance test plan is written against the requirements, and the criteria are agreed before design and procurement are finalised. A requirement that cannot be tested cannot be accepted, so wording that is vague, such as “fast” or “highly available”, is converted into measurable conditions at the start.
Trace every requirement to a test
A traceability matrix links each requirement to the tests that demonstrate it, and each test back to a requirement. Gaps in either direction show up immediately.
| Requirement | Acceptance criterion | Test reference | Method | Evidence |
|---|---|---|---|---|
| Site remains connected after loss of one WAN link | Failover to secondary link in under the agreed time with application sessions preserved | NET-FO-01 | Disconnect primary link; time recovery | Timed log, session record, signed test sheet |
Choose the right test levels
- Installation verification: equipment, labelling, cabling and power match the approved drawings.
- Configuration verification: devices match the approved configuration and parameters.
- Functional testing: each function works as specified.
- Integration testing: interfaces exchange data correctly, including invalid, duplicate, delayed and missing data.
- Performance testing: capacity and response times under defined load.
- Failover and recovery testing: redundancy operates and backups restore.
- Field validation: devices and communications perform at the real site.
- Documentation review: as-built records are complete and correct.
Write each test case so anyone can run it
- Identifier and the requirement it covers.
- Preconditions: system state, data, tools and access needed.
- Procedure: numbered steps without ambiguity.
- Expected result, stated measurably.
- Evidence to capture: logs, screenshots, measurements, witness signature.
- Pass or fail rule, including what is recorded when a result is borderline.
Define entry and exit criteria
Entry criteria say when testing may begin: installation complete, configuration frozen, test environment and data ready. Exit criteria say when it may end: all planned tests executed, results recorded, and defects at or above a stated severity closed.
Classify defects and agree closure rules in advance
| Class | Description | Effect on acceptance |
|---|---|---|
| Critical | A required function does not work, or safety or security is compromised | Blocks acceptance |
| Major | A required function works with a significant limitation | Blocks acceptance unless a workaround and a closure date are agreed |
| Minor | Does not affect the required function | Accepted; tracked to closure in the punch list |
Retest after each fix, and re-run affected regression tests, since a correction in one place can alter behaviour elsewhere.
Keep the evidence
Executed test sheets, measurements, logs and sign-offs form the acceptance record. They also serve later as a baseline: the performance measured at acceptance is the reference for future operations.