Two arguments happen on almost every software project. The first is at the start of a test cycle: the build has arrived, the testers try it, half the screens throw errors, and a day is lost before anyone admits the build was never ready. The second is at the end: the release date is tomorrow, eleven defects are still open, and the room splits between “ship it” and “we can’t”. Entry and exit criteria exist to settle both arguments before they start, by writing down the conditions while everyone is still calm.
Two definitions
Entry criteria are the conditions that must be true before a testing activity begins. Exit criteria are the conditions that must be true before it can be declared complete. The ISTQB glossary uses these terms, and agile teams express the same ideas as a “definition of ready” for work entering a sprint and a “definition of done” for work leaving it. The vocabulary changes; the purpose does not.
Criteria can be set for any testing activity: a system test cycle, a UAT phase, a performance test, or a single sprint. They usually live in the test plan, and they only work if the people who will be affected by them (developers, product owners, release managers) agreed to them in advance.
What good entry criteria look like
Entry criteria protect testers from wasting effort on something that cannot yet be tested meaningfully. For a system test cycle on a payments feature, a sensible set might be:
- Requirements or user stories for the scope are signed off, with acceptance criteria.
- The build is deployed to the QA environment and the deployment notes list what changed.
- Unit tests and the automated smoke suite have passed on that build.
- The payment gateway sandbox and any stubbed services are reachable.
- Test data (customer accounts, cards, UPI handles in sandbox) is loaded.
- No open blocker defects from the previous build in the areas being tested.
Notice that each item can be checked with a yes or no. “Build is reasonably stable” is not an entry criterion; “smoke suite passed” is. The single most valuable entry check on most projects is the smoke test, because it converts the vague question “is this build any good?” into a fifteen-minute automated answer.
What good exit criteria look like
Exit criteria describe “done enough” for the activity. Typical categories are coverage, results, defects and documentation. For the same payments feature:
- All planned test cases for high-priority requirements have been executed.
- At least 95 percent of all planned test cases have been executed, and every unexecuted one has a recorded reason.
- No open defects of critical or high severity.
- Open medium-severity defects have been reviewed by the product owner and either accepted with a workaround or scheduled.
- The regression suite has passed on the release candidate build.
- The test summary report has been shared.
The specific numbers vary by organisation and risk. A banking core system might demand 100 percent execution of planned cases and zero open high-severity defects. An internal reporting tool might accept less. What matters is that the number was chosen deliberately, not that it matches anyone else’s.
A scenario: the criteria under pressure
Consider a team two days from a festive-season release. The exit criteria say no open high-severity defects. There is one: when a UPI payment times out and the customer retries, the order occasionally shows as unpaid even though money was debited. The product head points out that it affects a small fraction of transactions and the campaign is booked.
Without written criteria, this becomes a contest of seniority. With them, the conversation changes shape. The criterion is not met. The team can fix the defect, or someone with the authority to accept the risk can formally waive the criterion, with the waiver recorded alongside the reason, the mitigation (perhaps a manual reconciliation job running every hour, and a customer care script) and the date by which a fix will follow. That is a legitimate decision. What is not legitimate is quietly reclassifying the defect to medium severity so that the criteria appear to pass.
This is the real value of exit criteria. They do not make the decision for anyone. They make it visible who decided, on what basis, and what risk was accepted.
Suspension and resumption criteria
Between entry and exit sits a third, less discussed pair. Suspension criteria say when testing should stop mid-cycle because continuing is pointless: for example, if a blocker defect stops more than a set share of planned tests, or if the environment is down for more than half a day. Resumption criteria say what must happen before testing restarts, typically a fixed build that passes smoke tests. These stop teams from logging dozens of duplicate defects that all trace back to one broken login service.
Common ways the idea fails
Criteria nobody can measure. “Adequate test coverage” and “acceptable quality” give no one anything to check. If a criterion cannot be answered yes or no from data you actually collect, rewrite it.
Metrics that reward the wrong behaviour. If exit depends on a 95 percent pass rate, people may be tempted to remove failing tests from the plan or mark them “not applicable”. Pairing execution percentage with a rule that every exclusion needs a recorded reason helps.
Treating entry criteria as a wall. Rigidly refusing to test anything until every entry criterion is met can waste time too. Many teams allow partial entry: if the payments module is ready but the reports module is not, testing starts on payments, and the criteria are applied per area.
Criteria written by QA alone. If developers and product owners did not agree to them, they will be ignored the first time they are inconvenient. The criteria are a contract between groups, not a QA rulebook.
Who makes the final call
A common misunderstanding is that meeting exit criteria means QA “approves” the release. In most well-run organisations, QA reports whether the criteria are met and what the residual risk is; the business owner (product head, release board, or change advisory board in larger enterprises) makes the go or no-go decision. Testing provides information. Accountability for shipping sits with the people who own the product and its outcomes. Exit criteria make that division of responsibilities explicit, which is better for everyone, including the testers.
Written by the RP Strativa Team. The criteria and thresholds shown are illustrative and should be set to suit each project’s risk.
Leave a Reply