First of all, software tests have to be stable, predictable, and reproducible. The same project circumstances must predictably produce the same test case outcomes, that is, passing or failing. Tests that sometimes pass and sometimes fail are not helpful at all. They either distract developers by providing false positive results or suppress actual bugs with false negative outcomes.

Circumstances that need to be taken into account are, among others, the current time, time zones, locales, randomly generated data, and concurrent execution of other tests that could interfere. The test scenarios should be predictably and explicitly set up, so that these circumstances have no influence on the outcome. If the tested functionality is in fact influenced by these factors, this is a sign that there is a need for additional test scenarios, considering different configurations.