the team is right, a flakey test is worse than having no tests come up with a strict definition of done build the knowledge and skill set, or hire, to enable higher quality tests here's a sample DoD, for us, anything that doesn't meet it cannot go to main, shift + deleted - no flake - no hard waits/sleeps - stateless, multiple entities can execute - cron job or semaphore where not possible - no order dependency; each it/describe/context block can run with .only in isolation - tests handle their own state and clean up after themselves - deleted or de-activated entities - tests live near the source code - shifted left, as possible - begins with local server or sandbox, works throughout deployments - low/minimal maintenance - enough testing per feature to give us release confidence - execution evidence in CI - some visibility, as in a test report