Definition
A collection of data validation rules — defined in YAML or JSON — that specify what 'good' looks like for a given dataset. Each expectation is a rule like 'row_count should not change by more than 5%' or 'column_values of revenue should not be null.' Expectation suites are the core primitive of data testing: they encode business logic into reproducible checks.
Why it matters
Expectations turn tribal knowledge into automated checks. 'We learned that revenue should never be negative the hard way' becomes an expectation: 'revenue_sum delta < 0 → fail.' Once encoded, expectations run automatically on every diff — so the same bug can't silently break production twice. CI passing doesn't mean data is correct; expectations passing means data behavior is within defined bounds.
How Well Tested handles it
Well Tested supports expectation suites in YAML format, runnable via `bun run expectations:check`. Suites can be uploaded via the file-compare API. Results from expectation runs are persisted as ExpectationRun records, feeding the cross-signal concern in release risk. Consistent passing expectations over 14+ runs can decrement release risk (dataset is stable); failed expectations add 20–40 points to the risk score.
Related terms
No commitment. 20 minutes. We review your release workflow and show you what's possible.