Data Quality

Define rules that watch a table and alert you when something looks wrong — the table emptied out, a sync stopped, a column filled with nulls, or the row count swung wildly. Rules run automatically on a schedule and keep a pass/fail history, so you find out a feed broke before your dashboards do.

Rule types: Row count minimum (fails below N rows); Row count drift (fails when the count moves more than X% vs the previous check — the first check just records a baseline); Null rate (fails when a chosen column’s NULL fraction exceeds a threshold); Freshness (fails when the newest row is older than N hours, measured from a timestamp column you name; no rows with a value there = stale); and Row count parity (row_count_parity — tunes the full-mirror source-vs-destination check that runs during a sync; per-entity tolerancePct/toleranceRows, or disable to silence its warning). See the admin control plane / DAB sections for the exact set_dq_rule config shapes.

Lives under Settings → Data Quality. Each rule binds to a connection (which supplies the alert recipients and email switch) and a table. Add / edit / delete rules, see a status pill per rule (Passing / Failing / Not checked), and open a rule’s result history (every evaluation with its measured value, pass/fail, timestamp). A rule can be toggled off (kept but not evaluated). Rules run on a background schedule (~every 5 minutes); Evaluate now re-checks one rule and Evaluate All runs every enabled rule immediately.

Alerts are kept quiet two ways: they respect the per-connection Email on warning switch (a violation is treated as a warning — off means the failure is still recorded in history but no email is sent), and they fire only on the pass→fail transition (one alert when a rule first goes red, not every cycle; a later recovery-then-fail is a fresh transition). Managing Data Quality rules needs the data.ingest permission (Admin role; the owner always has it) — the same tier as Import and the raw-SQL tools. Deleting a rule also removes its result history.