Four Hundred New Checks, One Missed IncidentQuality · medium · ~13 minThree months ago the platform team rolled out auto-generated quality checks across every critical dataset — freshness, volume, nulls, duplicates — and wired them all to #data-quality-alerts.Last week a genuinely corrupted revenue table sat undetected for two days. The check HAD fired. Nobody saw it: it was alert #83 of 214 that day.Leadership's conclusion: "the quality platform doesn't work." Yours needs to be better.Evidence

Metrics

Config

Logs

Checkpoints

The check fired and the incident was still missed for two days. What's the honest diagnosis?

31 of the alerts during the incident were downstream tables re-reporting the same root cause. Which mechanism fixes exactly this?

On day 4 the platform produced 214 alerts containing 2 true incidents. What is its alert precision, in percent? (round to one decimal)

Write the redesign that makes the 405 checks trustworthy — the alerting policies you'd change and the one metric you'd put on the platform team's own dashboard.