Alert on Broken Promises, Not Weird Numbers
Distinguish customer-impacting data symptoms from diagnostic cause signals when designing data observability alerts.
The move: separate the promise from the evidence. Symptom first In data observability, the symptom is the data product failing the job people rely on it to do. A dashboard is stale before an executive review. A modeled table violates its contract before a downstream machine-learning feature job runs. A metric changes enough that the business decision would change. These are promise failures. Causes second Cause signals are still valuable. Freshness lag, null spikes, schema drift, row-count deltas, warehouse saturation, and failed retries help you investigate. They become dangerous only when every cause signal receives page-level urgency. The alert should…
Sign up free — one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in