Blog
Notes on data that arrives late.
Practical writing on freshness, lineage and alerting, by the people who build Tidewater.
-
Late but green, when every job succeeds and the data is still stale
Most data incidents reach a dashboard through jobs that report success. Here is how to measure freshness directly and set expectations a team will trust.
-
How we build column lineage from SQL without running your queries
We read the SQL your warehouse already ran, parse it, and resolve every column back to its source. This is how it works, where it fails and how we check it.
-
A runbook for the dashboard that is wrong at 6 a.m.
A step by step routine for the moment someone says a number looks wrong, with the queries and API calls we use to find the cause in under 15 minutes.
-
Alert fatigue is a routing problem, not a volume problem
Teams mute noisy channels because nobody in them owns the alert. Route each alert to the owner, hold the rest for a digest and the volume stops mattering.
One short e-mail a month
New posts and changelog highlights. Unsubscribe whenever you like.