Three Numbers That Predict A Rebuild

Dashboard query time, pipeline failures per week, days to first data change. Pull all three this afternoon.

I pull these in the first week of any review.

1. Query time on your five most-used dashboards, today versus six months ago. Drift from 4 seconds to 40 happens quietly, one extra join at a time.
2. Pipeline failures per week, counted from the orchestrator logs. If it’s climbing while the number of pipelines stays flat, something underneath is wearing out.
3. Days until a new engineer ships their first data change. Under a week is healthy. A month means the system lives in two people’s heads.

Any one of them can have a boring explanation. A dashboard got a heavy new chart. A vendor had a bad week. Two moving the wrong way together is when I’d start planning.

A spreadsheet you update once a month is enough to track them.

If two are drifting, the checklist I use for the full picture is here: https://thomasnys.com/data-platform-assessment-checklist/