Notes on governed analytics
Why teams stop trusting their numbers, and what it takes to get the trust back.
A semantic layer isn't a doc. It's an owner.
A semantic layer works when one governed definition owns each metric and every human or AI query runs through it. A text document alone cannot enforce that agreement.
Read post →"How many customers do we have?" is never a simple question
A data consultant found two dashboards disagreeing on revenue, not from a broken formula but from two teams never agreeing what the metric meant. The fix is an owner and a single definition, not a debugging session.
Read post →Why agencies stopped sending clients dashboards and started sending 3 numbers
Agency client reports are shrinking to a short note with 3-4 numbers. The format works, but it only stays trustworthy if the numbers don't move underneath it.
Read post →You didn't get into analytics to spend an hour explaining a Slack thread
The hour lost explaining why two dashboards disagree is the tax of having no single metric definition to point to. Here's what removes the tax instead of paying it faster.
Read post →Three "correct" revenue numbers, one meeting, zero broken formulas
Three teams brought three different revenue numbers to one meeting, and no formula was broken. The real failure was an undocumented disagreement about what revenue means.
Read post →Why teams still keep a shadow Excel file even after buying a BI tool
Every company with a real BI tool still has a shadow spreadsheet. The tool was never the problem. The missing metric definition was.
Read post →