Member-only story
One Metric, One Definition
What a semantic layer actually is, and why analytics engineers should care
Read Free for non-members
It happens in almost every data team at some point. Someone asks a simple question in a meeting, what is our active user count this week, and three people open three different tools and come back with three different numbers. Nobody made a mistake. Each tool was just using its own definition of active user.
None of this is caused by bad analysts or bad tools. It happens because the logic that defines a metric lives in too many places at once.
The problem with scattered logic
When a company grows, it usually ends up with several BI tools, a few dashboards built by different teams, some Python notebooks, and a handful of spreadsheets that “just work”. Each of these rebuilds the same business logic from scratch. Revenue gets computed one way in the finance dashboard and a slightly different way in the product dashboard. Nobody notices until someone compares two numbers that should match and do not.
The instinct is often to fix this with more documentation. Write down what “active user” means, share it in a wiki, ask people to follow it. This helps for a while, but documentation does not enforce anything…
