Dashboards don’t usually fail because the charts are ugly.
They fail because the same metric means different things depending on who clicked which filter in which tool on which Tuesday.
That’s not analytics. That’s improv.
The problem: metric drift
Metric drift shows up when:
- definitions live inside dashboards
- spreadsheets become the “final version”
- different teams compute the same KPI with different logic
No one needs to be “wrong” for this to break trust. A metric can be consistent and still be different.
The fix: treat metrics like contracts
A metric should behave like a shared contract:
- Human-readable meaning (plain English)
- Machine-readable computation (formula + filters)
- Ownership (someone is accountable)
- Change control (versioning, approval, comms)
If the metric changes, the contract changes. And everyone should know.
Dictionary vs semantic layer (two different jobs)
Metrics Dictionary (meaning)
The dictionary is where you define:
- what the metric is
- why it exists
- what decision it supports
- what it is not
Semantic Layer (enforcement)
The semantic layer is how you make “define once” real:
- compute the metric centrally
- reuse it across dashboards, spreadsheets, and apps
- reduce the surface area for duplication
A dictionary is policy. A semantic layer is compliance.
The Decision-Grade Metric checklist
Before a KPI is allowed in an exec view, it should answer:
- What question does this KPI answer?
- What’s the exact definition?
- What’s the formula (inputs + filters + timing)?
- What systems feed it?
- Who owns it?
- What breaks it (common failure modes)?
- How do we know it’s still true? (integrity checks)
If you can’t answer these quickly, the KPI is a conversation starter, not an instrument.
One warning label
If your company’s highest-paid people spend time arguing about definitions, you’re using executives as a semantic layer.
That is… not the optimal use of a human.
Bottom line
Metrics should be:
- defined once
- governed like contracts
- reused everywhere
That’s how you get speed without superstition.