The question comes before the dashboard.
Most analytics work starts with a request for a dashboard and no agreement about the decision it supports. The result is a screen everybody looks at and nobody acts on. Naming the decision, the person who makes it and the frequency they make it at changes what gets built, and usually makes it smaller.
Two departments, two definitions, one argument.
Revenue, headcount, utilisation and on-time delivery all sound unambiguous until two teams calculate them differently and both are defensible. Until the definition is written down and owned, every report is contestable, and meetings are spent reconciling numbers rather than acting on them.
We define, then source, then present.
A short list of the decisions that matter, a written definition per measure with an owner, an honest assessment of whether the source systems can support it, then the model and the presentation. Where the data cannot answer the question yet, we say so and describe what would have to change, rather than producing a number that looks right.
What you are left with.
A measure dictionary that survives a change of staff, a data model that others can extend, reports tied to decisions rather than to curiosity, and a documented view of where the underlying data is weak and what it would take to fix.