Organising measures
Folders, naming, and keeping a growing model navigable.
A model with two hundred measures is only useful if you can find one.
Folders
Measures can be grouped into folders, which nest one level. A structure that mirrors how people ask questions — Revenue, Costs, Customers — beats one that mirrors how the data is stored.
Naming
Names appear in field lists, legends and axis labels, so they are user-facing text:
- Revenue, not
sum_amt - Revenue vs last year, not
rev_yoy_2 - Prefix consistently when it helps sorting —
Revenue,Revenue (LY),Revenue % changegroup naturally.
Build on other measures
Define Revenue once and build everything else from it. When its definition changes — a new exclusion, a corrected column — every measure built on it follows automatically. Redefining the same sum in six places is how six numbers end up disagreeing.
Deleting
A measure that other measures depend on cannot be deleted out from under them. You are told what depends on it, so the choice is informed rather than a surprise the next time somebody opens the dashboard.