When metric dictionaries stall (and how we unblock them)

Metric dictionaries often stall because ownership is fuzzy: analytics writes definitions, finance edits them in a slide deck, and product ships a dashboard that quietly diverges. We start by mapping the three artifacts and choosing one canonical row-level test per metric.

The second stall is vocabulary drift—teams reuse words like “active” without attaching a timestamp predicate. We force a silly-but-useful exercise: write the metric as a SQL comment first, then argue about nouns.

The third is calendar anxiety. If you cannot ship a perfect dictionary, ship a dated stub with explicit unknowns. Stakeholders tolerate visible scaffolding more than silent contradictions.

We keep appendices short on the blog; if you need primary sources or query snippets referenced in class, ask via the contact desk with the module name in the subject line.