A metric catalog is the register of every metric an organization maintains, with its definition, owners, sources and status. It is the single artifact that makes a measurement programme auditable, and it is usually the first thing to fall out of date.
Identifier and name, with the identifier stable for the life of the metric even if the name changes.
The full definition, versioned and dated.
Definition owner and value owner.
Source systems and the fields drawn from each.
Time grain, unit and direction of good.
Thresholds and their basis.
Status: active, deprecated or retired, with the date.
Which goals, risks or obligations the metric serves, so it can be retired when they change.
A catalog kept as a document or spreadsheet describes the metrics as they were understood on the day someone last edited it.
Meanwhile the computation lives somewhere else, in queries, scripts and spreadsheets that get modified as tools change. Nothing forces the two to agree.
Within two quarters the catalog is a description of intent rather than a description of what is being computed. And because it looks authoritative, it gets cited in audits, which is worse than not having one.
The test is simple. Pick three metrics from the catalog and check whether the formula recorded there is the formula actually producing the number this month.
A catalog with three hundred entries is usually evidence of metric sprawl rather than thoroughness. Most of them will have no value owner and no recorded consumer.
The useful discipline is an annual pass asking two questions per metric: who acted on this in the last year, and does its stated purpose still exist. Metrics failing both get retired rather than left dormant.
In Metric Maestro the catalog is not a separate document. The definition that appears in the registry is the definition that computes the number, because there is only one of them.
Each entry carries its version history, owners, sources and the record of every value produced under each version. That removes the drift between the described metric and the computed metric, which is the failure mode that makes document-based catalogs unreliable.
From the blog