A metric without an owner becomes a rumor with a chart.
Ownership should be visible, reviewable, and boring. The best analytics teams treat definitions the way engineering teams treat interfaces: small, named, versioned, and hard to change by accident.
This is product work because every metric creates a promise to its readers. If the name, grain, or source changes without notice, the product has broken that promise even when the SQL still runs.
{
"metric": "net_revenue_retention",
"owner": "Revenue Operations",
"source": "warehouse.finance.nrr_monthly",
"grain": "account_month",
"freshness": "24h",
"approved_by": "Finance",
"version": "2026-07-01",
"change_policy": "approval_required",
"review_cycle": "monthly"
}
The person or team that can approve a definition change and explain the result in plain language.
The table, stream, or application event that feeds the metric, with expected update timing.
Every material definition change, visible beside the metric instead of buried in a ticket.
SignalLake keeps metric definitions close to the data product, not hidden in a spreadsheet tab. Teams can still move quickly, but they move with a record.
The goal is not to slow analysts down. It is to make ownership obvious enough that a reader can trust the number without opening a ticket or asking which warehouse model happened to feed the latest screenshot.
Use the SignalLake metric contract to record the definition, source, review cycle, and approval trail.