Bootstrap the measurement plane: sources, schemas, pipelines, metrics, dashboards, and the rules that keep them honest.
metricv0.2.0When to run it
When the team has shipped features without instrumentation, when metrics are being computed from spreadsheets nobody trusts, or when an analytics infrastructure migration is underway.
What you'll have at the end
data_source records mapped to event_schema entities, metric definitions with their formulas and target outcomes, dashboard entities aggregating the metrics into views.
Common starting point
The three metrics the team currently quotes most often. Capture their definitions, the data sources behind them, and any disagreements over how they're computed.
Sequence summary
One step: run the data_analytics creation sequence — data sources, event schemas, metrics, dashboards.
List where the truth lives: the systems that emit events you can measure. Product, billing, support, CRM, external.
Define the events you will instrument. Schema first; instrumentation second. Each event needs a name, properties, and emit conditions.
Move data from source to warehouse. Define transformations and the resulting models the rest of the org consumes.
Define the numbers the team will look at. Group into North Star, input metrics, guardrail metrics, and diagnostic metrics.
Compose metrics into dashboards by audience: leadership weekly, team daily, on-call always-on.
Set the rules that guard against silent drift: freshness, completeness, accuracy, schema integrity.