One measurement plane, from data source to dashboard

A data source🛢️data_sourceA data source or integration emits the event schema🔠event_schemaAn event schema for tracking nodes a team instruments, defines the metric🔢metricA unified metric that measures progress, health, or behaviour across the product it grounds, is processed via a data pipeline🪣data_pipelineAn automated pipeline for data transformation, and is traced via data lineage🌳data_lineageA record of data origin and transformations. And each event schema tracks the funnel step🔽funnel_stepA stage within a conversion funnel it stands for.

The event is a typed node before any code emits it, so the metric names the exact event it counts. The column it reads is documented in the graph rather than inferred from a query.

A data pipeline🪣data_pipelineAn automated pipeline for data transformation reads from one data source🛢️data_sourceA data source or integration, writes to another, and feeds the data product🎁data_productA curated, reusable data asset the organisation consumes, and the data model🗃️data_modelA data model or schema it produces is persisted in a database schema🗄️database_schemaA database schema definition.

A number is traceable through every transformation rather than asserted at the end of the job. When a metric looks wrong, the lineage walks back hop by hop to the source it came from.

A north-star metric🔢metricA unified metric that measures progress, health, or behaviour across the product decomposes into the input metrics beneath it and drives the outcome🏆outcomeA desired business or user outcome it stands for. Each metric is validated by a data quality rule🧹data_quality_ruleA data quality validation rule, so a number that fails freshness or completeness is flagged at the source rather than in a review.

The tree makes the leverage explicit. Move an input and the graph shows what it rolls up to; question the north star and it names the inputs that move it. Guardrail metrics sit alongside, so a win on one number cannot quietly break another.

A metric🔢metricA unified metric that measures progress, health, or behaviour across the product is a connected number. It measures a key result📈key_resultA measurable result tied to an objective, drives an outcome🏆outcomeA desired business or user outcome, decomposes into its inputs, guards against the metric🔢metricA unified metric that measures progress, health, or behaviour across the product it must not break, and is segmented by a persona🎭personaAn archetype representing a user segment.

Measurement and strategy sit one edge apart. The question of what a metric is for resolves to the key result it proves and the outcome it serves, both named on the graph.

A dashboard🎛️dashboardAn analytics dashboard tracks the metric🔢metricA unified metric that measures progress, health, or behaviour across the product nodes an audience reads, contains the report📰reportA structured analytical report it drills into, and can even contain a live experiment run▶️experiment_runAn execution instance of an experiment that records actual conditions, observations, and raw results..

Each dashboard points to the same metric nodes everyone else uses, so a number on the leadership board is the same number the team and the on-call see. There is one metric node behind every view, so no second definition exists to drift.

A metric🔢metricA unified metric that measures progress, health, or behaviour across the product is validated by the data quality rule🧹data_quality_ruleA data quality validation rule nodes it must pass (freshness, completeness, schema integrity) and assessed by a dated metric quality assessment🩺metric_quality_assessmentAn assessment of whether a metric is well-defined, measurable, and aligned to a meaningful outcome..

A number that fails its rules is flagged at the source, and a stale or untrusted metric is marked as distinct from a fresh, verified one. Whether a number can be trusted is a property on the graph, recorded before anyone quotes it.