The problem
Seventeen dashboards disagreed about revenue. Finance ran one number, product ran another, and the exec team had stopped trusting any of them. The data warehouse had become a dumping ground with no shared definition of a metric.
What we did
We built real-time ingestion with Kafka feeding a Snowflake warehouse modelled in dbt, then put a single metric layer on top — one definition of revenue, one definition of ARR, one definition of churn, enforced in code.
Every team migrated onto the same definitions, dashboards were retired or rebuilt against the metric layer, and freshness tests now page a human the moment any pipeline goes stale. Metabase gives business users a safe slice without touching SQL.
Impact
Within a quarter, "what's the number?" stopped being a meeting topic. Event freshness holds at 5 seconds, four teams share one source of truth, and new metrics ship as code reviews instead of spreadsheet round-trips.