Skip to content
Automatizav1
CS-01

Cut executive dashboard latency from 8 s to 400 ms

Daily exec dashboard p95 query latency dropped 95%. Replaced ad-hoc cubes with dbt-first marts, partitioned by event date.

  • p95 latency8s → 400ms
  • Models migrated47
  • Engagement5 weeks
CONTEXT

A European sportsbook had grown a dashboard suite over 4 years. Each dashboard built its own pre-aggregations; the same metric existed in 3 different tables with 3 different definitions. Daily exec page took 8 seconds to load, and the head of trading didn't trust the number.

APPROACH

Audited 47 ad-hoc cubes, mapped them to a metric tree, rebuilt as dbt marts on date-partitioned tables, deprecated the duplicates, ran a 2-week parallel.

OUTCOME

Exec dashboard p95 < 400 ms. Two of the three metric definitions were wrong — corrected with documented owner sign-off. Trading committee reviews now reference one number, not three.

Wire a real pipeline to your warehouse

30-min stack review. We come back with a scoped plan, an owner, and a freshness contract. No SDR, no slide deck.