240 active dashboards. The CFO opened 8 last quarter.
We retired 232 boards, migrated the 8 that survived onto Apache Superset with custom whale + mekko charts, and gave the customer back $420K in year-one seat fees. Decision cycles dropped from three weeks to a workday.
Context.
Challenge. A Series C fintech had built 240 dashboards in three years. Eight got opened in the prior quarter. The CFO trusted one. The licence renewal was up for $480K and the head of data wanted out.
Constraint. Zero downtime, zero data loss, and the seat-fee saving had to be defensible to the audit committee.
Approach.
Catalogued the full estate by view-count and ownership. Killed 232 boards on a 5-view threshold with an opt-out window. Three people replied; the rest let the boards go.
Ported the 8 survivors to Superset. Built two custom chart plugins (whale curve for account contribution, horizontal waterfall for margin variance) and contributed three fixes upstream while we were in there.
Performance pass. Integrated ClickHouse for the top three high-traffic boards with pre-aggregated tables. Top board load time dropped from 15s to under 3s.
SSO + RBAC via OAuth, audit logging on every board, embedded the customer-facing dashboards into the existing portal with row-level security.
Stack.
30 minutes with the founder.
6-week engagement, fixed scope, fixed price.