Business Dashboard

The challenge

An executive team was spending hours per week reconciling reports from four separate systems: a CRM, a finance platform, a support tool, and an operations database. The spreadsheets were always out of date, always required a human to produce, and always raised more questions than they answered.

They wanted a single source of truth — in real time, not overnight batch.

What we built

• Data pipeline: Python/FastAPI service that polls each source on its own cadence, normalizes records into a shared schema, and writes to a PostgreSQL warehouse.
• Real-time layer: WebSocket push for the metrics that change fastest (sales, support ticket volume).
• Dashboard UI: React frontend with D3-backed visualizations — time-series, cohort analysis, geographic breakdown, and drill-downs that work on every chart.
• Role-based access: engineers, sales, finance, and executives each see a tailored view. Nothing visible that shouldn't be.
• Export: one-click CSV and PDF reports for the exec team's board decks.

Results

• Weekly reporting time dropped from ~10 person-hours to ~15 minutes (just a spot-check).
• New metric requests turn around in hours, not weeks.
• Exec team adopted it as the first tab open in the morning.

What we learned

The data pipeline is always harder than the UI. We spent roughly 60% of the engineering budget on the ingestion and normalization layer, and that turned out to be the right split — because when the data is clean and fresh, the UI practically writes itself.

Also: real-time isn't always the right answer. Some metrics update once a day and should. Pretending they're live would just add cost. Match the update cadence to the source reality.