One dashboard, a million receipts

Problem

A million transaction rows answer nothing by themselves. The questions a retailer actually asks — who comes back, who pays the bills, what's the real revenue — need an analysis and one screen that shows it.

What I built

A DuckDB pipeline over UCI Online Retail II (a real UK retailer's transactions, Dec 2009–Dec 2011 — a historical dataset, framed as such): cohort retention, revenue concentration, cancellation handling — rendered into a live single-view dashboard where every number is computed by the pipeline, none typed by hand.

Result

1,067,371 rows in. This retailer's wholesale-heavy customer base runs on loyalty: 96.8% of revenue comes from repeat buyers, and the top customer decile alone brings 63.9%. Median next-month retention is 20.5%, and 15.5% of invoices were cancellations the pipeline had to filter before any revenue number could be trusted. The dashboard is live on $0/month hosting.

What this costs you

This is exactly the $500 single-view starter. Full dashboards run $600–$1,500, live in 3–5 days.