Product · BI / Analytics
Enterprise BI Dashboard
Frontend Engineer — built from scratch
Under NDA — client and product details withheldThis project is under NDA — the client, product name, and any screenshots can't be shared. What follows is limited to my role and the technical approach: a business intelligence dashboard built for an enterprise Amazon seller doing $10M+ in GMV — revenue, order-level margin, and inventory rolled up across six figures of monthly orders, for an ops team that needed to catch margin erosion before it showed up in the P&L.
The Challenge
The seller's order history ran past 100,000 rows and kept growing, and the team needed to filter, sort, and export any slice of it — by date, SKU, or status — without the browser tab stalling or the export timing out. A dashboard that only handled the first page of data wasn't going to survive contact with the actual dataset.
The Approach
- Virtualized table rendering so only visible rows mount to the DOM, keeping scroll and filter interactions smooth over 100,000+ orders
- Cursor-based pagination against the API instead of offset pagination, avoiding the slow-count problem on a table that size
- Debounced, server-side filtering (date range, SKU, status) so the client never holds more than one page of results in memory
- Client-side CSV/PDF export built as a background job with a progress state, rather than blocking the UI while the file streams down
- Margin and revenue aggregates computed server-side and cached per period, so the dashboard renders instantly instead of summing rows in the browser
The Outcome
−45%
initial render time via virtualized tables + cursor pagination
100k+
orders filterable and exportable without a client-side stall