A 1,467-order revenue audit: 4 real bugs, and ~$197K found at month-end

This is real work from my time running BI and data automation for Food Service Direct – Unilever USA, an e-commerce business selling food-service products across the US and Canada. The revenue report in Power BI and the e-commerce platform were close, but not the same. Close is not good enough when leadership makes decisions from the report.

The setup

Revenue was reported in Power BI from a reporting view built on the store's database. The store platform (Magento) was the source of truth for orders. Small gaps between the two had become normal, and nobody could say exactly where they came from.

The check

Instead of comparing monthly totals, I checked every order from January to April 2026, one by one: 1,467 orders, the report against the platform.

• 1,234 orders matched exactly.
• 214 differences were expected: cancellations handled the way the business wanted.
• The rest pointed to 4 real bugs in the reporting view.

What it found

• Edited orders were counted twice, so revenue was overstated.
• Duplicate order lines also inflated revenue.
• One case went the other way and under-counted sales.
• ~$2,470 of refunded or cancelled orders needed a clear policy decision: count them or not.

I proposed a permanent fix for the reporting view, so the same problems could not come back next month, and wrote up the policy question in plain English for the business team.

The ~$197K month-end bug

The same habit paid off on a monthly sales export for a manufacturer partner. Its date range was edited by hand each month, and the end date quietly dropped the last day of every month. Across 2026 that was 559 rows and about $197K of sales missing from the partner's data. I automated the date range and fixed the cut-off, so the full month now goes out every time.

What this means for you

If your dashboard and your store platform don't agree, or you're not sure your numbers are right, this is exactly what my Data & Revenue Audit does: an order-by-order check, a written list of fixes in order of value, and the fixes built if you want them.