Meta reported 28 sales. The ledger had 22.

What we found

The payment provider's payout came in light. The money landing in the bank was lower than the platform's sales count said it should be. Over one seven-day window, the advertising platform reported 28 sales. The bank had settled 22.

What the numbers showed

A gap of 6 sales against 22 is a 27% overstatement. Going looking for the leak found three separate tracking integrations, all firing on the same purchase. Every real sale was being counted more than once by the platform, with no single integration at fault on its own.

Why it failed

A platform counts what it thinks it caused. When more than one tracking method is wired to the same checkout event, pixel, server-side event and a manual integration, for example, the platform's deduplication can fail to merge them, and one purchase gets logged more than once. The platform isn't lying. It's measuring its own contribution with instruments that were never built to check each other.

What we changed

The duplicates were killed, integration by integration, until the count matched. The payment ledger became the source of truth: a sale is a sale when the money settles, not before. A five-minute check now runs before every launch, buying the product on a live ad, checking the payment metadata carries the campaign, and confirming the sale lands once, not twice.

How to spot it in your own account

Put one row on your weekly reporting page: platform-reported sales against settled bank sales, for the same seven complete days. If the platform number runs consistently higher, list every tracking integration wired to your checkout, pixel, Conversions API, any third-party pixel manager, and check whether more than one of them is sending the same purchase event. Measure your own gap rather than trusting a vendor's quoted 'healthy' range; those figures carry no sample or method behind them.