2 min read
In brief
- Platform-higher-than-store is not attribution generosity — it is almost always double counting from two events describing one order.
- The two causes are mismatched event ids and multiple pixels or datasets claiming the same purchase.
- Fix it before increasing budget: bidding on inflated conversions buys more of whatever produced the duplicates.
Why this gap is the dangerous one
The familiar complaint is that Meta reports fewer purchases than the store, and it is usually explainable: attribution windows, cross-device, unowned traffic. A gap in that direction makes you suspicious of the platform and cautious with spend.
The reverse gap flatters you. Reported purchases exceed real orders, cost per purchase looks better than it is, and the campaign that appears to be winning is being rewarded for events that never happened. Every increase in budget compounds the error.
The four causes, in order of frequency
- 1
Two events, two ids, one order
The browser fires a purchase and the server fires a purchase with a different event id, or with none. The platform cannot merge them and counts two.
- 2
Duplicate pixel or dataset
An app installed its own pixel snippet alongside yours. Both report the same order into the same dataset, and both are believed.
- 3
Purchase fired more than once per order
A confirmation page that reloads, a thank-you route that re-renders, or a webhook that fires twice on a status change. Server-side duplicates are common because nobody watches the endpoint.
- 4
Test orders and manual orders
Your own test purchase, or an order created in the admin, counted as a real conversion. Small in number, disproportionate in confusion when volumes are low.
How to find it in an hour
- Pick one order from yesterday. Pull every event the platform recorded for that value and timestamp window.
- Count them. Two with different ids is the whole diagnosis.
- Read your server log for the order: one row or two? Two rows means the endpoint is firing twice, regardless of ids.
- Check the storefront source for a second pixel id — search for the id string, not for 'facebook'.
- Compare a week of purchase counts against a week of orders excluding cancelled and test orders. A consistent ratio like 1.8 or 2.0 is a structural duplicate, not noise.
The ratio tells you the cause
Near 2.0 means two full streams. A ratio that drifts with order volume but never reaches 2 usually means specific events — reloads, webhooks, a second pixel on a subset of pages.
A gap where the platform reports fewer sales is arithmetic. A gap where it reports more is a bug — and the only useful response is to trace one order through both sides until the ids match.
