Meta's numbers don't match my store's orders — which one do I trust?
Last updated:
Short answer
The two figures differ by nature rather than by error: the platform credits an order to the day of the ad inside an attribution window you chose, your system records it on the day of the order, and the platforms do not talk to each other, so each credits the same order to itself. Use your own system's numbers to judge profitability, and the platforms' numbers to compare campaigns and creative — after aligning the window and the date basis.
The attribution window and the date basis
The platform credits an order to the day the customer saw or clicked the ad, within a period you set — a seven-day click, for instance. Your system knows nothing about the ad, so it records the order on the day of the order.
So the two numbers do not match even with perfect measurement: an order clicked on Sunday and completed on Wednesday lands on two different days. Before any comparison, align the window and the date basis, and compare across two weeks or more rather than a single day.
Adding platforms together means double counting
A customer sees your ad on Instagram, searches your name on Google, then buys. Meta credits the order to itself, Google credits it to itself, and each is honest from where it stands.
So platform orders must not be summed and compared with your total sales: the sum will exceed your actual business. The correct total comes from your own system alone, and each platform is read against itself.
Pixel and conversions API: one event counted twice
Most accounts now send the event from the browser and from the server, which is correct and wanted. But if the two do not carry a shared identifier, orders are counted twice and the return looks excellent for no reason.
The quick check: divide platform orders by your system's orders for the same period. If the ratio sits consistently near double, the cause is deduplication, not campaign success.
Cash on delivery: an order counted, then cancelled
In Egypt and Saudi Arabia a large share of orders arrive cash on delivery. The platform counts the order the moment it completes on the site, and knows nothing about a refusal at the door.
So the cancellation and return rate has to come off before judging the return, and belongs inside the break-even calculation from the start — otherwise the campaign looks profitable on screen and unprofitable in the bank.
Which number belongs in which decision
For profitability and the overall budget decision: your own system, after returns. It is the only one that matches your bank account.
For comparing two campaigns or two creatives: the platform, because it is the only party that knows which ad the customer saw. Never compare a campaign on one platform with a figure from another.
For pausing and scaling: both — the platform says which campaign is better, your system says whether the overall level is profitable at all.
Reconciling the numbers in four steps
Fix the window
Choose one attribution window across platforms and always compare against it, instead of changing it each time you ask about results.
Align the date basis
Compare across two weeks or more rather than day by day, because the order and the ad do not fall on the same day.
Check pixel and conversions API deduplication
A steady ratio near double between platform and system means a duplicated event, not exceptional performance.
Take returns off before judging
Apply your actual cancellation rate to the platform's figures, then compare the return with your break-even point rather than a general average.
Frequently asked
- Which number goes in my monthly report?
- Your system's figure for sales and profitability, with each platform's numbers beside its own campaigns. A report that sums the platforms into one number and compares it with sales always paints a better picture than reality.
- A longer window or a shorter one?
- Longer gives prettier, less accurate numbers; shorter gives harsher, truer ones. What matters is settling on one, and knowing that changing it changes the numbers rather than the performance.
- Does a gap between the two numbers mean the setup is broken?
- Not necessarily. A reasonable, stable gap is normal, given the window and the date basis. What warrants inspection is a huge gap, or one that swings sharply month to month with nothing in the campaign to explain it.