How event deduplication works
Modern tracking sends the same action twice on purpose. The browser sends one copy through a tracking pixel such as the Meta Pixel, and your server sends a second copy through the Conversions API, so that if the browser copy is blocked — an ad blocker, browser tracking prevention, a tab closed too early — the server copy still arrives. That redundancy is the whole point, and it creates the problem it then has to solve: one purchase, two records.
Deduplication is how the platform decides those two records describe the same purchase. Both copies carry a shared identifier, normally an event ID generated once and passed into both paths, alongside the event name and its timing. When the second copy arrives and the platform recognises the pairing, it keeps one and discards the other. Meta matches on event name plus event ID; Google’s equivalents lean on transaction and order IDs in much the same spirit.
Why event deduplication matters
Without it, every sale that succeeds on both paths is counted twice. Reported revenue inflates, cost per acquisition looks better than it is, and return on ad spend flatters campaigns that have not earned it. The damage does not stop at reporting: automated bidding learns from those same inflated numbers, so budget drifts towards whatever happens to double-count most.
Getting it right is also what makes running both paths safe, which is the reason for adding a server connection in the first place. Deduplication is the condition attached to redundancy.
Where event deduplication goes wrong
The identifier is where nearly all failures live. If the browser and the server each generate their own ID, nothing matches and both copies survive. If the ID is regenerated on a page refresh, a customer who reloads the confirmation page creates fresh duplicates. If the event name differs by a single character between the two paths, the platform will not attempt the match at all.
Timing accounts for the rest. A server copy that arrives long after its browser twin can fall outside the matching window and be treated as new, and so can a queued request retried after a failed job. A confirmation page that customers can revisit or bookmark will fire again unless the site refuses to re-send an order it has already reported.
How to act on it
Generate the event ID once — on the server, or at the moment the order record is created — and pass that same value into both the browser call and the API call. Where an order or transaction number already exists, reuse it: it is naturally unique and already stored, so nothing has to be invented or kept in sync. Keep the event names identical, character for character, across both paths.
Then verify rather than assume. Events Manager reports whether events were received from both sources and whether they were deduplicated or processed separately, and a test purchase checked against your own order records will expose a gap quickly. Where a browser pixel and a server connection run together, treat deduplication as part of the pixel and Conversions API build rather than a later fix, and re-test after any change to the checkout.
