What an event ID does
When you report a conversion twice — once from the visitor’s browser and once from your server — the platform receives two separate messages describing what looks like two separate purchases. It has no way of knowing they are the same one unless you tell it. The event ID is how you tell it: a short reference generated at the moment the action happens and attached to both copies.
The platform matches on the identifier together with the event name. Two reports with the same name and the same ID, arriving close enough together, are treated as one action. Change either, and the match fails.
The value itself does not matter, only that it is the same on both sends and different for every action. An order number is ideal, because it already exists, is already unique, and is already known to both the page and the server. Where no such reference exists — a contact form, a newsletter sign-up — generate a random value and carry it through.
Why the event ID matters
Without it, a setup that sends browser and server events counts most conversions twice. Reported cost per acquisition halves, return on ad spend doubles, and every decision made from those numbers is wrong in the same direction: you conclude the campaign is working better than it is and keep spending on it.
The damage is not only in reporting. Automated bidding learns from the conversions it is told about, so inflated counts teach the system that certain audiences and placements perform better than they really do. It then buys more of them. A missing identifier is one of the few tracking faults that actively makes an account worse rather than just harder to read.
Common mistakes with event IDs
Generating the value separately on each send is the usual failure. Two random numbers created in two places will never match, so the setup looks correct in a diagram and duplicates everything in practice. Using a timestamp alone fails for the same reason, because the browser and the server rarely record the same moment.
The other frequent one is a value that repeats. A fixed string, or an ID derived only from the product, means many genuine conversions share an identifier and get collapsed into one, so the platform now under-counts instead. And on single-page sites, an ID generated once at page load is reused if the visitor submits twice.
How to get it right
Create the identifier once, at the point the action occurs, and pass it to whichever send happens second — usually the browser passes it to the server, or the server writes it into the confirmation page. Reuse the order or lead reference where you already have one, and keep the event name identical on both copies.
Then verify rather than assume. Both Conversions API setups and server-side tag containers report whether events were matched, and that report is the only proof deduplication is working. Re-check after any website change, because a rebuild that drops the identifier from the confirmation page breaks it silently — this is a standard item in a proper tracking review.
