Meta merges a browser Pixel event and a Conversions API (CAPI) event when both have the same event name and the same event ID and reach the same Pixel within 48 hours. So create one ID per event in the page (for example view_item_k3x9…, and purchase_<order number> for orders), give it to the Pixel as eventID, and send the same value to your server so the CAPI tag sends it as event_id. In our tests Meta then marked the second copy "Deduplicated", but only after a delay, and its deduplication report said "Not available" until there was more volume.
TagLinc hosts server-side GTM, so we are not neutral. The results below come from our own tests on 13 and 14 November 2026, with a test Pixel and a Test Events code, on a WooCommerce test store and a Shopify development store, each sending through a TagLinc server container. For the basics of the Conversions API itself, start with Meta Conversions API with server-side GTM.
Meta's rule in one paragraph
Meta's documentation recommends deduplicating by event ID and event name: the Pixel's eventID must equal the server's event_id, and the Pixel's event name must equal the server's event_name. If both copies reach the same Pixel within 48 hours of the first one, Meta keeps one and discards the other. When they arrive within about five minutes of each other, Meta says it favours the browser copy. The event ID can be any unique string; Meta suggests the order or transaction number for purchases and a random value for events without one, as long as both sides send the same value.
There is a second method, by fbp or external_id, but Meta says it generally only works when the browser event arrives first. Use event IDs.
Give every event one ID, in the page
The ID has to be made once, before either copy is sent, and then shared. The page is the only place both copies start from:
- Put the ID in the dataLayer push. Every ecommerce push gets an
event_id:view_item_,add_to_cart_,begin_checkout_plus a random part, andpurchase_plus the order number for orders (stable if the thank-you page is reloaded). - Web container: a Data Layer Variable for
event_id. Use it as Event ID in the Meta Pixel tag, and add it as anevent_idevent parameter on your GA4 event tag. The GA4 request carries it to your server container. - Server container: the Meta Conversions API tag takes
event_idfrom the event data the GA4 client built, and the event name from the same event, mapped to Meta's names (view_item→ ViewContent,add_to_cart→ AddToCart,purchase→ Purchase). - PageView: page views usually have no dataLayer push. Make one ID per page (a small Custom JavaScript variable that creates it once and reuses it) and use it for both the Pixel PageView and the Google tag's page_view.
On Shopify the checkout runs in Shopify's sandbox, where the Meta Pixel cannot load from GTM, so checkout and purchase events go to Meta from the server only. There is no browser copy to merge, so nothing is counted twice; only the storefront events (page view, product view, add to cart) need the pairing.
What Events Manager showed in our tests
| Test | What we saw |
|---|---|
| WooCommerce, every step from product view to purchase (13 Nov) | Each action arrived twice in Test Events, once from the browser and once from the server, with the same name and the same ID (for example add_to_cart_1bvrt2o1blm6m7). Test order #33 arrived as purchase_33 from the Pixel and the server: Meta marked the extra copies "Deduplicated". |
| Shopify storefront, page view, product view, add to cart (14 Nov) | Both copies arrived with matching names and IDs. Right after, Meta listed each pair as two separate "Processed" rows. About 45 minutes later it had grouped one pair, with the server copy marked "Deduplicated"; other pairs were still ungrouped at that point. |
| Event details, low volume (14 Nov) | "Total browser events received 15, total server events received 2" and "Event deduplication: Not available": the report needs more traffic before it shows a rate. |
Two lessons. First, do not judge in the first hour: two "Processed" rows straight after a test do not mean deduplication failed. Second, the dedup report on a quiet Pixel stays empty; check the IDs yourself instead (below).
Check the IDs yourself in five minutes
- Open a product page with developer tools. In the Console, type
dataLayerand find theview_itempush: note itsevent_id. - In Network, filter
facebook.com/tr: the Pixel request for ViewContent carries the same value aseid. - Filter
collect: the GA4 request to your tracking domain carries it as an event parameter (ep.event_id). - In Events Manager → Test Events (with your test code on the server tag), the browser and server rows show the same Event ID.
- Remove the test code from the server tag afterwards, or your live events keep going to Test Events.
Why deduplication fails
AddToCart, the server sends a custom add_to_cart. Map every event to Meta's standard name on both sides.Deduplication is not match quality
Merging copies stops double counting; it does not make Meta better at matching events to people. In our Shopify test the server copy carried only the IP address and browser: no fbp, no hashed email. That is low match quality, and it is fixed separately: see Meta Event Match Quality: how to raise your EMQ score.
We check your setup for free and send you what to fix.
FAQ
No. Meta says any unique string works, as long as the browser and the server send the same one for the same event. We use the event name plus a random part, and the order number for purchases.
It counts once, on its own. That is the case for Shopify checkout events and for visitors whose browser blocks the Pixel.
Events show in Events Manager within about 20 minutes, but in our test the grouping took about 45 minutes and was still incomplete then. Give it a few hours, and use Test Events for a quick check of the IDs.
Generally the first one received; when both arrive within about five minutes, Meta says it favours the browser copy.