How to check Meta Pixel and CAPI deduplication is working: proving the event_id matches.
- “how to check Meta Pixel and CAPI deduplication”
- “how to check event_id matches Meta”
- “verify Meta event deduplication”
- “Meta Events Manager deduplication tab”
- “Rate of Events Deduplicated Meta”
- “check event_id Pixel and Conversions API”
- “Meta CAPI deduplication test”
- “is my Conversions API deduplicating”
Baris Asa, tracking consultant, London · Last checked: November 2026
To check Meta is deduplicating, prove three things for one test order: the browser copy and the server copy carry the sameevent_id, they carry the same event name, and the Event Deduplication tab in Events Manager shows a deduplication rate for that event without a low-rate warning. Matching ids on their own prove nothing until the second and third are true.
This note is for the case where deduplication is already set up and you want evidence it works. If purchases are already being counted twice and you need the cause, start with Meta CAPI and Pixel double counting purchases. If you are still choosing how to send server events, see Meta Conversions API setup options compared. The fault is on file as F-102 .
What does Meta need to see to merge two copies?
Meta’s rule has four parts (Meta: handling duplicate events).
- The Pixel’s
eventIDmust match the Conversions APIevent_id. - The Pixel’s event must match the Conversions API
event_name. - Both go to the same Pixel ID.
- The second copy arrives within 48 hours of the first copy with that
event_id.
Meta says it generally prefers the copy it received first and discards the later one. So your check is a comparison of two requests, then a look at what Meta did with them.
Step 1: what does the browser send?
Open the shop in a browser with DevTools, Network tab, Preserve log on. Place a test order. Filter the requests on /tr.
Find the Purchase hit. Read two parameters: ev is the event name, and eid is the event ID. Write both down exactly, including case and any prefix. Also note the id parameter, which is the Pixel ID.
Three things to watch for:
- No
eidat all. The Pixel is sending no id. Nothing can merge. - Two Purchase hits for one order. That is two browser senders, not a deduplication problem. Meta documents merging a browser copy with a server copy, and says nothing is deduplicated when only one source is used. Switch one sender off. This is the case in F-101 .
- An
eidthat is empty or different on a slow page. The Pixel tag fired before the order number reached the data layer. That is timing, filed as F-103 .
Step 2: what does the server send?
Read the server request for the same order, from wherever you can see it:
- Server-side GTM: Preview mode, open the Meta tag for the Purchase, and read
event_idandevent_namein the outgoing request. - A plugin or app: its event log or debug screen, if it has one.
- Your own code: your server log of the request body.
Write down event_id and event_name. Now lay the two readings side by side:
| Browser | Server | Must be | |
|---|---|---|---|
| Event name | ev | event_name | The same word, same case |
| Event ID | eid | event_id | The same string |
| Pixel ID | id | the Pixel/dataset in the request | The same |
Compare as strings, character by character. 10482 and order_10482 are different. Meta doesn’t document how it handles id types, so send the same string on both sides and don’t rely on coercion.
Step 3: what does Meta say it did?
Meta gives you two screens. Use both, and know what each one proves.
Test Events. In Events Manager, Data Sources, your Pixel, Test Events. Send the server request with the test_event_code Meta shows you, and your server events appear in that window once the request is sent. Two cautions from Meta’s own page: the parameter is for testing only and must be removed from the production payload, and test events still flow into Events Manager and can be used for targeting and ads measurement. So place the test order once, then take the code out (Meta: using the API).
The Event Deduplication tab. Events Manager, your Pixel, pick the event, Event Details, Event Deduplication. Meta says this shows the Rate of Events Deduplicated per source, that higher percentages are better, and that a warning appears when the rate is too low. It publishes no target rate (Meta: verifying your setup).
Give it normal order volume. One test order is enough to read the ids, but a rate needs a few days of real orders to mean anything.
Which readings prove nothing?
- The Overview total. Meta says it shows events received before deduplication. A healthy setup that sends every purchase twice, once from each side, looks doubled there.
- Seeing the event arrive from both Connection Methods. Connection Method tells you which channel sent the event. Both channels arriving is expected. It says nothing about whether they merged.
- A good Event Match Quality score. It measures how well Meta can match an event to a person, not whether two copies became one. See Meta Event Match Quality low.
- Matching ids on one test order. If each side generates its own id (a counter, a random value, a timestamp), the two can agree on a quiet test and diverge under real traffic. Place two orders and compare each.
When the ids match but the rate is low
If the two readings agree and the tab still warns, work down this list.
- Different Pixel. The server sends to a dataset that isn’t the Pixel the browser uses, for example an old Pixel left in the theme or an agency’s Pixel.
- The id is made up per order by each side. It matches on your test and drifts in production, for instance a timestamp rounded differently.
- The server copy is late. Meta only merges copies within 48 hours of the first. A queue that retries or a batch that sends purchases days later falls outside the window.
- Event name drift.
Purchasefrom the browser,purchasefrom the server. - Only some orders are affected. That is usually the data-layer timing break from step 1, not a deduplication setting.
Meta also describes a fallback that matches on the event name plus fbp or external_id. It says this generally works only when the browser event arrives first, so don’t rely on it as proof.
What does a pass look like?
For one closed week:
- Two test orders, each with the same
eidandevent_id, the same name, the same Pixel ID. - The Event Deduplication tab shows no low-rate warning for Purchase.
- No low-rate warning after those days of real orders. Don’t use Ads Manager against shop orders as the pass test: attribution makes that gap unreliable.
Sales that were counted twice before the fix stay counted. The clean reading starts at the first clean week. If you would rather someone else ran the three steps and handed back the readings, that is a job for a Meta CAPI consultant.
Common questions
How do I check that Meta is deduplicating my Pixel and Conversions API events?
Read the event_id and event name from the browser request and from the server request for one test order, and confirm they are identical. Then open Events Manager, your Pixel, the event, Event Details, and the Event Deduplication tab. Meta shows a Rate of Events Deduplicated per source and a warning when the rate is too low.
What is a good Rate of Events Deduplicated?
Meta only says higher is better and shows a warning when the rate is too low. It publishes no target figure. Judge it alongside the ids: if both copies carry the same event_id and name, a low rate usually points to a different Pixel, a server copy later than 48 hours, or events that only one source sends.
Do the browser and server events need the same event name as well as the same ID?
Yes. Meta says a Pixel’s eventID must match the Conversions API event_id, and the Pixel event must match the event_name. Purchase from one side and purchase from the other are different names.
Why does the Events Manager Overview show roughly double my events?
Meta says the Overview shows the events it received before they were deduplicated. If every purchase is sent from the browser and the server, the Overview looks doubled even when deduplication works.
How long does Meta wait for the second copy?
Meta deduplicates events received within 48 hours of the first event with that event_id. A server copy that arrives later than that is counted as a separate event.
Sources
Read on 6 November 2026:
- Meta for Developers, Conversions API: handling duplicate Pixel and Conversions API events (
eventIDandevent_id, event andevent_name, 48 hours from the first event, first-received copy generally preferred,fbp/external_idfallback and its limits, no deduplication within one source) - Meta for Developers, Conversions API: verifying your setup (20 minutes, Overview counts before deduplication, Connection Method, Event Deduplication tab, Rate of Events Deduplicated and its warning)
- Meta for Developers, Conversions API: using the API (Test Events location,
test_event_code, remove before production, test events still used for targeting and measurement)