TRACKING SPECIALIST — BARIS ASA · LONDON
Synctag
Start the check → TRACKING SPECIALIST — BARIS ASA · LONDON

READING ROOM › FIELD NOTE › NOTE-012

NOTE-012 · FIELD NOTE FILED 08 NOV 2026 · REVISED 08 NOV 2026 F-102

What a Meta Conversions API setup should hand back: events, match keys, dedup evidence.

ABSTRACT — Nobody ranks Meta CAPI consultants, so judge one by what the setup hands back: a list of everything that sends purchases to Meta, an event map, a match-key table with Event Match Quality before and after, deduplication proof on a real test order, a refused-consent test, and a record of whose account holds the Pixel and the access token.

Baris Asa, Synctag, tracking specialist, London · Last checked: November 2026

Nobody ranks Meta Conversions API consultants. Judge one by what the setup hands back. Six things: a list of everything that sends purchases to Meta, an event map, a match-key table with Event Match Quality before and after, deduplication proof on one real test order, a test with the cookie banner refused, and a record of whose account holds the Pixel and the access token.

Synctag is my London practice: I set up and repair the Meta Conversions API (CAPI) for UK online shops and the agencies behind them, and every setup ends with this hand-back. I have no partnership or commission with any CAPI tool or host. CAPI is the route that sends a sale to Meta from your server as well as from the shopper’s browser.

How can you tell if your setup handed this back?

Two checks in your own Meta account:

  • Two routes, not one or three. Events Manager, Data sources, your Pixel, Overview. Under Connection Method, Purchase should show a browser channel and a server channel (Meta: verifying your setup).
  • No deduplication warning. Click Event Details next to Purchase, then the Event Deduplication tab. Meta shows a warning when the Rate of Events Deduplicated is too low (same guide).

Two server senders still show as one server channel, so only the sender list (piece 1) catches a third. Neither check proves the setup is right. They tell you whether to ask. If Meta reports more purchases than you took, see Meta CAPI and Pixel double counting purchases. To run the event_id test yourself, see how to check deduplication is working. This note covers what a finished setup should give you.

What should a Meta CAPI setup hand back?

Six pieces. If one is missing, ask for it before the job is signed off.

PieceThe question it answers
1. Sender listWhat sends purchases to Meta?
2. Event mapWhich events go by which route, keyed how?
3. Match-key tableWhich customer details reach Meta, and did matching improve?
4. Dedup evidenceDoes one order count once?
5. Consent testDoes a refusal stop the server copy too?
6. Access recordWhose account holds it, and who can switch it off?

1. Sender list. Every source that sends Purchase to Meta: the Pixel in the theme, the Shopify app, Google Tag Manager, a Gateway, a server container, a plugin. The target is two: the Pixel and one server route. A third sender is where double counting usually starts.

2. Event map. One row per event (Purchase, AddToCart, InitiateCheckout and so on): browser or server or both, the event name on each side, and where the shared event ID comes from. Meta’s recommended method merges two copies when the Pixel’s eventID matches the Conversions API’s event_id, the event names match, and both arrive within 48 hours (Meta: deduplication). For Purchase the ID is usually built from the order number, so it is the same on both sides.

3. Match-key table. Which customer details the server event carries, and in what form. Email and phone are SHA-256 hashed (one-way encoded); email trimmed and lowercased first, phone as digits with the country code, so a UK number starts 44, not 07. Browser IP, user agent and Meta’s fbp and fbc cookies are sent unhashed (Meta: customer information parameters). Next to the table: Event Match Quality for Purchase before and after, as dated screenshots. Meta scores it 1 to 10 and typically recommends 6.0 or higher (verifying your setup). Why it moves: Event Match Quality is low.

4. Dedup evidence. One real test order, followed end to end: the order number, the browser copy and the server copy side by side with the same event name and event ID, as seen in Test Events. Then, once real orders have flowed, the Event Deduplication tab with no warning. The test order proves the keys match. The tab proves it holds for real traffic.

5. Consent test. The server copy is sent by your server, not the shopper’s browser, so the banner’s choice has to be passed to it. My standard for UK shops is that a refusal stops the server copy as well as the Pixel, and the hand-back shows it: banner accepted, both copies arrive; banner refused, neither does. How the banner itself should be wired is a separate job: Consent Mode v2 audit.

6. Access record. Where the Pixel lives, who has access to it, where the access token was made and where it is stored. More on this below.

Two smaller items belong with the six. A change log: what was switched off, what was added, each GTM version named and dated, so any change can be reversed. And a line for what wasn’t set up or tested, such as a second store, an app, or offline sales. Not set up is on record as not set up.

What does dedup evidence look like?

The finding in this exhibit: the server copy of Purchase carried the bare order number, the Pixel copy carried a prefix, so Meta treated them as two sales.

Exhibit: one test order, as read in Test Events. Faulty cells in bold.

FieldBrowser (Pixel)Server (CAPI)Match?
Event namePurchasePurchaseyes
Event IDorder_10421042no
Value64.0064.00yes
CurrencyGBPGBPyes

Reconstructed · client data anonymised

Everything else matched, which is why nobody caught it from the Overview count alone. Each copy arrived, so both counted. Filed as F-102 .

Fix: build the event ID once, in the same format, and pass the same value to both routes. Re-test: the same table, every row “yes”, then the Event Deduplication tab a few days later.

Whose account should the setup stay in?

Yours. Three things to check in the access record.

  • The Pixel. It stays in your business portfolio (Meta’s name for the business account that owns your ad accounts and Pixels). A consultant or agency gets partner access to it. If they built it in their own portfolio, you don’t own your own data source.
  • The access token. The server needs a token to send to Meta. Meta offers two ways to make one: from your Pixel’s settings in Events Manager, which only people with developer privileges for the business can see, or from a system user in your business settings. Linking it through Manage Integrations creates a Conversions API app and system user for you (Meta: get started). Either way, it should be made in your business, and the record should say where it is stored, so it can be replaced.
  • The host. If the server route runs on a Gateway or a hosted server container, the hosting account and the bill should be in your name. Who maintains what on each hosting route: server-side GTM hosting.

The record also confirms the Test Events code is out of the live setup. Meta says to remove it before production, and notes that test events still flow into Events Manager and are used for targeting and ads measurement (Meta: using the API).

What does the hand-back not cover?

  • A choice of tool. Shopify app, Meta’s one-click setup, Gateway, Stape or server-side GTM: the choice comes before the setup. Compared in Meta CAPI setup options.
  • An audit of Google and GA4. This is Meta only. Checking every platform against your orders is a tracking audit.
  • A promise about ad results. A clean setup sends each sale once, with the details Meta matches on. What Meta does with it is Meta’s.

When should it be checked again?

Whenever something new sends to Meta: a new app, a theme change, a checkout change, a platform move. Each can add a third sender or change the event ID on one side. Re-run the dedup evidence with one test order and compare it with the hand-back. Meta’s Event Freshness tab shows how late server events arrive, on a scale from real time to weekly (verifying your setup); a server route that slips to hours late is worth a look, and past 48 hours Meta no longer merges the two copies (Meta: deduplication).

The questions to ask a Meta CAPI consultant before hiring are on the Meta Conversions API service page. The hand-back is how you check the answers afterwards.

Common questions

How do I know my Meta Conversions API setup was done properly?

Ask for the hand-back. In Events Manager, open your Pixel, then Overview: Purchase should list both browser and server under Connection Method. Event Details for Purchase, Event Deduplication tab, should show no low-rate warning. If nobody can show you one test order matching on both sides, it hasn’t been proven.

Who should own the Conversions API access token?

Your business. Generate the token from your own Pixel’s settings in Events Manager, or from a system user in your own business settings, and keep a note of where it is stored. A token tied to a consultant’s app or business means you depend on them to rotate or revoke it.

Do test orders sent during a Meta CAPI setup count in my reports?

Meta says events sent with a Test Events code still flow into Events Manager and are used for targeting and ads measurement. Keep test orders few, note their order numbers in the hand-back, and make sure the code is removed before go-live.

Sources

Read on 8 November 2026:

  • Meta for Developers, Conversions API: verifying your setup (Data sources and Overview; Connection Method; Event Details and Event Deduplication tab; Rate of Events Deduplicated and its warning; Event Match Quality 1 to 10, 6.0 or higher typically recommended; Event Freshness tab)
  • Meta for Developers, Conversions API: handling duplicate Pixel and Conversions API events (eventID matches event_id; event matches event_name; 48 hours)
  • Meta for Developers, Conversions API: customer information parameters (email and phone hashed; email trimmed and lowercased; phone with country code and no leading zero; IP, user agent, fbp, fbc not hashed)
  • Meta for Developers, Conversions API: get started (Generate access token in the Pixel’s settings, visible to users with developer privileges; Manage Integrations creates a Conversions API app and system user; system user token from business settings)
  • Meta for Developers, Conversions API: using the API (Test Events; remove test_event_code before production; test events used for targeting and measurement)

— B.A., filed from London

CONNECTED RECORDS F-102

REVISION HISTORY — revised 08 NOV 2026.

← back to the reading room