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

READING ROOM › FIELD NOTE › NOTE-014

NOTE-014 · FIELD NOTE FILED 08 NOV 2026 · REVISED 22 NOV 2026 F-302 F-101 F-301 F-205 F-204 F-102 F-104 F-201 F-603 F-602

Ecommerce tracking audit checklist: checkout, payment gateway, apps and server events.

ABSTRACT — What a detailed technical tracking audit for an online store checks, and how to tell a detailed one from a thin one: eight tests, each run with a real order, and one finding per fault with evidence someone else could open and repeat.

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

If GA4, Google Ads or Meta don’t match your store’s orders, a detailed audit finds out why with eight checks, each run on a real order. How detailed an agency’s or consultant’s audit is shows in its evidence: every finding comes with the order number, network request or screenshot behind it, so someone else could repeat the test. A tick is not evidence.
  1. Totals: do GA4, Google Ads and Meta agree with the store’s orders?
  2. Checkout: one purchase per order, with the right value
  3. Payment gateway: the sale keeps its source after PayPal or Klarna
  4. Apps: every app and snippet that sends its own tags
  5. Server events: browser and server copies counted once
  6. Conversion settings: what Google Ads and Meta bid on
  7. Consent: tags follow the banner on accept and reject
  8. Personal data: no email addresses or names in GA4

Check 1 needs only the store admin and the three platforms: start there. Between them, the eight cover the store’s own orders, the tags in the browser, the tag manager, the server and the settings inside each ad platform.

What does the audit hand back?

Each fault a check finds becomes one row in a findings register (a table of findings): the evidence, the number it distorts, how serious it is, who fixes it, the fix and how to confirm the fix worked. Around the register sit the totals table from check 1 and a list of what wasn’t checked. How to compare one auditor’s register with another’s is in what a tracking audit report should hand back.

Comparing agencies or consultants? Ask each for one finding as they would file it, with its order number and request.

Synctag’s tracking audit covers this ground read-only and files each finding against a fault code.

What do you need before you start?

  • Read access to GA4, Google Ads, Meta Events Manager and the store admin, and to Google Tag Manager (GTM) with Preview, so its tags can be watched firing. An audit changes no settings.
  • Real test orders, one per payment method. They are the one thing an audit adds: list their order numbers so they can be left out of reports, and refund them the same day. Some test gateways, such as Shopify’s Bogus Gateway, never leave the store, so they can’t test the return from the payment provider. Use the live checkout, or a sandbox that really redirects.
  • A landing link with test UTMs (the tags on a link that name its source), so you know what source each test order should carry.
  • Tag Assistant (Google’s tool for watching tags fire), Chrome’s network panel and GA4 DebugView (a live view of one test session) open during each order.

Start here: do the platforms agree with your orders?

This check needs no technical tools. Take one closed week. Set the store’s orders and revenue against GA4, Google Ads and Meta for the same days, by conversion date.

Evidence: one table, four rows, with the gap for each.

On the audits I’ve filed, a GA4 gap you can name is normal: close to your banner’s reject rate, plus a few per cent for ad blockers, unless GA4’s own consent modelling is switched on. Google Ads and Meta can add modelled conversions and credit the date of the click or ad view, so their gap can be smaller than the reject rate, or reversed. An exact multiple, or a gap nobody can name, usually points to one of the checks below. Value drifting from revenue is F-302 .

Before you read the gap, take out what shouldn’t be there: the test orders, refunds and cancellations, which GA4 keeps counting unless a refund event is sent, and visits from your own team (F-602 , excluding internal and bot traffic). Why the three platforms never show the same number: three purchase totals.

Checkout: is every order recorded once, with its value?

  • One purchase per order. Test: place an order and watch the network panel on the thank-you page. Evidence: one purchase request per platform, carrying the store’s order number as the transaction ID, never empty.
  • Reload. Test: reload the thank-you page. Evidence: GA4 ignores a second purchase from the same visitor with the same transaction ID (Google Analytics Help), and so does Google Ads within the same conversion action, as long as the tag sends the ID. Meta may not: a reload can send a second pixel Purchase unless the setup fires once per order, so check Events Manager for that order.
  • Value and currency. Test: open the purchase in DebugView. Evidence: transaction ID, items, value and currency present, and value equal to item price times quantity, with shipping and tax in their own fields.
  • Funnel steps. Test: one session from product page to purchase. Evidence: view_item, add_to_cart, begin_checkout and purchase each arrive once.

Those field rules are Google’s: transaction ID and items are required, currency is required when value is set, and value leaves out shipping and tax (event reference). If Google Ads or Meta is sent a different basis, write the difference down so the totals check reads right.

A purchase sent twice is F-101 (purchase event fires twice). Missing steps or item detail is F-301 . The transaction ID must be unique to each order: a hardcoded or reused one makes Google Ads undercount (Google Ads help).

Payment gateway: does the sale keep its source?

Customers leave for PayPal, Klarna or a 3-D Secure screen (the bank’s card check) and come back. If the return starts a new session, the payment provider gets the credit for the sale.

  1. Arrive on the test UTM link, pay with each method, finish the order.
  2. In GA4, find each order by transaction ID and read its session source.
  3. Open the referral report for the last 30 days and look for payment domains with purchase revenue.

Evidence: test order numbers, payment method, and the source GA4 gave each. Every row should show the test UTMs.

A payment domain as source is F-205 . A paid click that arrives as (direct), because the click ID or UTMs were stripped on the way in, is F-204 . The fix for the first is GA4’s unwanted referrals list (Google’s guide). It doesn’t correct sessions already recorded, so the audit notes the date.

Apps: what does each one send on its own?

Store apps, theme code and GTM can each send a purchase to the same platform. A common version: an old Google Ads snippet left in the theme after GTM went in. Both fire, and nobody owns the old one.

Build an inventory, one row per sender:

SenderWhere it livesSends
e.g. Meta sales channel appShopify, Customer eventsMeta: all standard events
e.g. GA4 tagGTM containerGA4: funnel and purchase
e.g. old Google Ads snippetTheme codeGoogle Ads: purchase

Evidence: the inventory, with the requests from one test order matched to its rows. Two senders with the same platform and event is a finding.

On Shopify, app and custom pixels sit under Settings, Customer events. They load on the storefront, checkout, thank-you and order status pages. In markets set to require consent (usually the UK and EEA), they run only once the visitor grants the permissions set on the pixel (Shopify help). What the common Shopify apps send: Shopify tracking apps: what each sends.

Server events: is each sale counted once across browser and server?

Many stores send purchases twice on purpose: from the browser and from their server, so that sales blocked in the browser still arrive. Each platform needs a shared key to keep one copy.

  • Meta: the pixel’s eventID must match the Conversions API (Meta’s server feed) event_id, and the event names must match. Within 48 hours Meta discards the later copy and generally keeps the one received first (Meta’s guide). Check one test order in both sources, then the Event Deduplication tab in Events Manager (Meta’s verification guide).
  • Google Ads: a browser tag and a server upload to the same conversion action should carry the same transaction ID. Google drops repeats only within one conversion action (Google Ads help).

Evidence: the browser and server copies of one order, side by side, with matching IDs. Missing or different IDs is F-102 . The Meta side in detail: check event_id matches.

Conversion settings: what are Google Ads and Meta bidding on?

This is the check closest to the money: the platforms bid on whatever they are told is a sale.

  • Google Ads. Test: Goals, Conversions, Summary. Evidence: a screenshot of the action list, showing which actions are Primary (the ones a campaign bids on, when they sit in its goals) and how each counts. A GA4-imported purchase and a Google Ads tag purchase are two actions; Google only drops repeat transaction IDs within one action, so if both are Primary in the same goal, each sale counts twice (F-104 , primary and secondary actions).
  • Meta. Test: the conversion event each active sales ad set optimises for (in Ads Manager, at ad set level), against Events Manager. Evidence: every sales ad set optimises on the deduplicated Purchase from the server events check.

When the banner is wired right, accepted visitors are measured in full, and Google can model some of the rest where the account meets its thresholds.

  1. Open Tag Assistant on a fresh session. In the Consent tab, the On-page Default column shows all four consent types set before any tag fires.
  2. Accept. The On-page Update column changes and the tags fire.
  3. In a new session, reject. Tags that need consent are blocked. In basic consent mode Google’s tags are blocked too; in advanced mode they send cookieless pings. Either passes if it matches what the store chose.
  4. Place a test order in each state.

Evidence: Consent tab screenshots for default, accept and reject, and the tag list for each (Google’s debugging guide). A default set late, or an update that never arrives, is F-201 . Step by step: how to check Consent Mode v2 is working.

Personal data: are emails or names reaching GA4?

In a GA4 exploration (or a customised Pages and screens report), use the dimension Page path + query string and screen class and filter for ”@” and “%40” (an @ written into a URL), and check event parameters for email addresses, phone numbers or names. Order confirmation and account pages first.

Evidence: the affected pages and parameters, with dates. Google’s policy forbids sending information that identifies a person, such as an email address, phone number or name (Google’s guidance). Any hit is F-603 .

What should the checklist leave out?

Anything not tested. A market, a payment method or a server container (the server-side tag setup) nobody gave access to goes on a separate list, as unchecked, not as clean.

A review of the GA4 property alone (streams, key events, filters, retention) is a different job, with no tag or checkout testing: a GA4 audit.

I run Synctag’s tracking audits myself, read-only, through these checks. I have no partnership, affiliate or commission arrangement with any app or platform named on this page.

Common questions

What should an ecommerce tracking audit check?

Eight things, tested with real orders: platform totals against the store’s orders; one purchase per order with transaction ID, value, currency and items; the sale keeping its source after the payment provider; every app or snippet that sends its own tags; browser and server copies counted once; which conversions Google Ads and Meta bid on; tags that follow the banner on accept and reject; and no email addresses or names reaching GA4.

How many test orders does a tracking audit need?

One per payment method, because each provider returns the customer differently, plus one with the banner accepted and one with it rejected. Use the live checkout, or a sandbox that really sends the buyer to the provider and back. Some test gateways, such as Shopify’s Bogus Gateway, never leave the store, so they can’t test the return. Note the order numbers and refund them the same day.

Can I run this checklist myself?

The first check, totals against orders, needs only the store admin and the three platforms. The rest needs Tag Assistant, Chrome’s network panel and GA4 DebugView, and access to wherever the tags are set up. If the gap is still unexplained after the checkout and payment checks, or nobody can open the server container, that is the point to bring someone in.

What’s the difference between an agency tracking audit and an independent one?

The checks shouldn’t differ: a detailed audit of an online store runs the same tests whoever does it. What differs is who runs them (one named person or a team), how much of the tracking they set up themselves, and who you go back to with questions. Neither is better by default, and many agencies audit well. Ask either for the order numbers and network requests behind each finding, so someone else could repeat the test. The wider choice for a brand is in independent tracking consultant or agency.

Sources

Read on 8 November 2026, re-checked 22 November 2026:

← back to the reading room