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

← The field manual

MANUAL · CH.01

How a tracking audit works: six stages, in order

Filed

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

A tracking audit tests what your tags record against something that isn’t a tag, usually the store’s orders. Nothing on the live site changes while it runs. It works in six stages, in a fixed order: agree the question, freeze the starting point, list everything that sends data, test real orders, match the totals with the store’s orders, then file each finding against a fault code.

What it leaves behind is a baseline: a dated record of what your tracking does today, which every later fix is measured against.

StageWhat it needsWhat it produces
1. The questionThe number that disagrees, and the source you trustOne written question, a period, what is out of scope
2. FreezeRead access, the live GTM versionA dated starting point
3. InventoryGTM, theme code, apps, any server containerOne row per thing that sends data
4. Live testsReal orders, Tag Assistant, DebugViewWhat each platform received, order by order
5. Order matchA closed week, the store’s order exportGA4 purchases set against orders, one by one
6. FileEverything aboveFindings by fault code, and the baseline

This chapter is the method inside an audit. The engagement around it (scoping call, access, findings call) is on the tracking audit page. What each test checks: the ecommerce tracking audit checklist. What the report hands back: the report note. Whether you need an audit at all: tracking audit vs GA4 review.

Stage 1: what question is the audit answering?

Stage 1 writes down one sentence before anyone asks for access: which number disagrees with which source, over which dates. “GA4 shows more purchases than Shopify since the theme change.” “Google Ads reports more conversions than we have orders.” The source has to be something that isn’t a tag: the store’s orders, or the CRM on a lead-gen site.

The question decides which platforms are in, which period gets matched and what is left out (a second market, the app, a subdomain nobody uses). Write the exclusions down. An audit without a question checks everything and ranks nothing.

If the honest question is “is any of it right?”, that is fine. Tracking that was installed and never checked against the store is filed as U-001 : a state, not a bug. It may be fine; nobody can say. The audit ends it either way, with a list of named faults or a dated record that says the numbers hold.

Stage 2: how is the starting point frozen?

Stage 2 records the Google Tag Manager (GTM) version that is live today and exports it, before anything is tested. In GTM, the Versions tab lists every saved version; the most recent one with a date in the Published column is the one live now. Export it (Admin, Export Container, choose the version, Download). If your access level doesn’t allow the export, the container owner can send the file. Write down the date and who owns tracking.

The reason is simple: a container that changes during the audit can’t be measured. So ask for a quiet period, with no changes to tags, theme, checkout or consent banner while the audit runs, or a note of any that can’t wait. A change halfway through moves the starting point, and the before-and-after is lost.

The audit runs on read access. GTM’s Read permission lets you view the container and browse tags, triggers and variables. GA4’s Viewer role shows settings and data. GTM’s preview mode shows only in the browser that opened it, or to someone sent the preview link, never to ordinary visitors. Publishing stays with whoever owns the container.

Stage 3: why is every sender listed before testing?

Stage 3 lists every tag, pixel and app that sends data, because a test order is only readable if you know who could have sent each request. On the audits I’ve filed, the second purchase is mostly not in GTM. It sits in a theme snippet, a hard-coded Google tag, a platform app such as Shopify’s Google & YouTube channel, or a server container (a second GTM container running on a server rather than in the browser) nobody mentioned.

The inventory has one row per sender: where it lives, what it sends, on which pages, and who added it if anyone knows. Read the GTM export, search the theme code for the Google tag (gtag) and the Meta pixel (fbq), list the installed apps, and open any server container.

This list is also the first draft of the documentation that F-501 says is missing.

Stage 4: how are real orders tested?

Stage 4 places real orders on the live checkout and records what each platform received. One order per payment method, because each provider sends the customer back differently, plus one with the cookie banner accepted and one rejected. Each starts from a landing link with test UTMs (the tags on a link that name its source), so you know which source the order should carry.

Watch each order with Tag Assistant (Google’s tool for watching tags fire), the browser’s network panel and GA4 DebugView (GA4’s live view of incoming events). Accept analytics cookies before reading DebugView; it doesn’t show events from visitors who declined. On Shopify, tags inside a custom pixel run in a sandbox that Tag Assistant’s troubleshooting doesn’t detect. Use the Shopify Pixel Helper or the network panel there.

For every order, write down which events fired, which sender sent them, the transaction_id (the order number your tags send with each purchase), value, currency and source. List the order numbers so they can be left out of reports, and refund them the same day. The eight things each test checks are in the checklist.

Nothing gets fixed at this stage, even when the fault is obvious. Write it down. The fix comes after the findings, against a starting point that hasn’t moved.

Stage 5: how are the totals matched with the store?

Stage 5 matches one closed week of GA4 purchases with the store’s orders, order by order. It comes after the tests because the test orders show which store identifier your tags send as transaction_id: order number, order ID or checkout token. That is the column to export.

Take one full week that ended at least two days ago. Google says GA4 processing can take 24 to 48 hours and reports may change in that time. The 30-day comparison on the tracking audit page is the quick self-check; the audit’s match is narrower and goes order by order.

Sort every order into four columns: matched; in the store but not GA4; in GA4 but not the store; test or cancelled, set aside. Then look for any transaction_id with more than one purchase. Google deduplicates purchases with the same transaction_id from the same user, so a second copy usually survives only with a different ID or from a different sender.

The reading you can check today:

  • GA4 purchases ÷ store orders, for the same days in the same timezone, with test and cancelled orders taken out.
  • Steadily below 100%, with the missing orders explained by declined cookies and blocked scripts: expected.
  • Above 100%, an exact multiple, or a step on a date you can name: a fault, and it gets a code.

An illustrative week: 190 orders in the store, 212 purchases in GA4, a ratio of 112%. The 22 extra purchases sit in the third column with checkout tokens instead of order numbers: two senders, F-101 .

Write down what Google Ads and Meta report for the same week, with the settings each used. They are not meant to equal the store. Which gaps are normal: GA4 revenue higher than Shopify and GA4 and Google Ads don’t match.

Stage 6: how does a finding get filed?

Stage 6 gives every gap traced to a cause a fault code, so the same fault reads the same way on every audit. Two purchases for one order is F-101 . A sale counted by both browser and server is F-102 . A purchase with no value or items is F-301 . A paid click filed as direct is F-204 . An email address in a page URL is F-603 .

Each finding carries evidence someone else can open again: an order number, a network request, a screenshot, an export. A gap that wasn’t traced is written as unexplained, with what was ruled out. It isn’t guessed. How findings are graded and laid out is in the report note.

What does the baseline do after the audit?

The baseline is the dated record of stages 3 to 5: the inventory, what each test order produced, the order match and the platform figures for the same week. Every fix is measured against it.

Repairs come after the findings, as named GTM versions published by whoever owns the container, so each one can be read later and rolled back. After each fix, the same test order runs again and the ratio should move the way the finding said it would. If it doesn’t, the finding was wrong, and the record shows it.

Repeat the order match monthly. Repeat the whole baseline after any change to checkout, theme, consent banner or purchase tags.

Where does the order usually go wrong?

  • Matching an open week. GA4 is still processing, so the gap moves while you read it.
  • Testing before the inventory. A second purchase arrives and nobody can say who sent it.
  • Fixing as you go. The starting point moves, and the audit can no longer show what changed.
  • Auditing one platform. Faults such as F-102 , F-104 or F-403 sit between GA4, Google Ads and Meta, not inside any one of them.

Common questions

How does a tracking audit work?

In six stages, in this order: write down the question, freeze the starting point with read-only access, list every tag, pixel and app that sends data, place real test orders while watching each tag fire, match a closed week of GA4 purchases with the store’s orders, then file each gap against a fault code. Nothing on the live site changes until the findings are written.

Which stage takes the longest?

Stages 3 to 5. The inventory grows with every app and snippet on the site, the test orders grow with every payment method and market, and the order match grows with the number of senders to untangle. Stages 1 and 2 are usually an afternoon. The whole audit usually takes one to two weeks from access to written findings.

Which period should a tracking audit compare?

One full week that ended at least two days ago. Google says GA4 processing can take 24 to 48 hours and reports may change in that time, so a week that is still open, or ended yesterday, gives a gap that moves while you read it.

Can I run a tracking audit myself?

Stages 1, 2 and 5 need only access and an order export. Stages 3 and 4 need Tag Assistant, the browser’s network panel, GA4 DebugView and someone who can read a GTM container and the site’s theme code. The U-001 page sets out a smaller version, a tracking baseline, that one person can build in a day or two. If nobody in-house can read the container and theme code, that is the part the tracking audit covers.

Also filed as

  • how does a tracking audit work
  • tracking audit process
  • steps of a tracking audit
  • how to run a tracking audit
  • what happens during a tracking audit
  • tracking audit methodology

Faults explained here

Sources

  1. Tag Manager: Export and import containers (Admin, Export Container, choose a version or workspace, Download; checked 9 Nov 2026)
  2. Tag Manager: Versions (a version is a snapshot of a container configuration; every publish records a version; entries with a date in the Published column have been live, the most recent is the one live now)
  3. Tag Manager: User permissions (Read: view the container and browse tags, triggers and variables; Publish: full rights including publishing)
  4. Tag Manager: Preview and debug containers (the debug interface appears only in the browser that enabled preview, or for users sent a shared preview link; not visible to regular website visitors)
  5. Analytics: Access and data-restriction management (Viewer: can see settings and data)
  6. Analytics: Data freshness (processing can take 24 to 48 hours; reports may change in that time)
  7. Analytics: Monitor events in DebugView (events from users who decline analytics cookies do not appear)
  8. Analytics: Measure ecommerce, transaction_id (the transaction ID deduplicates transactions from the same user; the same ID should not be used across different users)
  9. Shopify: Create a Google Tag Manager custom pixel (Tag Assistant's troubleshooting doesn't detect Google tags in a custom pixel; test with the Shopify Pixel Helper; checked 9 Nov 2026)
WHO RUNS IT

Baris Asa, a London tracking specialist working as Synctag, runs this audit for ecommerce brands and the agencies behind them.

Tracking audit →