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

READING ROOM › FIELD NOTE › NOTE-026

NOTE-026 · FIELD NOTE FILED 12 NOV 2026 · REVISED 12 NOV 2026 F-101 F-103 F-201 F-501 F-603

GTM container audit: what gets checked and what you get back.

ABSTRACT — A Google Tag Manager audit reads one container without changing it: tags, triggers, variables, the dataLayer, custom scripts, consent, access and versions. You get back an inventory with a keep, fix, remove or add call on each item, a findings register with evidence and severity, an access list and a list of what wasn't checked.

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

When GA4 and your store disagree on orders, the Google Tag Manager container is the first place to look. A GTM container audit reads that container and changes nothing: tags, triggers, variables, the dataLayer, custom scripts, consent, access and versions. You get back an inventory with a keep, fix, remove or add call on every item, a findings register with evidence for each fault, an access list, and a list of what wasn’t checked.

Every finding names the tag, trigger or variable, the container version it was read in, and the evidence, so a developer can act on it without ringing me.

This note covers the container only. If the gap might sit in GA4 settings, the ad platforms or the store itself, that’s a tracking audit, and its cross-platform report is described in what a tracking audit should hand back.

What does a Google Tag Manager audit check?

Thirteen areas: install, coverage, tags, missing tags, triggers, variables, the dataLayer, custom scripts, consent, naming, access, versions and, where there is one, the server container. The dataLayer is the order and page detail your site hands to the tags. Each area produces rows in the inventory or the findings register, never a single “checked”.

AreaWhat gets checked
InstallSnippet high in the head on every page type; a second container; Google tags hard-coded in the theme
CoveragePages Google has seen without the tag (Tag coverage summary), checked against the sitemap
TagsWhat each sends and where; paused tags; old Universal Analytics tags
Missing tagsA lead event, begin_checkout, Google Ads click data on landing pages
TriggersPurchase tags on page load instead of the order event; triggers firing twice
VariablesWhere each value comes from; variables used by nothing
dataLayerOrder ID, value and products on purchase, there before the tag reads them
Custom scriptsEach Custom HTML tag: who added it, what it loads, whether it slows the page
ConsentThe setting on each tag; the default runs before any Google tag
NamingTag names, folders and notes a stranger could follow
AccessWho has Publish and Administrator, including people who have left
VersionsNamed versions with notes; drafts (workspaces) left out of date
Server containerWhat the web container sends it, when access is given

Two notes on the table. A container that loads a Google tag carrying your Google Ads destination (AW-) on every page doesn’t also need a Conversion Linker (the tag that keeps the Google Ads click so a later sale is credited); a GA4-only Google tag doesn’t count, and a container with neither on its landing pages needs one. And the consent default can be set inside GTM on the Consent Initialization trigger, which Google fires before every other tag, or in page code above the GTM snippet. Either is fine if it runs first.

What does a GTM audit hand back?

Four things: an inventory with a keep, fix, remove or add call on every tag, trigger and variable; a findings register with evidence, severity, fix and re-test for each fault; an access list; and a list of what wasn’t checked.

1. The inventory.

ItemWhat’s wrongCall
Tag · GA4 purchaseFires on page load, products emptyFix
Tag · UA all pagesUniversal Analytics no longer processes dataRemove
Trigger · All clicksUsed by nothingRemove
Tag · Meta PurchaseEvent ID matches the server copyKeep
(none)No Google tag or Conversion Linker on landing pagesAdd
Illustrative rows, not from a client container.

Anything marked remove is checked first for what else depends on it.

2. The findings register. One row per fault, holding:

  • the GTM item and the container version it was read in
  • the evidence: a network hit, a screenshot, an order number
  • what it distorts, and the severity
  • the fix, and how to re-test it

Severity follows what the fault distorts:

  1. Wrong numbers that ad platforms bid on, and personal data reaching Google.
  2. Tags firing against consent, or before the consent default.
  3. Missing report detail. Purchases without products, parameters that never arrive.
  4. Housekeeping. Names, unused items, container size (Google suggests optimising once the Size indicator on the Versions page passes 70%).

3. The access list. Every user on the account and the container, their permission level, and whether they still need it.

4. What wasn’t checked. A server container nobody gave access to, a payment method there was no way to test, a market on another domain. On record as unchecked, not as clean.

Which faults should an audit look for first?

Five, each with its own fault record:

  • The purchase sent twice. A Google tag hard-coded in the theme and a GTM tag both send the order. GA4 and Google Ads drop a repeat with the same transaction ID, so it counts twice when the ID is missing or differs, or when each copy feeds a different conversion action. Fault F-101; worked through in GA4 purchase event fires twice.
  • The purchase sent before the order data is there. The tag fires on page load, so value or products arrive empty on some orders. Fault F-103; the dataLayer side is in the GA4 ecommerce dataLayer in GTM.
  • Consent left unset. Set right, Consent Mode lets Google model back part of the conversions from visitors who decline, once traffic clears Google’s thresholds. Left unset, those are lost, or tags fire against the visitor’s choice. The usual cause is tags in Google’s “Consent Not Configured” list, or a default that runs after other tags. Fault F-201; the order of setup is in Consent Mode v2 in GTM, and a full consent check is a consent mode audit.
  • Email addresses in URLs. A form or sign-up redirect puts the customer’s email in the page URL. GA4 collects the URL as page_location, and any tag that reads the URL passes it on. The fix is at the source, with GA4’s Data redaction as a backstop on web streams. Fault F-603.
  • Nobody knows what half the tags do. No names, no version notes, no owner. Fault F-501. The audit flags it; writing down why each tag exists, so someone else can run the container, is tracking documentation.

Read Google’s Tag Diagnostics first (Google tag settings › Tag quality › View diagnostics). It flags some consent mode and transaction ID problems, though some diagnostics are still rolling out and the transaction ID checks only cover particular setups. It reports what the Google tag sends, not how GTM is configured, so it’s a starting point, not the audit.

How is a container audit done?

Read-only, in three passes:

  1. Inventory. The live version is exported as a JSON file (Admin › Export Container) and filed with the report; if the auditor’s access can’t export it, you do. Every tag, trigger and variable is listed with what it sends, when it fires and what depends on it.
  2. Test. Each tag is checked on the live site with Tag Assistant and the browser’s network panel, to see what actually leaves the page: page views, add to cart, checkout, and a real order or form where those are tracked.
  3. Settings. Consent, versions and workspaces are read in the GTM interface. Consent Overview only shows if someone has switched it on in Admin › Container Settings; if it’s off, each tag’s consent settings are read one by one. The Tag coverage summary (Google tags › your tag › Admin › Tag coverage) lists the pages Google has detected, up to 10,000, plus any URLs added by hand or by CSV, so the sitemap goes in as a CSV. A page missing from it is a lead, not proof.

Can you check your own container first?

Yes, without changing anything. Step 1 needs only GA4 and your store; for steps 2 to 4, ask whoever holds GTM access to read them to you.

  1. Your own numbers. Compare GA4 purchases with the store’s paid orders (test and cancelled orders left out) for 7 complete days ending at least two days ago, since GA4 can take a day or two to process. A small gap is normal (ad blockers, declined consent, test orders). GA4 at double, or an exact multiple, points at a purchase sent twice (F-101). GA4 well below the store points at consent or a purchase tag firing too early (F-201, F-103).
  2. Admin › User Management (needs an account Administrator). Count who has Publish. Someone who has left the business or the agency is a finding.
  3. Versions. Read the last five version names. “Version 47” with no description means nobody can roll back with confidence.
  4. Tags. Look for anything named “test”, “old”, “copy” or “UA”. Each is a question someone should be able to answer.

Any of those is a finding. An audit puts it on record with the evidence.

What doesn’t a GTM container audit cover?

Anything outside the container: GA4 property settings, Google Ads conversion actions, Meta’s Events Manager, apps installed on the store platform and the store’s own order numbers.

The audit also doesn’t fix anything. The clean-up is a separate job, which I can do or hand to your developer: in a workspace, tested in Preview, published as a named version by whoever holds Publish. If a release goes wrong, an earlier version can be published again from the Versions page.

What access does a GTM audit need?

Read access to the container. Google’s Read level lets a user browse the tags, triggers and variables without making changes. Edit, Approve and Publish aren’t needed, so the audit stays read-only. The user list sits with account Administrators: for the access list, an Administrator sends a screenshot or export of Admin › User Management.

If the work is GTM-only, this is the container review I run as a Google Tag Manager consultant, for brands and for agencies working in their clients’ containers. Send the container ID (GTM-XXXXXXX) and what looks wrong; I reply with what the audit would cover. Scope is agreed on a short call. You provide Read access and a way to place a test order or submit a test form.

Common questions

Can an automated GTM audit tool replace a manual audit?

Partly. A tool can list tags, spot paused or unused items and flag naming problems. It can’t tell you what a tag is for, whether a purchase reaches GA4 with the right value on a real order, or which finding costs you most. Use a tool’s output as a starting inventory, not as the report.

How long does a GTM audit take?

It depends on the container: the number of tags, how many sites and markets it serves, and whether there’s a server-side container. The date is agreed on the scoping call, once the container has been seen.

Does a GTM audit change the container?

No. It records and reports. The clean-up is a separate job, done in a workspace and published as a named version, so it can be rolled back.

How often should a GTM container be audited?

When it changes hands or shape: a new agency or developer, a theme or checkout change, a new cookie banner, or before a move to server-side tracking. Otherwise once a year, before peak trading.

Is a GTM audit the same as a tracking audit?

No. A GTM audit stops at the container’s edge. A tracking audit follows the data on into GA4, Google Ads and Meta and checks it against your orders.

Sources

Read on 12 November 2026:

← back to the reading room