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

READING ROOM › FIELD NOTE › NOTE-008

NOTE-008 · FIELD NOTE FILED 06 NOV 2026 · REVISED 22 NOV 2026 F-101 F-102 F-201 F-205 F-603 F-601 F-501

What a tracking audit checks, and what the report should hand back.

ABSTRACT — A tracking audit checks five things: the tags installed, what each event sends, platform totals against your orders, what ad platforms bid on, and consent. Judge an audit by its report. A detailed technical audit for an online store hands back a findings register: every fault with evidence you can open, the number it distorts, a severity, an owner, the fix and a re-test. Plus a check against your orders and a list of what wasn't checked.

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

A tracking audit checks that the numbers in GA4, Google Ads and Meta match what happened on your site. It usually looks at five things: the tags installed, what each event sends, platform totals against your real orders, what the ad platforms count, and consent. It hands back a list of faults, not a score.

What does a tracking audit check?

This is about web analytics and ad-platform tracking, not a tracking plan review or a finance audit trail. Five areas:

AreaThe question it answersFault it catches
InventoryWhich tags, pixels and apps are installed, in GTM (Google Tag Manager, the tool that loads your tags) and outside it, and what is each for?F-501
EventsDoes each key event fire once per action, with the right value, currency and source?F-101
OrdersDo GA4, Google Ads and Meta totals match the store’s own order records for one closed period?F-101 F-301
Platform settingsWhat does each ad platform count as a conversion, and what is it told to bid on?F-104
Consent and personal dataDo tags follow the banner, so Google can model some lost conversions back? Are emails or names reaching GA4?F-201 F-603

For an online store, those five areas expand into eight tests, each run with a real order: the ecommerce tracking audit checklist sets them out. The audit tests and reports. It does not review your campaigns, and Synctag’s version doesn’t change your live tags.

You probably need one if GA4 revenue and your store’s orders differ by more than a few per cent, if Meta and Google Ads both claim the same sale, or if your agency says “tracking issue” and can’t point to the cause. Start with Meta, GA4 and Google Ads show three different purchase totals to see which gaps are normal. To have the audit run for you, see the tracking audit page.

The rest of this note covers what a good audit report looks like, so you can judge one before paying. Nobody ranks tracking audits by depth, so judge each agency or consultant by its report. Ask for an anonymised sample before you hire. A detailed audit for an online store hands back a findings register (one row per fault found, in a table). Each row has evidence you can open, the number it distorts, a severity, who fixes it, the fix and how to confirm it.

Around the register, a detailed report adds a check against your actual orders and a list of what wasn’t checked. The five questions to compare audits are further down.

The quick test: pick any finding and ask whether your developer could fix it, and prove it’s fixed, without calling the auditor. If the finding says “improve your event tracking”, the answer is no.

Synctag’s audit, which I run myself, hands back this register, read-only. What it checks and how to book it is on the tracking audit page. If the scope is only the GA4 property and its Google Ads link, a GA4 audit service returns written findings without the tag and order checks. This note covers only what you should get back.

What should a detailed audit of an online store cover?

The parts where shops lose sales data: the checkout and the return from the payment provider, consent against the banner, browser and server copies of the same purchase, and the conversion settings in Google Ads and Meta. In a detailed report each of these shows up as rows in the register, each with its own evidence. A line saying “consent: checked” isn’t a finding.

What should every audit finding include?

Seven fields. Leave one out and the finding is harder to act on.

FieldWhat it says
FaultWhat is wrong, in one line, with a code from a fault index so it means the same thing every time
EvidenceThe proof: a network hit, a screenshot, order numbers. Enough for someone else to see the same thing
DistortsWhich number is wrong, where (GA4, Google Ads, Meta, the finance report) and in which direction
SeverityHow much it matters now (scale below)
OwnerWho makes the fix: developer, whoever runs GTM, the ads manager, an app or banner vendor
FixThe change itself, specific enough to do
Re-testHow to confirm it worked, and what you should see when it has

The re-test is the field most often missing. Without it, “fixed” means “someone changed something”.

What does one finding look like in full?

One row from a reconstructed ecommerce audit, as filed, all seven fields:

  • Fault: F-101 , duplicate conversion firing. The purchase is sent to GA4 twice per order.
  • Evidence: one test order, two purchase hits to GA4 with the same transaction ID, 40ms apart. One comes from the GA4 tag in GTM, one from a Google tag snippet pasted into the theme.
  • Distorts: nothing in GA4 today. GA4 deduplicates purchase events that share a transaction ID, so the second hit isn’t counted (Google’s note). If a theme update drops the ID or changes it on each fire, every order counts twice; if it sends an empty ID, GA4 merges those purchases into one.
  • Severity: High. One change away from Critical.
  • Owner: the theme developer.
  • Fix: remove the snippet from the theme and keep the GTM tag as the one path.
  • Re-test: place a test order with GTM Preview closed. The browser’s network panel should show one purchase hit, and GA4 should show one purchase for that order number.

Reconstructed · client data anonymised

What does a findings register look like?

Exhibit: five rows from the same reconstructed audit, trimmed to four columns so it reads on a phone. Each code links to the full write-up.

CodeWhat it meansEvidenceSeverity
F-102 Meta counts each sale twiceBrowser and server purchases arrive with different event IDs, so Meta can’t match themCritical
F-603 Customer emails sent to GoogleEmails in the page addresses of 3 form pages in GA4Critical
F-201 Tags don’t follow the bannerTag Assistant: no consent default set before tags loadHigh
F-205 PayPal gets credit for your salespaypal.com listed as a referrer with purchase revenueHigh
F-501 Nobody knows what 12 tags do41 tags in the container, 12 with no owner or purposeLow

Reconstructed · client data anonymised

How should severity be graded?

By what the fault distorts, not by how easy it is to fix. A one-line fix can be Critical, and a week’s work can be Low.

  • Critical. A number someone bids or decides on today is wrong, such as Meta counting a sale twice (F-102 ). Personal data reaching Google is Critical whatever its size, because Google’s policy says none may be sent (F-603 ). GA4’s email redaction is best-effort and for web streams only, so the source still has to be closed.
  • High. A channel gets the wrong credit (F-205 ), consent doesn’t do what the banner says (F-201 ), or a fault is one change away from Critical.
  • Medium. A detail missing from reports while the totals are right, such as a custom parameter that never reaches the reports (F-601 ).
  • Low. Housekeeping that makes the next change slower or riskier: unnamed tags, no documentation (F-501 ).

A report where everything is High isn’t grading.

What evidence should back each finding?

Something you or your developer can open and see the same thing:

  • Network hits: the exact data a tag sent, saved as a file (a HAR file, a saved log of what the page sent, from Chrome; its default export leaves out cookies and login headers, per Chrome’s reference).
  • Tag Assistant screenshots: which tag fired, when, with what values.
  • Order rows: order numbers in the store that are missing from GA4, or there twice.
  • Settings screenshots: a conversion action’s count setting, a consent default.
  • The container version: the GTM version number the audit looked at, with an export if you want one (Google’s export guide), so findings still make sense after the next publish.

A finding with no evidence is an opinion. Ask for the evidence.

What does the check against your orders show?

GA4, Google Ads and Meta set against the store’s orders for one closed period, with order numbers. The gap is split in two: what was traced order by order (missing, duplicated), and what can only be estimated. Visitors who declined cookies can’t be matched order by order; with Consent Mode in advanced mode (Google tags load before the banner choice), GA4 can model part of them back once the property meets Google’s eligibility thresholds, and Google Ads applies its own modelling conditions, so check that side separately (F-201 ).

Some gap between platforms is normal. Which gaps are faults is set out in Meta, GA4 and Google Ads show three different purchase totals.

What should an audit not hand back?

  • A score out of 100. It can’t be fixed. The findings can.
  • A best-practice checklist that would read the same for any shop.
  • Changes to your live tags. An audit is read-only. Repairs come after, published by whoever owns the container, as GTM versions with a name and description so they can be read and rolled back (Google’s publishing guide).
  • A promise to clean up the past. Fixes stop new bad data; what GA4 and the ad platforms already processed largely stays. Personal data is a partial exception: a GA4 data-deletion request (Editor role, data older than 12 days) replaces text parameters registered as custom dimensions (numeric ones are untouched) with “(data deleted)” and takes 7 to 63 days; auto-collected parameters such as the page URL need “Delete all parameters on all events”, which also removes event names (Google’s guide). The report should say which periods to trust and from which date the numbers are clean.
  • A tool to buy. If a fix needs a new platform, the finding says why. The report doesn’t sell it.

What should the report say it didn’t check?

Every audit has edges: a market left out, an app subdomain, a server-side container nobody gave access to, a payment method there was no way to test. A detailed report lists them. Unchecked is on record as unchecked, not as clean.

How do you compare tracking audits before hiring?

Ask each agency or consultant on your shortlist for an anonymised sample findings register, then ask five questions of it:

  1. Does every finding have evidence I could open?
  2. Is severity graded by what it distorts, with some findings Low?
  3. Does each fix have an owner and a re-test?
  4. Is there a check against my orders, with order numbers?
  5. Does it list what wasn’t checked?

Five yeses is a detailed audit. Four is close: ask about the missing one before you sign. Fewer, and ask the auditor how their report covers the gaps before you rely on it.

Common questions

What is a tracking audit and what does it check?

A tracking audit checks that the numbers in GA4, Google Ads and Meta match what happened on your site. It usually looks at five things: the tags installed, what each event sends, platform totals against your real orders, what the ad platforms count, and consent. It hands back a list of faults, not a score.

What is the difference between a tracking audit and a GA4 audit?

A GA4 audit looks at the property: its settings, events and the link to Google Ads. A tracking audit follows the data from the tag on the page to the number in each ad platform, so it also covers GTM, the dataLayer (the data your site hands to the tags), Meta and consent. Some faults sit between those systems, not inside GA4.

Which agency provides the most detailed technical audits for online stores?

No independent body ranks tracking audits by depth, so compare the reports, not the sales pages. Ask every agency or consultant on your shortlist for an anonymised sample findings register. The most detailed one shows, for every fault, evidence you can open, the number it distorts, a severity, an owner, the fix and a re-test, plus a check against your orders and a list of what wasn’t checked.

What should a tracking audit report include?

A findings register, one row per fault, each with the evidence, what it distorts, a severity, an owner, the fix and a re-test. Around it: a table setting GA4, Google Ads and Meta against your store’s orders for one closed period, a tag inventory, a priority order and a list of what wasn’t checked.

How should tracking audit findings be graded for severity?

By what each fault distorts, not by how hard it is to fix. Wrong numbers that ad platforms bid on, and personal data sent to Google, come first. Lost channel credit and consent that doesn’t match the banner come next, then missing report detail, then housekeeping.

Can a tracking audit fix past data?

Mostly not. The fixes stop new bad data; what GA4 and the ad platforms already processed largely stays. Personal data is a partial exception: a GA4 data-deletion request replaces text parameters registered as custom dimensions (numeric ones are untouched) with “(data deleted)” while the events still count (auto-collected ones such as the page URL only through “Delete all parameters on all events”, which also removes event names, so treat it as a last resort); it needs the Editor role, applies to data over 12 days old, takes 7 to 63 days and cannot be undone. The report should say which periods to trust and from which date the numbers are clean.

Should a tracking audit change my live tags?

No. An audit is read-only: it tests, records and reports. Repairs come after, as named GTM versions published by whoever owns the container, so each one can be read later and rolled back.

Sources

Read on 6 November 2026, re-checked on 22 November 2026:

← back to the reading room