tracking-checker

Read a website's public Google tagging and report what is likely broken, without ever touching the Ads or Analytics account.

It exists because the same handful of failures keep costing advertisers real money, and all of them are visible from outside: a conversion tag wired to the page-load event, a missing conversion linker, one conversion action fired by two tags with nothing to deduplicate on, two GA4 properties on the same page.

$ tracking-check example.com

Scanned: https://example.com
Containers: GTM-ABCD123

[CRITICAL] An Ads conversion fires on page load
    Evidence: GTM-ABCD123 tag #7 conversion 11248148062 triggers: event=gtm.js
    The tag fires before any success happens, so it counts page views as conversions.
    Smart Bidding optimises on that signal.
    Verify: In Tag Manager, open this tag and read its trigger. If it is All Pages, the
    conversion should instead fire on the confirmation event, or on a URL only reachable
    after the form is submitted.

Install

No dependencies beyond the standard library. Python 3.10+.

git clone https://github.com/<you>/tracking-checker
cd tracking-checker
python3 -m tracking_checker.cli example.com

Or install it:

pip install -e .
tracking-check example.com

Use

tracking-check example.com                          # one site
tracking-check a.com b.com c.com                    # several
tracking-check example.com --lang fr                # French output
tracking-check example.com --format json            # machine readable
tracking-check example.com --format markdown > report.md

Exit code is 1 when at least one critical finding was raised, so it can sit in a cron job or a CI step and actually fail.

from tracking_checker import check

result = check("example.com")
print(result.critical, [f.id for f in result.findings])

What it checks

id	severity	what it means
conversion-on-pageview	critical	An Ads conversion tag is gated only by the page-load event, so it counts visits rather than successes
ga4-conversion-on-pageview	critical	A GA4 event named like a success fires on page load. If it is a key event imported into Ads, bidding optimises on page views
no-conversion-linker	critical	Ads conversions with the linker off and no standalone linker tag, so the click id is not stored first-party
conversions-not-visible	info	No Ads conversion tag at all, so the conversions are decided in the accounts and cannot be read from outside
duplicate-conversion	warning	One conversion action fired by several tags with no transaction or lead id to deduplicate on
multiple-ga4	warning	Several GA4 properties measuring the same page
enhanced-conversions-off	warning	Enhanced conversions unchecked on every conversion tag
consent-no-default	warning	A consent banner is present but no consent default is visible in the served HTML
consent-signals-incomplete	warning	A consent default that does not cover all four signals
multiple-containers	info	Several Tag Manager containers on one page, the usual leftover of an agency change
paused-tags	info	The container holds paused tags, a frequent cause of "the conversion just stopped"
server-side	info	Tagging served from a server-side container, which cannot be read from outside
no-tagging	warning	No Google tagging in the served HTML

What it deliberately does not do

It never puts a number on the damage. It reads a public container, not an account. It cannot see spend, conversion volume or revenue, so it will never tell you that you are losing a percentage of anything. Every finding says what was observed and how to confirm it yourself, usually in about two minutes. A checker that guesses is worth nothing the moment someone opens the account and looks.

Three concrete blind spots, printed at the end of every report:

• JavaScript is not executed, so tagging injected after load is invisible.
• The Ads and Analytics accounts are not read.
• A server-side container's forwarding rules cannot be read from outside.

A clean report is not proof that a setup is correct. It means the public container shows no known failure pattern.

How it works

Two public sources. The HTML the server returns for the URL, and googletagmanager.com/gtm.js?id=GTM-XXXXXXX, which is the compiled published container: its tags, its triggers, and the conditions on those triggers. Resolving triggers onto tags is what makes conversion-on-pageview detectable at all.

parse.py describes what is there. checks.py decides what is wrong. They are kept apart on purpose, so the description stays true even when the judgement changes.

Tests

python3 -m pytest tests/ -q

Fixtures are synthetic containers built by hand, not captures of real sites.

Licence

MIT.