Tracking-tag regression watch on your own site

Why it matters

Every decision about ads and channels rests on the numbers your tags send. If the purchase event stops firing, a campaign that is working can look like it is failing, and the reverse. The cause is usually mundane: an app update, a new page template, a consent setting changed by someone else. The fix is to check the same pages the same way after every change.

No loop covers this. The four loops, all in build and none on sale, look outwards at rivals and at your ad accounts; none watches your own tags. So this recipe is by hand. Every other report you read rests on these tags, which is the reason to run it.

What you need

• - A list of your key pages: home, a category, a product, basket, checkout and order confirmation
• - Access to Google Tag Manager preview and Tag Assistant, or your platform's equivalent
• - A list of the tags that should fire on each page (analytics, ad platforms, consent)
• - A test order method your platform supports, so you can reach the confirmation page without a real purchase

Setup

1. Write down, for each key page, the tags and events that should fire, and in what consent state.
2. Open Tag Manager preview with Tag Assistant and walk the pages in order, recording which tags fired and in which order.
3. Repeat the walk twice: once accepting all cookies and once declining, so you can see how consent changes what fires.
4. Paste your expected list and the observed results into the prompt below.
5. Repeat the check after every theme update, app install, checkout change or consent banner change, and keep each result.

The prompt

Paste this into any capable model, followed by the evidence you gathered in the setup.

You are helping a UK online business check whether its tracking still works after a site change.
Below are: an EXPECTED list (page, tag or event, consent state in which it should fire) and an OBSERVED list from a Tag Assistant walk (page, tags that fired, order, consent state), plus a note of what changed on the site recently.
Do this:
1. Compare expected and observed page by page. List every tag or event that was expected but did not fire, and every one that fired but was not expected. Quote the observed rows.
2. For each mismatch, give the most likely cause, marked as a hypothesis, linked to the recent change if the evidence supports it.
3. Check the consent behaviour: does anything fire before consent that should wait, or fail to fire after consent is given?
4. List fixes in order, each rated for effort (low/medium/high) and impact (low/medium/high), with the reason.
5. Say which decisions (for example ad budget changes) should wait until the fix is confirmed.
Rules:
- Quote the evidence for every point, word for word, and say where it came from (which rival, which page or ad, which date I captured it).
- Label each point OBSERVATION (you can see it in the evidence) or HYPOTHESIS (your reading of why). Never present a hypothesis as a proven cause.
- Never invent a figure, a date, a price or a result. If the evidence does not show it, write "not in the evidence".
- Use British English and plain words.

Expected, observed and recent changes:
[paste here]

Doing it by hand today

This is a by-hand job today. Google Tag Assistant works with Tag Manager preview to show which tags fired and in which order, which is exactly the record you need.

Pay attention to consent. Consent mode passes a visitor's choices to Google tags using the types ad_storage, analytics_storage, ad_user_data and ad_personalization; a misconfigured banner can block tags you expected, or fire them when it should not.

Keep each check sheet with its date. When a report looks odd, the first question is whether tracking changed, and a dated record answers it. The ad spend tool is a useful sense check once you trust the numbers again.