When to run this checklist
Run it after any change to the tag manager container, the pixel, a conversion action or the consent banner, and after any site release that touches a form, a checkout or the page templates the tags depend on. Run it before a campaign launches, because the campaign launch QA checklist hands tracking questions to this list rather than answering them. Run it when a property or an account changes hands, alongside the GA4 setup checklist, since an inherited setup is often half-finished. The checklists library routes the other situations, such as a landing page or an email send, to the list that hands off to this one.
Run it on a schedule too. Tracking rots quietly: a developer renames a button, a plugin update changes the data layer, a consent vendor changes a default, and nothing alerts anyone until a report looks wrong weeks later. A monthly pass through the list on the key journeys is cheaper than reconstructing a quarter of missing conversions, and the record it produces is the evidence that the numbers in the reports were trustworthy on a given date. The tutorial on checking conversion paths after a release covers the other half of that monthly pass, walking each journey from the visitor's side to the order, lead or booking.
Before you start
Have the tracking plan open; the tracking plan guide explains what it should contain, and if there is no plan the first group of items cannot pass. Have access to the GTMTag management system. container with preview rights, the GA4Google's event-based analytics platform. property with DebugView, the Google Ads conversion settings and the Meta Events Manager for the pixel in use. Use a clean browser profile or a private window so that the consent banner appears in its default state and no earlier cookies interfere.
Arrange a way to make a test conversion that can be recognised and removed later: a test product, a form submission from a known email address, or a test-mode transaction. Know how internal and test traffic is excluded in the property, so that the verification itself does not inflate the data. Finally, have somewhere to record the result, with screenshots and the date, because a verification nobody can find afterwards is a verification that did not happen.
The checklist
Work through the groups in order, from the plan outward. The items marked in the plan, tag manager, analytics and ad platform groups mirror the compact tracking list on the resources page; the consent and reporting groups extend it to the areas most often skipped. Each item is a pass-or-fail check made with a real test action, not a reading of a configuration screen. Ticks are saved in your browser only, so keep the written record alongside.
Plan
Tag Manager
Google Analytics
Ad platforms
Consent and privacy
Reporting
If an item fails
A plan failure is fixed before anything else, because every later item is a comparison against the plan. Add the missing events, name the owners and agree the key events; the tracking plan guide has the structure. Tag manager failures are worked through with preview mode, and the debugging with preview mode tutorial shows how to read the event timeline, the variables and the data layer state for each event. A tag that fires on every page usually has a trigger scoped to all pages rather than to the action; a tag that reads an empty value usually fires before the data layer push it depends on.
Analytics failures split into property problems and event problems. If DebugView shows nothing, or key events are not defined, the GA4 setup tutorial covers the property settings, and the key events and conversions guide explains what to mark and why. Ad platform failures are fixed in each platform: the Google Ads conversion tracking setup guide covers the status labels and what each one means, and the Meta pixel setup guide covers Events Manager and the test event tool. Where the tags fire correctly but an inherited Google Ads account counts the same event through several actions, the conversion action clean-up tutorial removes the duplicates and resets which actions are primary. If browser and server events are both arriving but counted twice, the Conversions API guide explains deduplication through a shared event identifier; Conversions APIServer-side event sharing with Meta. without it inflates every number.
Consent failures are the ones with legal consequences. If a tag fires before consent, the consent mode guide covers the tag manager side and the privacy and consent guide the policy side. A reporting gap that cannot be explained is not a failure of the checklist but a finding; the platform versus analytics discrepancies guide lists the usual causes, from attribution windows to consent, and most gaps are explained by two of them.
A form event that fired twice and doubled the lead count
A recruitment agency notices that Google Ads reports 212 leads in a month while the CRM shows 108. The
verification pass finds the cause in the tag manager group: the form's success trigger fires once on the
submission and once more when the thank-you message re-renders, so every lead produces two generate_lead
events and two conversions. The fix is a trigger that fires once per page load, confirmed in preview and
in DebugView. The team then re-runs the ad platform group, sees the conversion action still marked as
recording, and saves the record with screenshots. The next month Google Ads reports 115 leads against 111
in the CRM, a gap of four that the attribution window explains.
Common mistakes
- Verifying from the configuration screen. A tag that exists is not a tag that fires. Every tick needs a real test action traced through to the system that should receive it.
- Testing with consent already granted. The browser that built the tags has long since accepted everything. A private window shows what a new visitor's first visit actually sends.
- Skipping the value check. An event that arrives with no value, or the wrong currency, passes most people's eye test and ruins every return calculation downstream.
- Treating deduplication as a Meta setting. The shared event identifier has to be generated on the site and sent by both the browser and the server; no platform setting creates it for you.
- Leaving the test conversion in the data. A verification that adds a fake order to the month's revenue has replaced one error with another. Exclude the test device or remove the test afterwards. The guide to test environments and test data sets up marked test orders and their exclusions before the test runs.
- Not writing it down. The record with screenshots and a date is what lets the next person, or you in three months, know that the numbers were right on that day.
