What turns up in GTM and GA4 audits
The same few problems come back audit after audit, on sites of every size. Most of them never show up as an error.
The short answer
Across recent audits, the problems that repeat are key events that count something other than the outcome they’re named for, tags that load before a visitor answers the cookie banner, events nobody can trace to a source, and settings left at defaults that cost data. The reports look normal. The problems appear when the container, the live site and GA4 are checked against each other.
On this page
Key events that count the wrong thing
- One key event used a click counter that went up every time Tag Manager read it, rather than once per click. It recorded far more conversions than the site had leads, and those went into the conversion totals in every report.
- Another key event had recorded nothing in 90 days. Its trigger looked for a button on a page that no longer had it.
- A lead generation site still counted purchase as a key event, alongside time-on-site and scroll events. None of those are leads.
- One property kept ten key events from an old sign-up funnel that hadn’t recorded data all year.
Tags that don’t wait for the cookie banner
- On several sites the banner was on the page but not connected to Tag Manager. Analytics, advertising and session-recording tags loaded before anyone made a choice, and the container set no consent mode defaults.
- One site had no consent platform at all, so GA4 received no ads measurement or personalization signals.
- On a site whose checkout runs on another company’s platform, the website and the checkout handled cookie choices differently.
Events nobody can account for
- Every audit turns up events in GA4 that nobody on the team can trace. Some come from a third-party checkout or member portal, some from code in the page, and some from old tracking removed in one place but left in another.
- A paused tag’s events stopped on the date of its last edit. That explained a gap that looked like a tracking failure.
- GA4 and Google Ads were loaded in the page code and again through Tag Manager.
- One third-party checkout ran three Tag Manager containers owned by three different parties. One of them held changes nobody had published.
The same action counted twice
- GA4’s enhanced measurement already records form starts, and Tag Manager sent a second form start event beside it.
- Lead tier events (hot, warm and cold) copied the individual form events, then drifted from them, so the two sets of numbers stopped matching.
Settings left at their defaults
- Event data retention at two months, which limits explorations to recent data.
- Email and URL parameter redaction turned off.
- Debug mode left on in production, so ordinary visits appeared in DebugView and the developer traffic filter couldn’t work as intended.
- Internal traffic filters created and then left in testing.
- Staging and build domains suggested for cross-domain linking, or added to the conversion linker by the dozen.
- Search Console linked to an old URL-prefix property, or not linked at all.
Traffic that isn’t people
- A sudden run of Direct traffic from one country distorted a property’s numbers until it stopped as abruptly as it began.
- On more than one site, Direct was over half of all sessions, partly from customer portal visits that belong in their own view.
