Conversion tracking setup that catches what’s wrong with your numbers

Compare three numbers for the same period: what analytics reports, what each ad platform reports, and what your CRM or order system recorded. They’ll never match exactly, but the size and direction of the gaps tell you a lot.

Ad platform conversions far above recorded enquiries usually means double-firing. A sudden step change with no campaign behind it usually means a site release. A total that’s suspiciously round and stable for months usually means something fires on every page view.

It lets your tags respond to the choice a visitor made in your consent banner. When consent is granted, tags behave normally. When it’s denied, Google tags don’t read or write the identifiers they otherwise would, and send a limited signal instead.

Google then models some of the missing conversions. That’s legitimate, but it’s an estimate, not a count, so a report that mixes the two without labels will rightly get challenged by finance.

Consent mode isn’t a way to keep collecting from people who said no. A vendor describing it that way is describing a compliance problem.

Sometimes, and less often than it’s recommended. It gives control over exactly what data leaves for each platform, resilience against browser limits on cookies, and fewer third-party scripts in your visitors’ browsers.

It doesn’t recover consent-declined data, doesn’t fix a badly defined event, and costs money every month. We recommend it only when signal quality, page weight or data control justify the cost.

Usually because the booking system sits on another domain, and without cross-domain measurement the arrival counts as a new visit. The original source is lost, one session becomes two, and the conversion is credited to a referral from your own website.

The fix is configuration, not a rebuild: declare the domains to each other and pass the identifier across. The tell-tale sign is your own domain showing up as one of your biggest referral sources.

That’s the main constraint on how we work. Nothing gets deleted for looking unused. We document each tag, run the replacement alongside the existing setup where the platform allows, and compare the two before switching anything off.

Where a change is unavoidable, we make it on a set date and annotate the reporting, so nobody later mistakes an implementation change for a change in performance.

Often you don’t, and where a developer has built a data layer properly, we’d rather build on it than replace it.

The usual gap is definition, not technical skill. A developer builds what was asked for. Whether a form submission and a quote request count as the same conversion, and how either matches what sales calls a lead, is a commercial decision rarely made before the tags are written.