Conversion tracking

A tag fires at the moment the action completes, typically on a confirmation page or in response to a client-side event.

A value and an identifier usually travel with it, so revenue and deduplication are possible downstream.

The advertising platform matches that event against a click it recorded earlier, which is the attribution step.

Server-side implementations do the same thing from the backend, which is more reliable and does not remove the consent question — it relocates it.

It happens when the same tag is installed both through a tag manager and directly in a template, and neither owner knows about the other.

The effect is not subtle: every conversion counts twice, so reported revenue doubles and every derived figure — cost per acquisition, return on spend — is wrong by a factor of two.

It is also self-reinforcing, because the inflated numbers make the channel look good enough that nobody questions them.

Several distinct identifiers for the same product are a different matter and frequently deliberate — a separate property per subdomain, an agency tag alongside a client's.

Which is why the two are distinguished rather than reported together.

A tag placed on a page load rather than bound to the completion event counts everyone who reached the page.

On a confirmation page that is nearly the same thing, until someone refreshes it, bookmarks it, or reaches it without completing.

On a checkout page it is not the same thing at all, and the reported conversion rate becomes a measure of intent.

The diagnostic is straightforward: reported conversions materially exceeding the records in the system that actually processed them.

Which advertising and analytics tags are present, detected from real script hosts and markup fingerprints.

Whether the same tag appears more than once, kept distinct from several deliberate identifiers.

Whether conversion tracking is present at all on a commerce site, which is a common and expensive omission.

Not whether it fires correctly, and not what it reports — both require completing the action, which a scan does not do.

A commerce site with no conversion tracking is spending on acquisition with no way to tell which spending worked.

A site with tracking that double-counts is worse off than one with none, because the decisions are being made confidently on wrong numbers.

Each tag is also a third-party script on the critical path, so the inventory is a performance question as well as a measurement one.

And each is a data transfer to another company, which puts it in the consent design rather than outside it.