Comparing numbers from an advertising platform and a site analytics tool almost always produces a gap. The gap is structural, and treating it as a fault leads to time spent chasing something that cannot be closed.
Each system counts from where it sits
An advertising platform records what it delivered and what it can associate with that delivery, while a site analytics tool records what happened after arrival.
The two observe different events at different moments, and neither has visibility into the other's half of the journey.
A visitor who clicks and abandons before the page loads exists in one system and not in the other, which produces a gap before any definitional question arises.
Definitions differ in ways that compound
Session boundaries, what constitutes a new user and how long an interaction remains attributable all vary between tools, and each difference multiplies through the report.
One system may attribute a conversion to the click that started the visit while another attributes it to the last channel touched before purchase.
Both are defensible and they cannot produce the same number, so aligning them requires choosing one definition rather than reconciling two.
Attribution windows create the largest divergence
Platforms typically count a conversion occurring within a set period after an interaction, including views that never produced a click.
Site analytics generally requires a traceable arrival, so view-based conversions are invisible to it entirely.
The resulting difference is often the single largest component of the gap, and it grows with the length of the purchase cycle being measured.
Blocking and consent remove data unevenly
Browser protections, extensions and consent choices suppress measurement scripts at different rates depending on the audience and the device.
Because different tools rely on different mechanisms, the same suppression removes different proportions from each, changing the gap without changing behaviour.
Modelled estimates are then applied to fill the missing portion, and the models differ between vendors, which adds a further layer of divergence.
Stability is the metric that matters
A consistent gap is a definitional artefact and can be worked with, since trends remain readable even when absolute figures do not agree.
A gap that changes suddenly points at something real, such as a broken tag, a tracking parameter stripped by a redirect, or a consent banner change.
Monitoring the ratio between systems is therefore more useful than monitoring either number, because it turns an unresolvable discrepancy into a working diagnostic.