What a consent choice does to your reporting data.

A consent decision is a signal travelling with the page, telling the tags on it what they are permitted to do: whether an identifier may be stored, whether an event may be sent onward, whether what is sent may be used for advertising as opposed to measurement.

It is a set of permissions rather than one switch, and different permissions gate different things. That is why a visitor is rarely simply in or out. They are commonly counted for one purpose and not for another, in the same session, and the report that shows them is assembling several different populations without saying so.

Once you can see these separately, most of the confusing numbers stop being confusing.

• Consented and observed

  Measured the way analytics has always described. This is the group every explanation of how tracking works is implicitly about, and it is no longer the whole of a client's traffic.

• Present but not observed for that purpose

  The visit genuinely happened and part of it is not collected. These people are in the client's server logs and absent from the reports, which is why log volume and analytics volume stopped agreeing.

• Estimated on your behalf

  Platforms fill part of the gap with modelled figures derived from the population they can still see. A modelled conversion is a real conversion that was not observed, expressed as a number. It is neither invented nor measured, and it is reported in the same column as the observed ones.

This is the part that costs teams the most time. A platform's own interface generally shows observed and modelled together, because that is the figure it wants you to optimise against. What arrives in your warehouse through an export is often the observed part, or a different blend of the two.

Neither number is wrong. They are answers to slightly different questions, and the difference between them is not a pipeline fault. Teams lose weeks to this, because the gap has every appearance of a bug: it is consistent, it is one-directional, and it does not reproduce when you check a single row.

The practical response is to decide, per metric, which of the two is your reported figure, and to write that down where the metric is defined. A report that silently combines a modelled figure from one source with an observed figure from another cannot be reconciled by anybody, including the person who built it.

The size of those three populations is not stable. It moves when a banner changes, when a platform changes how it models, and when a client's traffic mix shifts between places with different rules. None of those is a change in marketing performance.

So a metric can move for measurement reasons and look exactly like a result. Before explaining a drop to a client, check whether what actually changed was the share of visitors being observed. It is the first question worth asking and it is almost never the first question asked, because the dashboard offers no hint that it is a question at all.

Record the consent rate as a metric in its own right, on the same cadence and beside the metrics it affects. It is the denominator underneath everything else, and it is the only way to separate a measurement change from a performance change after the fact rather than by memory.

One boundary worth stating plainly, because this subject attracts confident writing that should not be trusted. This is a page about what happens to numbers. What a client is obliged to do, and where, is a question for somebody qualified to answer it, and the answer differs by jurisdiction and by what the client collects. Nothing here is a substitute for that.

What a platform reports is what lands, daily, in a warehouse you own, so the observed and modelled question is visible to you rather than resolved silently inside a reporting tool. The raw figures sit alongside the modelled tables, which is what makes the gap inspectable at all.

Consent rate can be ingested from wherever the client records it, including a spreadsheet or an emailed file if that is where it lives, so it sits beside the metrics it explains rather than in a separate system. Because syncs add to the tables rather than replacing them, a shift in that rate shows up as a change you can point at instead of an unexplained step in a conversion line.