CMP statistics can quickly become a collection of numbers without a clear purpose.
Acceptance rate, rejection rate, preference panel openings, category choices and service-level choices can all theoretically be measured.
The more useful question is which information helps understand consent behaviour without turning the CMP into another visitor tracking tool.
Acceptance rate is the most obvious metric
The first indicator people generally look at is the acceptance rate.
It answers a simple question: among expressed choices, what proportion represents acceptance?
This number can be useful, particularly when comparing how one website evolves over time. It becomes much more questionable when treated as a commercial KPI that must always be maximized.
A CMP is not a conversion funnel. Its purpose is to collect a valid choice, not to turn every visitor into a positive consent.
Rejection rate matters just as much
Honest consent statistics obviously need to include rejection.
If an interface reports 70% acceptance, the remaining 30% are not visitors the banner “failed” to convert. They simply expressed another choice.
Following both values can also reveal unusual changes.
A sudden fall in rejection after a visual change may deserve investigation: did the interface genuinely become clearer, or did the reject button simply become less visible?
Preference panel openings provide different information
The detailed panel allows visitors to move beyond the binary “accept all"or"reject all” choice.
Measuring panel openings can therefore show whether this possibility is actually being used.
A very low rate is not necessarily a problem. Many visitors simply want to make a global choice quickly.
But the event can help explain how the interface is used without needing to know the identity of the person using it.
Service-level choices can be useful
On a website using several services, individual choices provide additional information.
Visitors may accept audience measurement while rejecting an advertising service, or authorize an external video individually.
This information can help identify which services are commonly accepted and which are frequently rejected.
It may even raise a useful product question: if almost everyone rejects a service, is that service genuinely necessary on the website?
Every event does not need to become a profile
Calculating an acceptance rate requires counting events.
That does not require building a detailed visitor profile containing browsing history, viewed pages and every successive choice.
I prefer distinguishing statistical measurement of CMP behaviour from individual tracking of a person.
The former can answer operational questions using far less data.
Statistics and proof of consent are different subjects
These statistics should not be confused with the legal question of demonstrating consent.
French privacy guidance states that controllers relying on consent need to be able to demonstrate that valid consent was obtained.
This does not mean that a CMP’s global acceptance rate is by itself proof for every processing activity.
Conversely, the question of proof does not necessarily mean that the statistics tool needs to become a marketing database individually profiling every visitor.
The measurement context matters
To understand a statistic, you need to know how it was produced.
The number of visitors shown the banner, the number of expressed choices and the way events are counted all influence the result.
A rate based on sessions, visitors, banner displays or only expressed clicks does not answer exactly the same question.
Comparing two CMPs solely from a published percentage without knowing the methodology is therefore unreliable.
A very high acceptance rate is not automatically good news
An interface may obtain more acceptance because it explains choices more clearly.
It may also obtain more because the accept button occupies half the screen while rejection is gray, small and difficult to find.
The number alone cannot distinguish between those situations.
This is why I am cautious about commercial comparisons based solely on promises to increase consent rates.
Statistics are particularly useful for detecting anomalies
I find statistics especially useful when observed over time.
A sudden change after an update may reveal an interface issue. Events suddenly disappearing may indicate an integration error. More preference panel openings may follow a wording change.
In this context, statistics become a diagnostic tool rather than a competition for the highest acceptance rate.
You still need to inspect technical behaviour
An excellent consent rate says nothing about blocking quality.
A CMP could theoretically report 60% acceptance while allowing a tracker to fire before the choice.
Interface statistics should therefore remain complementary to technical verification.
DevTools can be used to directly inspect requests and cookies.
Which events should be measured?
A few events are generally enough to provide useful information without unnecessarily multiplying data:
- banner display;
- global acceptance;
- global rejection;
- preference panel opening;
- validation of a customized choice;
- optionally service or category choices when that information is useful.
This foundation already provides a reasonably precise understanding of how the interface is used.
Measure to understand, not to push
I find statistics useful when they help detect a problem or understand how a CMP is used.
They become much less useful when their only purpose is finding the button color that makes more visitors accept.
A CMP needs to provide a genuine choice. A rejection rate is therefore not a failure rate.
The best statistic is one that helps explain what is happening without changing the purpose of the tool itself.
