Google Consent Mode v2

How it works

When a visitor rejects cookies, Consent Mode v2 prevents the analytics cookie from being written but still fires a “cookieless ping” — a request that records the event without any identifier. Google then aggregates the cookieless pings across many properties and uses a machine-learning model trained on the consenting visitors to estimate what the rejecting visitors probably did: how many sessions, how many conversions, by channel. The estimates appear in GA4 and Google Ads as if they were measured.

Modelled, not measured

This is the important distinction. Consent Mode v2 fills the gap; it does not close it. The number you see in GA4 with Consent Mode enabled is the consenting 13–40% of EU visitors plus a statistical estimate of the rest. The estimate is useful when you need a ballpark — directionally correct for cross-channel comparisons in stable markets — and unreliable when you need exact reconciliation against CRM revenue, when a new channel mix breaks the training assumptions, or when you are auditing for compliance and the data subject asks “was my visit measured?”.

When measurement is the right answer

For board-level revenue decisions, for CFO reconciliation, for the cost-of-customer calculations a serious finance team will defend — modelling is the wrong layer. Cookieless analytics measures every visitor on the same anonymous-aggregate basis, with no model in between. See the architectural argument on the complete data pillar.