Consent Mode v2 and modelled conversions: what advertisers in the EU need to know
Consent Mode v2 became unavoidable for anyone advertising to EU and EEA users: without it, Google's advertising products stop accepting the signals that power audiences and conversion measurement for those users. Most advertisers implemented it under deadline pressure, which means most implementations were done quickly rather than well. A year on, the difference between the two shows up directly in account performance.
What the framework actually does
Consent Mode is a signalling layer between your consent banner and Google's tags. Two signals matter for advertisers: ad_storage (may advertising cookies be used) and ad_user_data / ad_personalization — the v2 additions governing whether user data may be sent to Google for advertising and whether it may drive personalisation. Signals default to denied for regulated regions and update when the user chooses.
The design's point is what happens on denial. In the "advanced" implementation, tags still fire — cookieless, stripped of identifiers — sending anonymous pings that tell Google an ad click happened and a conversion happened, without linking them to a person. Those pings feed conversion modelling: Google estimates, from observable consented behaviour and the anonymous signals, how many unconsented conversions occurred, and feeds those modelled conversions into reporting and Smart Bidding. In the "basic" implementation, tags don't fire at all until consent — compliant, but modelling gets far less to work with, and your measurement in high-denial markets goes dark.
What modelled conversions mean for your numbers
Modelled conversions are statistically real and individually fictional: right in aggregate — Google's stated aim — but not attached to any actual customer record. Practically, expect three effects. Your Google Ads conversion counts in EU markets will exceed what you can match to CRM rows; the gap is the model. Day-to-day numbers become slightly less stable, since models re-estimate. And the modelling quality depends on volume — small accounts and small conversion actions get coarser estimates, which argues for consolidated conversion actions in low-volume EU markets.
None of this is a reason to avoid modelling; the alternative is not "accurate data", it is a biased dataset missing precisely the privacy-conscious segment. It is, however, a reason to reconcile platform numbers against payment-provider truth monthly, and to expect the ratio to differ by market in line with consent rates.
The implementation details that matter
Defaults must load before tags. The consent default command has to execute before any Google tag, in every template and single-page-app route. The most common defect we find is a race condition where tags occasionally beat the default — intermittent, invisible, and non-compliant.
Use a certified CMP, wired correctly. Google requires a certified consent management platform for EEA traffic. But certification covers the CMP, not your wiring: verify that banner choices actually update all four consent signals, including the v2 pair, and that region targeting matches your legal advice.
Pair it with server-side tagging and Enhanced Conversions. Consent Mode governs whether data may be sent; server-side infrastructure improves how well consented data survives the journey. Together with Enhanced Conversions' hashed first-party data for consented users, they recover most of what a naive setup loses. Separately, each is a partial fix.
Monitor consent rates as a KPI. Banner design moves consent rates dramatically, and consent rate now feeds directly into measurement quality. Track it by market alongside your media metrics; a UX tweak that lifts consent five points is worth real money.
The regulation is not going away, and its cousins outside the EU are multiplying. Advertisers who treat consent infrastructure as a compliance checkbox will keep bleeding signal quietly; those who treat it as measurement architecture keep their optimisation loop intact. The second group is buying customers with better information — the compounding kind of advantage.