Full Answer

When a visitor declines consent on a Consent Mode V2 banner, Google tags switch to cookieless pings that carry only timestamps, user agents, and referral URLs — no personal identifiers and no conversion value. Google's behavioral modeling layer then attempts to reconstruct the missing conversions statistically, filling the gap between what was measured and what likely happened.

The recovery rate depends heavily on traffic volume. Google's own documentation states the modeling needs sufficient signal to be accurate, and independent testing shows that stores below 1,000 daily sessions see modeled data that is materially less reliable than the 65–70% headline figure. For a WooCommerce store doing 200 sessions per day, the practical recovery may be closer to 40%, leaving more than half of declined-consent conversions completely unrecoverable through modeling alone.

The structural issue is that Consent Mode operates entirely in the browser. If a visitor declines and leaves, the cookieless ping fires but carries no purchase data — there is nothing to model against. Server-side tracking captures the conversion event at the server before the consent decision affects the tag, sending it directly to Google Ads or Meta CAPI. This does not bypass consent law — consent still governs what data is sent — but it ensures the event reaches the endpoint rather than depending on a browser-side model to guess what happened. For stores where every tracked conversion influences Smart Bidding, the difference between modeled recovery and actual measurement is the difference between optimised spend and wasted budget.