Server-side tracking and Consent Mode v2, explained without the jargon

What problem server-side solves

The browser is a hostile environment for measurement. Ad blockers strip pixels, Safari's Intelligent Tracking Prevention shortens cookie lifetimes, and privacy extensions block third-party requests outright. The result is that a meaningful share of real conversions never reach the ad platforms.

Sending the event from your server instead means the conversion is reported regardless of what the browser allows — with your server as the trusted source of truth about what actually happened.

How deduplication works

Both the browser pixel and the server event should fire, each carrying the same unique event ID. The platform receives both, recognises the shared ID, and counts one conversion — taking whichever arrived with the richer data.

Skip the shared ID and you double-count everything, which is worse than the problem you set out to solve.

• Generate one event ID per order or lead
• Send it on both the browser event and the server event
• Include hashed email, phone, name and location for match quality
• Verify in Events Manager that deduplication is being applied

Consent Mode v2 in plain terms

Consent Mode v2 adds two signals — ad_user_data and ad_personalization — on top of the existing analytics and ads storage flags. Before the visitor decides, everything is set to denied and no identifying cookies are written. After they choose, the tags adjust behaviour accordingly.

When consent is denied, Google can still model conversions from aggregate patterns, which recovers much of the lost reporting without storing anything on the visitor's device.

A correct implementation, end to end

In our builds, consent defaults to denied, the banner writes the visitor's choice, and every subsequent tag respects it. Conversions fire from the server with hashed identifiers and a shared event ID, click IDs and UTM values are captured on entry and persisted, and GA4 runs alongside with the same consent state.

Nothing identifying leaves the server unhashed, and no raw email address or phone number is ever transmitted to an ad platform.

What to expect after switching

Reported conversions usually rise in the first two weeks, because events the browser previously lost are now being counted. Match quality scores improve, audiences rebuild more accurately, and cost per acquisition typically falls over the following four to eight weeks as the algorithm learns from a fuller picture.

Do not compare the month before and the month after as though the numbers are on the same basis — they are not. Note the change date in your reporting and treat it as a new baseline.