Analytics
Consent Mode v2 for ecommerce: what changes in your numbers
Consent Mode v2 changes what GA4 and Google Ads can count when a shopper declines. The two new signals, basic versus advanced, and how to read reports after.
By CartKernel · Published
Consent Mode v2 is the mechanism by which your consent banner tells Google’s tags what a shopper agreed to, and switching it on changes what you see in GA4 and Google Ads from that day forward. Sessions fall, or reports start carrying modelled data. Google Ads conversions take on a modelled component. Revenue shifts between channels in attribution. None of that is a defect. The reports are reflecting consent choices that were being made all along, and this article explains each change, why it happens, and what to compare against so decisions can still be made.
What v2 adds to the original signals
Consent Mode began with two signals: analytics_storage, which governs analytics cookies, and ad_storage, which governs advertising cookies. Version 2 adds two more.
ad_user_data: whether data about the user may be sent to Google for advertising purposes.ad_personalization: whether that data may be used for personalised advertising, which is what remarketing lists depend on.
Google requires these signals for visitors in the European Economic Area before audience and measurement features in Google Ads will use their data. A Canadian or US store that ships to Europe is covered by that requirement for those visitors. Closer to home, Quebec’s private-sector privacy law and several US state laws mean most stores already run a banner of some kind, and the same four signals are how that banner’s choices reach Google’s tags. The mechanics are in Google’s consent mode setup guide, and the short definition is in the consent mode entry.
Basic or advanced: the choice that sets your numbers
There are two ways to implement it, and they produce different reports.
| Basic implementation | Advanced implementation | |
|---|---|---|
| Tags before consent | Do not load | Load, but send cookieless pings only |
| Declined visitors in GA4 | Absent | Counted as cookieless pings; modelled where eligible |
| Google Ads conversions | Modelled with a general model | Modelled with advertiser-specific data |
| Remarketing lists | Consented visitors only | Consented visitors only |
| Reporting shape | Fewer users, clean observed data | Fuller totals, part observed and part modelled |
Under the basic implementation, a declined visitor is invisible. Under the advanced one, the tag still sends a ping without cookies or identifiers, carrying the event name and its parameters, and Google uses those pings for modelling. Google’s consent mode help page describes the two approaches. The advanced route reports more; the basic route is simpler to defend if your privacy counsel prefers that nothing loads before consent. Decide with counsel, then read the rest of this piece with that decision in mind.
What changes in GA4
Users and sessions. With basic mode, GA4 loses every declined visitor. With advanced mode, cookieless pings arrive, but without a cookie each ping looks like a new user, so the raw counts are unreliable until behavioural modelling takes over.
Behavioural modelling has entry conditions. At the time of writing, Google’s documentation says a property must collect at least 1,000 events per day with analytics_storage denied for at least seven days, and have at least 1,000 daily users sending events with analytics_storage granted for at least seven of the previous 28 days, with the reporting identity set to blended. Check the behavioural modelling page for the current thresholds. A small store may never qualify, in which case advanced mode still helps Google Ads but GA4 shows observed data only.
Purchases and revenue. In advanced mode a purchase from a declined visitor still arrives as a cookieless ping with its value, so revenue totals hold up better than session totals. What the ping lacks is the identity that links it to the session and campaign that brought the shopper, so those purchases drift toward direct and unassigned in attribution reports. If direct traffic looks high or unassigned traffic grows after launch, this is the usual reason.
Exports. Modelled data appears in standard reports and explorations. The BigQuery export contains observed events only, so anything built on the export will be lower than the interface.
What changes in Google Ads
Conversion counts rise relative to a strict setup. Google models conversions it could not observe and includes them in the conversions column. Smart Bidding uses the modelled total. That is the intended behaviour, and it is also why Google Ads reports more conversions than the store in a way that is expected rather than wrong.
Remarketing lists shrink. Only visitors who granted ad_storage and ad_personalization are added, so audience sizes drop on the launch date and campaigns that depend on them may serve less. Budget that was spending on remarketing tends to move to prospecting on its own.
Enhanced conversions need ad_user_data granted. Hashed email from the checkout is only sent when that signal allows it, so the match rate you see in the conversion diagnostics falls to the consented share. The enhanced conversions entry covers what is sent and when.
If conversions fall rather than rise after launch, something in the implementation is wrong, and conversions dropped after consent mode walks through the usual causes.
Implementation details that move the numbers more than the mode
- Regional defaults. Consent defaults can be set per region. Denying by default only where the law requires it, and granting by default elsewhere, keeps observed data as complete as the law allows. A banner that denies everything for everyone reduces data in regions where it did not have to.
- Banner acceptance rate. The share of visitors who accept is the largest single determinant of how much observed data you keep. Banner wording, placement and the equal prominence of accept and reject are a design job with a measurable outcome.
- Load order. The consent default command must run before any Google tag loads, and
wait_for_updategives the consent platform a moment to respond before tags fire. Tags that fire before the default is set behave as if consent were granted, which is the most common audit finding. - URL passthrough and ads data redaction. Optional settings that pass click identifiers through URLs when cookies are denied, and strip identifying values from denied pings. Each is a decision to document.
- Server-side tagging changes nothing about consent. A server container honours the same signals. What it changes is browser blocking, which is a separate source of loss that consent mode does not address; server-side tagging explains the distinction.
How to read the reports afterwards
- Annotate the launch date in GA4 and in every dashboard that reads from it.
- Keep Shopify or your platform’s order count as the revenue of record, and compute the share of orders GA4 observes each month, before and after the change. The monthly reconciliation is the method.
- Read GA4 channel revenue as shares rather than totals, and expect the direct and unassigned shares to rise.
- Report the banner acceptance rate from the consent platform as a metric of its own, by region.
- Bid on Google Ads modelled conversions, and set budgets on blended numbers such as the marketing efficiency ratio that do not depend on the tag seeing every order.
An illustrative before and after
Say half of a store’s visitors are in regions where the banner defaults to denied, and six in ten of those accept. Observed data then covers all of the visitors who saw no banner plus six in ten of the rest, which is four in five sessions. Under the basic implementation, GA4 reports four fifths of the traffic and every channel share is computed on that sample. Under the advanced implementation with modelling eligible, GA4 reports something close to the full picture with roughly a fifth of it modelled, and Google Ads conversions include a modelled share of similar size. The proportions are invented to show the mechanics; a store’s own banner acceptance rate replaces them. Which attribution model to read the result through is a further choice, covered in which attribution model should an ecommerce store use, and the full picture of the reporting effects is in how does consent mode affect ecommerce reporting. The ecommerce analytics service handles the implementation and the reporting change together, because one without the other produces numbers nobody trusts.