Privacy Analytics: How to Measure Without Losing Insight
Most websites treating privacy analytics as a compliance checkbox are missing the bigger picture. The real question isn’t…

If your site runs Google Analytics or Google Ads on European traffic, Google Consent Mode v2 has already changed how your data works — whether you’ve implemented it or not. The March 2024 enforcement deadline came and went. Sites that missed it are now running Google Ads without modeling, building remarketing audiences that quietly stopped refreshing, and reporting GA4 conversion data that has a growing blind spot.
This article covers the technical side: what the v2 signals actually do, why basic mode leaves money on the table, how to wire up consent initialization in GTM, and what your GA4 reports look like after. If your question is more “do I even need a consent banner on my site?” — that’s a separate legal question, and I covered it fully in the cookie consent banner decision guide.
Consent Mode is a Google framework that tells your Google tags how to behave when a visitor hasn’t consented to cookies. Instead of the old binary — fire or don’t fire — it lets tags operate in a restricted, cookieless state. The difference in v2 is two new signals that directly address the EU’s Digital Markets Act.
The original Consent Mode shipped with two signals: ad_storage and analytics_storage. Version 2 adds ad_user_data and ad_personalization. Those two additions aren’t cosmetic. The EU Digital Markets Act (DMA), which designated Google a gatekeeper in 2023 and set a March 2024 compliance deadline, explicitly requires it to obtain user consent before combining data across its own services. The new signals are the mechanism for communicating that consent.
If you only have the original two signals in place, you’re still running Consent Mode v1. That’s not compliant with current Google Ads requirements for EEA traffic.
Each signal maps to a specific data behavior. Here’s the full picture:
| Signal | Controls | When denied |
|---|---|---|
ad_storage |
Advertising cookies (_gcl_au, conversion cookies) |
No ad cookies written to browser |
analytics_storage |
Analytics cookies (_ga, _gid) |
No analytics cookies; cookieless ping still fires |
ad_user_data |
Whether personal data is sent to Google for advertising purposes | No personal data leaves for Google ad systems |
ad_personalization |
Remarketing and personalized advertising | User excluded from remarketing audiences |
The first two existed in v1. The new pair — ad_user_data and ad_personalization — closes the gap the DMA identified: a user could technically block ad cookies but still have their data shared with Google’s ad infrastructure. These signals prevent that.
All four should default to denied for EEA visitors. When a visitor accepts cookies through your consent banner, the signals update to granted and your tags begin operating at full capacity.
There are two overlapping regulatory pressures here. The GDPR requires explicit consent before setting analytics or advertising cookies — that’s been the case since 2018. What changed in 2024 is the DMA specifically requiring Google to get verifiable consent before combining user data across its products (Search, YouTube, Maps, Analytics, Ads). Google’s response was to make Consent Mode v2 mandatory for any publisher using its advertising products in the EEA.
Practically speaking, if you run Google Ads and don’t implement v2, your remarketing audiences stop updating and your conversion modeling degrades. Google’s own documentation is explicit: without the ad_user_data and ad_personalization signals, Google can’t use EEA user data to power Smart Bidding. That directly hits campaign performance.
For analytics specifically, the consequence is less dramatic but still real. Without Consent Mode, you’re choosing between blocking GA4 until consent fires (losing all pre-consent data) or firing it without consent (GDPR violation). The framework gives you a legal middle path. How you structure that choice — and whether you legally need consent at all — depends on your setup; the consent banner decision guide walks through that.
This is the decision that most implementations get wrong, or at least underexplained. The names are a bit misleading — “Basic” doesn’t mean simpler to implement; it means less data.
| Feature | Basic Mode | Advanced Mode |
|---|---|---|
| Tags fire before consent | No — tags blocked until consent granted | Yes — tags fire immediately in restricted state |
| Cookieless pings sent | No | Yes (for non-consenting users) |
| Behavioral modeling in GA4 | Not available | Available (requires minimum traffic threshold) |
| Conversion modeling in Ads | Not available | Available |
| Data from non-consenting users | Zero | Anonymous, modeled estimates |
| Best for | Sites with minimal analytics needs, very conservative privacy stance | Most sites running GA4 and/or Google Ads |
With Basic mode, Google tags do nothing until the visitor clicks “accept.” Simple, conservative, but you lose all insight into users who decline or dismiss the banner. On a European site — in the EEA projects I’ve run, consent rates land somewhere around 50–70% — that’s a significant chunk of traffic you’re flying blind on.
Advanced mode fires tags immediately, but in a restricted state: no cookies are written, and only anonymized, non-identifying signals are sent. That cookieless ping is what enables Google’s modeling. The trade-off is that your tags technically execute before explicit consent — which is why your privacy legal setup needs to account for this, and why the cookieless ping alone is designed not to carry personal data.
For most sites running GA4 and any Google advertising, Advanced mode is the right choice. The data recovery through modeling is substantial, and the cookieless pings themselves don’t trigger GDPR requirements.
The first time I wired up Advanced Consent Mode in GTM, the modeled conversions didn’t appear in Google Ads for two days after the launch. Nothing was broken — there’s just a processing lag before Google has enough cookieless signals to begin modeling. If you don’t know to expect that, it looks like the setup failed.
Here’s how the implementation works end-to-end:
denied by default for EEA:
gtag('consent', 'default', {
'ad_storage': 'denied',
'analytics_storage': 'denied',
'ad_user_data': 'denied',
'ad_personalization': 'denied',
'wait_for_update': 500
});
The wait_for_update parameter (in milliseconds) tells Google tags to pause before firing while your CMP loads. 500ms is a common value; adjust based on your CMP’s load time.
consent update commands automatically when a user makes their consent choice. Search for your CMP’s name directly in the gallery.
analytics_storage; “Marketing” or “Advertising” cookies map to ad_storage, ad_user_data, and ad_personalization. The exact category names depend on your CMP’s setup.
analytics_storage. Google Ads Conversion Tracking needs ad_storage and ad_user_data. If a tag doesn’t have consent requirements set, it may fire regardless of consent state.
If you’re not using GTM, you can implement the consent defaults directly in your gtag.js snippet. The Google developer documentation has the direct gtag() approach. If you’re using a data layer setup, make sure your consent defaults load before any other data layer pushes — order matters here.
A misconfigured Consent Mode is worse than no implementation, because you think you’re covered when you’re not. Test every change:
collect or g/collect. Before consent, look for gcs=G100 in the query string (all denied). After consent: gcs=G111 (all granted). Any other value warrants investigation.Always test in a fresh incognito window or a browser profile where you’ve cleared cookies. A cached consent acceptance will make everything look fine even if your initialization tag is broken.
This is where honest expectations matter. After implementing Advanced mode, your GA4 reports won’t look the same.
When a user denies consent, GA4 still receives a cookieless ping — a minimal, non-identifying signal with page URL, timestamp, and user agent. Google’s machine learning uses these signals, combined with behavioral patterns from consenting users, to estimate what non-consenting users did. This is called behavioral modeling. The estimates show up blended into your standard reports alongside observed data.
A few things to know about how this modeling works:
From the sites I’ve helped through this transition, my observation is roughly 20–40% of EU traffic coming through modeling when consent rates are typical — though that varies significantly by market. Germany in my experience tends toward 30–50% opt-in; France and the Nordics often run higher. These are field observations, not published benchmarks from Google. The modeled data won’t be as precise as observed data, but it’s far better than the zero you were getting from those users before.
This is also where the GA4 side of consent mode connects to larger data strategy. If you’re running a privacy-conscious stack and considering first-party data as a fallback, I wrote about that in the first-party data strategy guide. And if you’re evaluating whether server-side tracking might complement or replace some of this client-side setup, that’s a different trade-off worth understanding.
Consent Mode is not a magic compliance solution. A few things it doesn’t do:
If your site gets minimal EU traffic and you’re running a simple setup, Basic mode is fine. You’re just accepting the data loss in exchange for a simpler, unambiguously conservative implementation.
Google Consent Mode v2 adds two signals that close a genuine gap in the original framework — ad_user_data and ad_personalization exist because the DMA specifically required them. If you’re using Google Ads in the EEA and haven’t added those signals, your campaigns are already missing optimization inputs.
Advanced mode is worth the extra setup effort for any site running GA4 alongside advertising. The cookieless pings and behavioral modeling recover meaningful signal from non-consenting users without touching their personal data. Basic mode is simpler and still compliant — it just costs you that recovery.
The GTM implementation isn’t complicated once you know the sequence: Consent Initialization tag first, CMP template second, consent mapping third, then tag-level consent requirements. Test with GTM Preview, verify signal states, and check the GCS parameter in your Network tab. When the modeled conversions take a day or two to appear in Google Ads after launch — that’s normal. Give it 48 hours before troubleshooting.
First published . Last updated .