Conversion tracking and Consent Mode v2: why your data is missing
After consent arrived, some sites lost half their conversions and modelling will not return them to everyone. How Consent Mode works, where it is most often set up wrong, and how to tell whether your numbers still mean anything.
- Basic mode sends nothing before consent; advanced mode sends anonymous signals that modelling can work from.
- Modelling needs volume; vendors quote the order of a thousand events a day, which smaller sites do not have.
- The most important check is not in Google Ads: compare conversions with real orders in your own system.
Measurement is the most boring and most expensive part of PPC. Boring because nothing about it is visible. Expensive because a campaign with broken measurement does not get optimised — it only gets dearer. The algorithm dutifully produces more of whatever you labelled as success, even when it is not.
On top of that, since 2024 a layer of consent has cut off part of the data entirely. This article explains what is actually happening, and above all how to tell whether your numbers can still be trusted.
What Consent Mode v2 actually is
It is the mechanism by which your site tells Google what the visitor has consented to. It works with four parameters:
| Parameter | What it permits | What it affects |
|---|---|---|
ad_storage | Storing cookies for advertising | Conversion tracking from clicks, remarketing |
analytics_storage | Storing cookies for analytics | Data in GA4, telling new from returning visitors |
ad_user_data | Sending user data for advertising | Enhanced conversions |
ad_personalization | Personalised advertising | Remarketing and similar audiences |
Remarketing needs both ad_storage and ad_personalization. Enhanced conversions rest on ad_user_data. That is why a campaign can suddenly lose its audience without anything in it having changed.
Basic versus advanced mode
- Basic mode. Until consent is given, nothing is sent. No data, no basis for modelling. Simpler to deploy and to defend, but the loss is total.
- Advanced mode. Tags load immediately but store no cookies before consent and send only anonymous signals without identifiers. Google then models the missing conversions from them.
The difference in practice is dramatic — advanced mode is the only way to get part of the missing data back. It does, however, require careful deployment, because “tags load immediately” has to mean “and store nothing”.
How much data is lost
The figures circulating are uncomfortably scattered and mostly come from vendors of consent management tools — that is, from firms with an interest in the problem sounding large. Even so they broadly agree on the order of magnitude:
- One vendor reports a loss of 30 to 60 % of measured data after enabling Consent Mode v2 for sites without enough volume for modelling.
- A documented case from April 2026 describes a 90 % overnight drop in measured conversions after a consent misconfiguration. After the fix, about 40 % of attribution data returned through modelling; the rest was gone for good.
- Modelling needs volume. One vendor speaks of the order of a thousand events a day from both consenting and non-consenting users — more than smaller sites have.
They come from vendors, from different markets and with different methodologies; the ninety per cent drop is one specific case of misconfiguration, not a norm. What follows reliably is the direction: smaller sites lose more, because modelling needs volume, and a configuration error can be an order of magnitude worse than refusal of consent itself.
The legal side differs by country: ePrivacy rules are implemented nationally and enforcement sits with each national data protection authority. The specific interpretation for your situation is a legal question, not a marketing one — and this article does not replace it.
Where it is most often set up wrong
- The consent banner loads after the tag. Then the first pageview is measured regardless of consent, and your problem is somewhere other than you think.
- The default state is not “denied”. Before any interaction all parameters must be denied. Setting “granted” as the default is the most common serious error.
- Consent is not updated. The user clicks “accept”, but the signal never reaches the tags; data is missing even though you have consent.
- Some parameters are missing. Typically
ad_user_dataandad_personalizationadded later and only in some places. - The banner is designed so nobody consents. A dark “reject” button on top and a grey “accept” below is safe, but then do not be surprised by the numbers.
The check to run regardless of all of the above
This is the most important paragraph in the article. Independent of Consent Mode, attribution and modelling, one thing holds:
Compare the number of conversions in the ad platform with the real orders or enquiries in your own system, over the same period. If they do not match, everything else is inaccurate.
What to look for when they do not match:
- More conversions than orders: duplicate tracking (tag deployed twice), counting reloads of the thank-you page, counting cancelled and unpaid orders.
- Fewer conversions than orders: missing consent, script blocking, conversions off the site (phone, email, in store), purchase on another device.
- Counts match but values do not: value including VAT or shipping, returns not deducted, wrong currency.
For enquiry-driven sites one more question belongs here: how many enquiries turned into business. Without it you are measuring the cost of a contact, not the cost of a customer — see the difference between CPA and CAC.
What to do about the data loss
- Deploy advanced mode, correctly and verified. It is the only way to recover part of the data through modelling.
- Turn on enhanced conversions and send first-party data. It can return a substantial share of missing conversions. Detail in first-party data.
- Consider server-side tracking. More resistant to blocking and to browser cookie limits.
- Connect offline conversions. For services it is often the only route to the truth.
- Evaluate over a longer window. At lower volumes a weekly view is mostly noise.
- For big decisions, measure by experiment. A geo test works even when half your conversions are missing, because it measures total revenue rather than attribution.
What to take away
Complete data is not coming back. The point of measurement today is not perfect accounting of conversions but a stable enough basis for decisions and for algorithms to learn from. The minimum every account should have: correctly deployed advanced Consent Mode, enhanced conversions, one main conversion with a truthful value, and a monthly check against real orders. Everything else is an extra.
Common questions
What is Consent Mode v2 and do I need it?
It is the mechanism by which a site passes the visitor’s consent state to Google across four parameters: ad_storage, analytics_storage, ad_user_data and ad_personalization. For advertisers targeting users in the European Economic Area it is the condition for conversion tracking, remarketing and personalisation to work at all. Without it you lose both measurement and audiences. The actual legal obligations, however, follow from ePrivacy rules and the GDPR rather than from Google’s policies, and their interpretation for your situation is a legal question.
Why did my conversions drop after introducing consent?
Because without consent no cookies are stored and some conversions cannot be tied back to a click. Vendors report losses of roughly thirty to sixty per cent of measured data for sites without enough volume for modelling. If the drop is substantially larger it is usually not consent but a configuration error — a documented case from April 2026 describes a ninety per cent drop caused by misconfiguration, of which about forty per cent of data returned through modelling after the fix while the rest was permanently lost.
How do I verify my conversion tracking is correct?
Compare the number and value of conversions in the ad platform with the real orders or enquiries in your own system over the same period. More conversions than orders usually means duplicate tracking or counted cancellations; fewer means missing consent, script blocking or conversions happening off the site. If counts match but values do not, the usual culprits are VAT, shipping or undeducted returns. Run this check at least monthly, regardless of how well everything else is set up.
- Consent Mode V2 enforcement is silently breaking Google Ads conversions, PPC LandSource of the documented April 2026 case with a 90 % drop in measured conversions after misconfiguration and the subsequent return of about 40 % of data through modelling. A single case, not a statistic.
- Consent Mode v2 Data Loss: Why GA4 Traffic Dropped 30-60%, SeresaSource of the 30 to 60 % loss range for sites without enough volume for modelling. A vendor of consent management tooling, with a commercial interest in the problem sounding large.
- Google Consent Mode V2: EU Advertiser Setup Guide 2026, Audit SocialsSource of the four-parameter list and their link to remarketing and enhanced conversions. A vendor blog; verify technical detail in Google’s documentation.
- Google Consent Mode v2 Implementation: 2026 Guide, CookienoxDescription of the difference between basic and advanced mode. A consent management vendor.