12 Months On from Consent Mode v2: Does GA4 Still Work Effectively for Marketing Attribution?

Estimated reading time 11 minutes

TLDR: Google Analytics 4 (GA4) now combines observed measurement with modelling where direct observation is limited. Google's behavioural modelling is subject to eligibility thresholds, including at least 1,000 daily users who have granted analytics consent and sufficient denied events over the required periods. This means many lower-volume websites may not qualify for behavioural modelling. Across the client implementations we have analysed, our Website Attribution API has identified materially more attributable website activity than was represented within GA4, averaging 36% more attributed conversions in the analysed implementations. When that attribution is connected with calls, form fills and eCommerce transactions, marketers have another source of evidence for understanding performance and optimising advertising investment.

I’m not calling out Google…or am I?

Having spent my life at Hilton Hotels, Halfords and Virgin poring over marketing attribution data to work out which marketing activity worked in driving sales, Google Analytics was my go-to platform and source of truth for all my online reporting.

But ever since Google Analytics 4 (GA4) was launched, the reporting got way too complex to create, what was a simple data view became super complex and I was starting to question why our client sites were dropping traffic, but increasing sales.  Site conversion was going off the charts!

Having developed our marketing attribution software platform C360 to connect phone calls, form fills and eCommerce transactions with marketing source data, including ROAS down to keyword level where available, we were able to compare GA4 reporting with the attribution and conversion data we were observing elsewhere. That comparison raised some important questions, particularly around the growth of traffic classified as Direct. (Direct as a source on GA4 is automatically chosen when the source cannot otherwise be identified.)

Consent Mode 2 and the death of browser attribution

Browser restrictions, particularly when on an Apple device, fundamentally changed what marketing measurement can be seen through GA4. Google’s response to this gap in attribution was to move GA4 away from purely observed measurement towards a combination of observed data and modelling.

The regulatory environment around privacy and digital advertising has also changed significantly. Google’s fine from the EU is one example of the wider regulatory pressure facing major technology platforms. Separately, the EU's Digital Markets Act introduced additional obligations for designated gatekeepers, including requirements affecting how personal data may be combined or used for advertising purposes.

Consent Mode v2 added two important consent signals, ad_user_data and ad_personalization, alongside the existing consent controls. Its purpose is to allow Google's products to modify their behaviour according to a user's consent choices while retaining some measurement capability where direct observation is limited.

The aim is to make measurement more privacy-resilient in an environment where not every interaction can be directly observed.

In practice, this creates a significant challenge for many smaller sites because behavioural modelling has eligibility thresholds. Google Analytics 4 (GA4) requires at least 1,000 daily users who have granted analytics consent (analytics_storage='granted') for at least 7 of the previous 28 days, alongside Google's denied-event threshold and other model-quality criteria. In simple terms, many lower-volume B2B, regional and specialist websites may not generate enough eligible data for behavioural modelling to become available.

Where modelling is used, it should be understood for what it is: an estimate derived from available signals rather than a complete, directly observed count. Google's published behavioural modelling guidance explains the eligibility and modelling approach. For attribution decisions, marketers should therefore understand which parts of their reporting are observed and which are modelled.

Why does marketing attribution look so wrong in GA4?

You've got to read the detail behind GA4 to understand why reported source and visitor volumes can differ from other measurement approaches. For Google to model behaviour effectively, it needs a substantial amount of eligible data, and many sites simply will not generate that volume.

With an advanced Consent Mode implementation, Google tags can still load when consent is denied, but they operate without the normal persistent identifiers. Google describes the resulting signals as non-identifying information such as consent state and country, which can contribute to modelling. Google explicitly recommends allowing tags to load in this way because blocking or delaying Google tags until after the consent interaction reduces the accuracy of conversion modelling and makes GA4 behavioural modelling unavailable.

This distinction is critical.  Consent Mode does not magically reconstruct an individual customer’s journey.  It creates additional signals from which Google can model the behaviour that it cannot directly observe.

That does not make modelling inherently inaccurate. It does mean the result is an estimate rather than a complete observed record, which is an important distinction when attribution data is being used to make commercial decisions.

For large ecommerce businesses, those thresholds may be trivial.

For smaller B2B organisations, specialist retailers, regional businesses or websites with relatively low conversion volumes, they are anything but.

But won’t server attribution fix this?

You may have heard of server attribution before. Moving parts of measurement from the browser to a server-side architecture can improve control over first-party data collection, reduce the impact of some browser restrictions and ad blockers, and improve continuity of attribution signals. It does not, however, guarantee that every user or every original source can be identified, and it is not automatically more accurate in every implementation.

Moving tracking from the browser to the server can improve data quality and give organisations greater control over collection and processing. But unless that attribution is connected with actual business outcomes, such as calls, form fills, eCommerce transactions, applications or CRM sales, server-side tracking can still provide only part of the customer journey.

Server-side tracking also does not remove an organisation's privacy obligations. Whether consent is required depends on the technologies used, the information accessed or stored on a user's device, the purpose of the processing and the applicable requirements of UK GDPR and PECR.

There is a security consideration too, but an API is not inherently a security vulnerability. APIs and collection endpoints must be designed, authenticated, configured and monitored securely, just like any other internet-facing component.

The benefits of server-side measurement can therefore be significant, but the implementation still needs to address privacy, security and how attribution data will ultimately connect to business outcomes and advertising platforms such as Google Ads and Meta.

Website attribution without the modelling: welcome to our Website Attribution API

Faced with the challenge of generating more directly observed website attribution data to append to the conversions we track within C360, we created our Website Attribution API. It does not depend on GA4 behavioural modelling. Instead, it collects first-party attribution data within the client's website implementation and passes the relevant attribution information into C360, subject to the client's consent and privacy configuration.

This means organisations of different sizes can build a directly observed first-party attribution dataset and map it to conversions including calls, form fills and eCommerce transactions. Where the necessary privacy, consent and platform requirements are met, conversion information can also be returned to advertising platforms such as Google Ads and Meta to support campaign optimisation.

In practice, the results we have observed across analysed C360 client implementations have been striking. By connecting our attribution data with the conversions we track and using that information to improve campaign optimisation, we have seen four particularly interesting outcomes:

1. Website visitor measurement showed a material difference: Across the client implementations we analysed, the number of website users represented within GA4 averaged approximately 64% of those identified through our Website Attribution API, with some implementations closer to 50%, after known bot traffic was excluded. That does not mean GA4 is '64% accurate'. It means there was a material measurement difference between the datasets, and that difference deserves investigation before marketing decisions are made.
2. Direct attribution fell significantly in the implementations we analysed: Some clients had previously seen more than 40% of traffic reported as Direct within GA4. Within the C360 attribution data for those environments, genuine Direct traffic averaged approximately 8%, with more activity attributed to identifiable sources such as Google Ads, Google Organic and Meta Ads. These are client results, not a universal benchmark, but they demonstrate why a large Direct category deserves investigation.
3. Cost per conversion reduced by an average of 38% in the implementations where we measured subsequent optimisation outcomes: Better attribution allowed clients to connect calls, form fills and eCommerce transactions with the marketing activity associated with them and then optimise campaigns using that information. Across those analysed implementations, average cost per conversion reduced by approximately 38%. In one implementation, cost per conversion reduced from £151.21 to £20.33 within two weeks following improved attribution and campaign optimisation.
4. Sales increased by an average of 27% across the clients included in our analysis: With better attribution data, clients were able to optimise campaigns using a clearer view of which activity was associated with conversions. Across the analysed clients, sales increased by an average of approximately 27% following implementation and subsequent campaign optimisation. C360 is not a button that increases sales by 27%; the commercial value comes from the decisions made using better attribution evidence.

So what should marketeers do?

I'd never recommend that you stop using GA4 altogether because it remains useful for understanding onsite behaviour. But if a conversion tracking or attribution solution in your marketing stack depends heavily on GA4 data, you need to understand what is observed, what may be modelled and where other business conversion data can validate the picture.

I’d recommend some investigation first:

1. Define whether your site is likely to qualify for GA4 behavioural modelling: If you meet Google's eligibility thresholds, modelling may help fill some measurement gaps. If you operate a smaller online entity, review what GA4 is telling you and understand the extent to which your reports depend on directly observed versus modelled data.
2. What part of your marketing stack uses GA4 data? If you are using call tracking, a CRM, an eCommerce platform or another attribution product that consumes GA4 data, understand how those dependencies affect the reporting that ultimately reaches your marketing and board teams.
3. Are you seeing acquisition costs rise, visitor numbers fall or Direct traffic increase? Those movements may have several causes, but they are good reasons to compare GA4 with other first-party conversion and attribution evidence rather than assuming one dataset tells the complete story.
4. Are you under pressure to deliver more from the same budget? If you cannot connect marketing activity with the resulting calls, forms, eCommerce transactions and CRM outcomes, it is harder to build a confident case for where more budget should go. More than half of the C360 clients we have analysed subsequently increased their marketing budgets within six months of implementation, which we believe reflects greater confidence in linking spend to business outcomes.

Only when you understand where you are currently, can you then make an informed decision of what to do next to fix your marketing attribution problem.

The future is end-to-end customer journey visibility, with your CRM system being the single source of truth.

Twelve months on from Consent Mode v2, I think the marketing industry needs to move beyond treating GA4 as the only source of truth for visitor attribution.

Consent Mode v2 reflects a digital environment in which privacy choices increasingly affect what can be directly observed. For larger websites, Google's modelling capabilities may help address some of those measurement gaps. For organisations that do not meet the required thresholds, or where attribution needs to extend beyond the website into calls, forms, eCommerce transactions and CRM sales, additional first-party attribution may be required.

That is where we believe solutions such as our Website Attribution API and C360 can add value: connecting website attribution with real business conversions and, where appropriate, returning better-quality conversion signals to advertising platforms such as Google Ads and Meta.

Server-side tracking can also form part of that strategy, but no technology alone creates an end-to-end view of the customer. Attribution needs to connect the original marketing interaction to the resulting conversion and, ultimately, the commercial outcome recorded within the CRM.

Ensuring that your CRM contains reliable source data for each customer record is key to understanding what is working and what is not. GA4 can contribute to that picture, but it should not necessarily be the only evidence used to make the attribution decision.

The objective is not to replace GA4. It is to understand its limitations, combine it with other reliable sources of attribution and conversion data, and make better marketing decisions from the evidence available.