The Data Gap Problem

Privacy regulations have created a measurement gap for digital advertisers. In the European Economic Area, cookie consent rates typically range from 50 to 70 percent, meaning that 30 to 50 percent of website sessions generate limited or no conversion data. For advertisers relying on Google Ads, this gap directly impacts campaign optimisation, audience targeting, and return-on-ad-spend calculations.

Google's response to this challenge is conversion modelling — a machine learning approach that uses observed data from consented users to estimate conversions from non-consented sessions. When implemented through Consent Mode V2, this modelling can recover a substantial portion of lost conversion data while maintaining full regulatory compliance.

What Conversion Modelling Actually Is

Conversion modelling is not guesswork and it is not simple extrapolation. It is a machine learning system that analyses behavioural patterns from users who did consent to cookies and uses those patterns to predict the likely behaviour of users who did not consent.

The model considers signals such as:

• Time of day and day of week — conversion rates vary by time, and this pattern applies across consented and non-consented users.
• Device type and browser — conversion behaviour differs between mobile and desktop, and between browser types.
• Geographic region — conversion rates vary by location, and this remains true regardless of consent status.
• Page content and navigation patterns — what pages a user viewed and in what sequence can predict conversion likelihood.
• Referral source — the channel that brought the user to the site is a strong predictor of conversion intent.

By combining these signals, Google's models can estimate conversion counts with reasonable accuracy, giving advertisers a more complete picture of campaign performance.

How Cookieless Pings Work

The foundation of conversion modelling is the cookieless ping — a lightweight HTTP request sent to Google's servers when a user has denied consent. Understanding what these pings contain (and what they do not) is essential for both technical implementation and privacy compliance.

A cookieless ping includes:

• Consent state: The explicit information that the user has denied analytics_storage, ad_storage, or both.
• Page URL: The page the user is viewing.
• Timestamp: When the event occurred.
• User agent: The browser and device information (not unique enough to identify an individual).
• Functional information: Whether the page interaction was a page view, scroll, click, or form submission.

A cookieless ping explicitly does not include:

• Any cookie identifier (_ga, _gid, or any other).
• Any cross-site tracking information.
• Any personally identifiable information.
• The user's IP address in a form usable for identification (it is used only for coarse geolocation and then discarded).

These pings give Google enough contextual information to feed into conversion models without compromising individual user privacy.

Advanced Mode vs Basic Mode: A Critical Choice

Consent Mode V2 offers two implementation levels, and the difference between them is dramatic in terms of data recovery:

Basic mode sends absolutely no data to Google until the user grants consent. It is the simplest implementation — essentially just blocking Google tags until consent is given. The advantage is maximum simplicity; the disadvantage is zero data recovery from non-consented sessions. No cookieless pings means no modelling inputs.

Advanced mode sends cookieless pings for non-consented sessions while fully respecting the user's consent decision by not setting any cookies or collecting any personal identifiers. This is what enables conversion modelling. Google recommends Advanced mode and requires it for full Consent Mode V2 compliance.

The choice between these modes has a direct, measurable impact on your data. A site with a 60 percent consent rate using Basic mode sees only 60 percent of its conversion data. The same site using Advanced mode can potentially see 80 to 90 percent of estimated conversions through the combination of observed (consented) and modelled (non-consented) data.

Activation Thresholds: When Modelling Kicks In

Conversion modelling does not activate automatically for every website. Google requires minimum data volumes to ensure the models are statistically reliable. The key thresholds are:

• Google Ads conversion modelling: Approximately 1,000 ad clicks per day for at least 7 consecutive days. Below this threshold, there is not enough consented conversion data for the model to learn from reliably.
• GA4 behavioural modelling: Lower thresholds, but Google still requires a minimum of 1,000 events per day with consent granted for each event type being modelled.
• Consistency requirement: The thresholds must be met consistently. Sporadic traffic spikes followed by quiet periods will not activate modelling.

For websites that do not meet these thresholds, conversion modelling will show as unavailable in your Google Ads account. This is a common frustration for small and medium-sized advertisers, but it reflects a genuine statistical limitation — models trained on too little data would produce unreliable estimates.

How Much Data Can You Actually Recover?

Google's published data and independent studies suggest that conversion modelling through Consent Mode V2 Advanced mode can recover 50 to 70 percent of conversions that would otherwise be lost due to consent rejection. The exact recovery rate depends on several factors:

• Consent rate: Sites with higher consent rates provide more training data for the model, improving its accuracy for the non-consented segment.
• Traffic volume: Higher traffic means more signals and better model performance.
• Conversion type: Simple, high-volume conversions (like pageview-based goals) are easier to model than complex, low-volume conversions (like enterprise lead submissions).
• User behaviour diversity: If consented and non-consented users behave very differently, the model has a harder job. In practice, the main behavioural difference tends to be the consent decision itself, so models perform reasonably well.

To put this in practical terms: a site with a 60 percent consent rate losing 40 percent of its conversion data can expect to recover roughly 20 to 28 percentage points through modelling, bringing total observed-plus-modelled conversions to approximately 80 to 88 percent of the true total.

Impact on Smart Bidding and ROAS

Conversion modelling is not just about reporting accuracy — it directly affects automated bidding strategies. Google Ads Smart Bidding algorithms (Target CPA, Target ROAS, Maximise Conversions) use conversion data as their primary training signal. When conversion data is incomplete due to consent gaps, these algorithms make suboptimal bidding decisions.

Without modelling, Smart Bidding sees fewer conversions than actually occurred and may:

• Underbid on valuable keywords, losing impression share to competitors.
• Misattribute performance across campaigns, shifting budget away from effective campaigns that happen to have higher consent rejection rates.
• Report artificially low ROAS, potentially leading to budget cuts for campaigns that are actually profitable.

With conversion modelling active, Smart Bidding receives a more complete picture of true conversion volume, enabling it to bid more aggressively where warranted and allocate budget more effectively across campaigns.

Why This Matters for Advertisers Right Now

Google has made Consent Mode V2 a requirement for advertisers targeting EEA and UK users. Since March 2024, personalised advertising features for these audiences are only available to advertisers using a Google-certified CMP with Consent Mode V2. This is not optional — it is an enforced requirement.

Advertisers who have not implemented Consent Mode V2 with a certified CMP are already experiencing:

• Loss of remarketing audience data for EEA/UK users.
• No conversion modelling, resulting in larger data gaps.
• Degraded Smart Bidding performance.
• Reduced ability to measure and optimise campaign ROAS.

FlexyConsent is a Google-certified CMP that implements Consent Mode V2 in Advanced mode by default. It automatically manages all required consent parameters — analytics_storage, ad_storage, ad_user_data, ad_personalization, and functionality_storage — ensuring your Google Ads and GA4 configurations receive the signals they need to activate conversion modelling.

With IAB TCF 2.3 support, native integrations for WordPress, Shopify, and PrestaShop, and plans starting from EUR 0 per month, FlexyConsent removes both the technical and financial barriers to proper Consent Mode V2 implementation.

Key takeaway: Conversion modelling is not a nice-to-have — it is the mechanism that bridges the gap between privacy compliance and effective advertising measurement. Without it, you are making bidding and budget decisions based on incomplete data. With Consent Mode V2 Advanced mode and a Google-certified CMP, you can recover 50 to 70 percent of lost conversions and give your Smart Bidding algorithms the signal they need to perform optimally.