ON THIS PAGE
- What Is Conversion Modeling?
- How does Conversion Modeling Work?
- Observed Data vs Training Data vs Modelled Data
- Differences Between Conversion Modeling and Behavioral Modeling
- Benefits for Marketers of Conversion Modeling
- Best Practices for Conversion Modeling
- How to Start Using Conversion Modeling?
- Frequently Asked Questions
Starting in March 2024, Google switched to Google Consent Mode v2 to comply with the privacy laws in the European Economic Area (EEA) and the United Kingdom. Read more about Google Consent Mode v2, how to implement it, and how to choose a Google-certified CMP partner for ad complianceow to choose a Google-certified CMP partner for ad compliance.
This privacy-first approach affects heavily digital marketing. When a user doesn’t consent to ads cookies or analytics cookies, Google Consent Mode automatically adjusts the relevant Google tags’ behavior to not read or write cookies for advertising or analytics purposes. This enables advertisers to respect user privacy choices. Without cookies, advertisers experience a gap in their measurement and don’t get user paths on their websites. They are no longer able to directly tie users' ad interactions to conversions and don’t know whether the users are repeat visitors or whether they have arrived from paid or organic traffic sources. Thus, this gap in advertisers’ measurement does not allow you to evaluate the efficiency of your marketing campaign and could decrease the conversion rate.
Read this blog to learn how to increase your website conversion rate using Google Consent Mode v2.
What Is Conversion Modeling?
Google introduced conversion modeling to fill in data loss in digital marketing measurement when it’s not possible to observe the path between ad interactions and conversions.
Google Consent Mode conversion modeling uses Google AI to analyze observed conversions and to predict lost conversions without identifying any individual user.
Results from Google Ads have shown that, on average, conversion modeling could recover up to 70% of paths between ad interactions and conversions lost due to user Cookie Consent choices. Keep in mind that these results may vary widely, depending on user cookie consent rates and the advertiser’s Consent Mode setup.
Google has found that user conversion rates vary based on user consent status. Consented users are typically 2-5 times more likely to convert than unconsented users. However, this varies widely depending on consent rates, industry, and conversion type.
Conversion modeling does not use individual user data. Instead, Google algorithms are used to estimate the likelihood of conversions based on aggregated data, such as historical conversion rates, location, device type, browser, etc. Thus, there are no privacy issues for non-consented users (opted-out users). This is a preparation for a cookie-less future when third-party cookies are supposed to go away in near future.
Conversion modeling is a tool, developed by Google, to overcome the restrictions to limit the usage of cookies while continuing to enable marketers to measure their marketing performance.
Scan your website for free to see all your website cookies, local storage, and session storage in use.
How does Conversion Modeling Work?
When a user doesn’t consent to advertisement cookies or analytics cookies, Google Consent Mode v2 adjusts the relevant Google tags’ behavior and does not read or write cookies for advertising or analytics purposes. This leads to a gap in advertisers’ measurements. Conversion modeling solves this problem using observable data and AI.
Google AI analyzes observable data and historical trends from consented users to get behavioral data and quantify the relationship between consented (opted-in) and unconsented (opted-out) users. Then, using observed data from users who have consented to cookie usage, Google AI models assess attribution paths for unconsented users. In this way, conversion modeling allows to link between ad interactions and conversions while respecting user consent choices.
Convention modeling works by performing the following steps:
- Data collection. Google AI collects and analyzes data from consented users who have consented to cookies and other tracking technologies. This includes ad interactions, user paths on the website, the website’s response, and conversions like purchases or sign-ups.
- Finding the pattern. Google AI searches this data for common patterns and relationships between consented and unconsented user paths. It tries to find a typical pattern of consented user behavior before converting, considering factors like demographics, browsing habits, visited web pages, and time spent on specific pages.
- Predicting the behavior. Based on the behavior and activity patterns of consented users, Google AI tries to predict the behavior of unconsented users. It compares the behavior of consented and unconsented users, tries to find similarities, and calculates the probability of conversion of unconsented users. This approach allows us to predict the behavior of unconsented users even without direct user tracking.
Conversion modeling does not change the total number of conversions collected by Google Analytics 4 (GA4), but it changes the channels that those conversions came through. For example, conversions could be attributed to these channels: direct entry, paid search, organic search, email, referral, etc. After conversion modeling, different channels will be attributed to conversions to reflect the real situation of user behavior and patterns for conversion.
Observed Data vs Training Data vs Modelled Data
There are three categories of data used for conversion modeling:
- Observed data
- Training data
- Modeled data.
Observed data
Observed data comes directly from users who have consented to use their data for analytical and advertising purposes using cookies or app IDs. This type of data provides reliable information about user behavior like the number of users, sessions, page views, events, conversions, and total paths to conversions.
Training data
Training data is a combination of observed data and labeled data used to train Google AI algorithms behind modeled data.
Labeled data is raw data that has been assigned labels to add context or categories, which is used to train machine learning models in supervised learning.
Examples of labeled data in GA4:
- Specific events like “Add to cart” and “Checkout completed” are labeled as conversion steps to train the algorithms about the user conversion funnel.
- A high number of page views and long average session time are labeled as engaged users to predict future engagement behaviors of unconsented users.
- User demographics and purchasing history are labeled for user categorization to improve user segmentation and personalization.
The training data has a huge impact on the effectiveness of modelled data.
Modeled data
Modeled data is the estimated data through conventional modeling for unconsented users (opt-out users). Google machine learning algorithms analyze behavior from consented users and use these patterns to predict the behavior of similar unconsented users. It means that modeled data, obtained through conventional modeling, can provide realistic results of the behavior of unconsented users without collecting data from them. Keep in mind that it’s an estimation and in some cases may not be as precise as observed data. However, modeled data gives us much more information than not having data at all.
Differences Between Conversion Modeling and Behavioral Modeling
Both behavioral modeling and conversion modeling are used when a user did not grant user consent for analytical or advertising cookies and the user could not be tracked.
Behavioral modeling estimates user behavior when they don't consent to cookies.
Conversion modeling estimates user behavior patterns and engagement on the user paths on the website, which allows us to evaluate the performance of marketing when there is a loss of conversions for unconsented users.
Benefits for Marketers of Conversion Modeling
Modeled conversions provide marketers with a more accurate view of their marketing campaign performance.
For example, you might have spent $1,000 on a paid search campaign and received in total 100 observed conversions. 50 of these observed conversions could be attributed to your marketing campaign, and another 50 - to direct entry. This would mean that you paid $20 per conversion.
After conversion modeling evaluates the data, the total number of conversions remains the same- 100. It could find that 80 of these observed conversions could be attributed to your marketing campaign, and just 20 of them- to direct entry. In this case, it would mean that you paid $12.5 per conversion.
This would make a huge difference in the effectiveness of the marketing campaign.
In general, conversion modeling helps to understand user behavior and allows companies to:
- Increase the conversion rate. Understanding conversion paths across devices and channels resulting from ad interactions helps to identify barriers, overcome them, and improve the conversion rate.
- Improve automatic bidding. Conversion modeling fills a data loss of unconsented users. This helps to improve automated bidding since it relies on accurate information about website or app activity rather than assumptions.
- Improve their marketing ROI. When you get more accurate results of your marketing campaign, you can make data-driven decisions, update your marketing strategy, and spend more efficiently your marketing budget by targeting customers who are more likely to convert. This will lead to higher conversion rates.
- Get accurate yet privacy laws-compliant data. Conversion modeling allows us to get accurate data while complying with privacy laws in the EEA and the UK at the same time.
- Be competitive. Many companies will struggle due to data loss once Third-Party Cookies go away. By implementing conversion modeling now, you can get a competitive advantage and set yourself up for success.
Best Practices for Conversion Modeling
Successful conversion modeling is essential for getting reliable data, optimizing marketing strategies, and improving conversion rates. While there are no prerequisites for conversion modeling, follow these best practices to employ conversion modeling to increase your website conversion rate:
- Enable Google Consent Mode v2 across all pages or app screens.
- Collect at least 700 ad clicks over 7 days for a domain per country.
- Collect at least 300 conversions per month across the entire website.
How to Start Using Conversion Modeling?
You can’t enable or disable conversion modeling in Google Analytics 4. Google Consent Mode conversion modeling is integrated directly into advertisers’ Google Ads campaign reports- you can’t choose whether to use it or not. Reports that use conversion modeling aren’t distinguishable from those that don’t.
So, GA4 automatically applies conversion modeling to data that meets the criteria defined above, you don’t have to do anything to use it in GA4. You just need to collect a sufficient number of ad clicks and conversions to get accurate modeled data.
Frequently Asked Questions
What Is Conversion Modeling?
Google introduced conversion modeling to fill in data loss in digital marketing measurement when it’s not possible to observe the path between ad interactions and conversions for unconsented users. Google Consent Mode conversion modeling uses Google AI to analyze observed conversions and to predict lost conversions without identifying any individual user. Use CookieScript CMP to implement Google Consent Mode v2 and use conversion modeling.
How does conversion modeling work?
Google AI analyzes observable data and historical trends from consented users to get behavioral data and quantify the relationship between consented and unconsented users. Then, using observed data from consented users, Google AI models assess attribution paths for the unconsented users. In this way, conversion modeling allows to link between ad interactions and conversions while respecting user consent choices. You also need to use Google Consent Mode v2. CookieScript CMP is a Google-certified CMP, recommended by Google, for the implementation of Google Consent Mode v2.
What are the categories of data used for conversion modeling?
There are three categories of data used for conversion modeling: observed data, training data, and modeled data. Observed data comes directly from consented users. Training data is a combination of observed data and labeled data used to train Google AI algorithms. Modeled data is the estimated data through conventional modeling for unconsented users.
What are the differences between conversion modeling and behavioral modeling?
Behavioral modeling estimates user behavior when they don't consent to cookies or app IDs. Conversion modeling estimates user behavior patterns and engagement on the user paths on the website, which allows us to evaluate the performance of marketing when there is a loss of conversions for unconsented users. You need to implement Google Consent Mode v2 for both. CookieScript CMP is a Google-certified CMP, recommended by Google, for the implementation of Google Consent Mode v2.
Are there prerequisites for conversion modeling?
While there are no requirements conversion modeling, follow these best practices for an effective conversion modeling: enable Google Consent Mode v2 across all pages or app screens, collect at least 700 ad clicks over 7 days for a domain per country, and collect at least 300 conversions per month across the entire website. CookieScript CMP is a Google-certified CMP, recommended by Google, for the implementation of Google Consent Mode v2.
How to Start Using Conversion Modeling?
Google Analytics 4 automatically applies conversion modeling to data that meets the criteria defined above, you don’t have to do anything to use it in GA4.
What countries is consent mode v2 available in?
Due to Google's explicit consent requirement, Google Consent Mode v2 only works in the EEA, and the UK if you use a Google-certified CMP and your website displays cookie consent for those visitors. CookieScript CMP is a Google-certified CMP, recommended by Google.