Predictive Audiences in GA4 Part 4

The following is part 3 in this series, if you need to catch up, here is part 1  part 2 and part 3 or download the complete guide.

November 10, 2022 -

The following is part 3 in this series, if you need to catch up, here is part 1  part 2 and part 3 or download the complete guide.

In this final installment we’re looking at why and how to use predictive audiences:

    Using Predictive Audiences

    Now that we know what predictive audiences are, what we can do with them, and how can they be helpful?

    There are two main use cases to leverage predictive audiences in marketing activities, Remarketing audiences, and re-engagement campaigns

    Remarketing audiences is a list of users grouped together who are likely to convert and more easily convinced to complete the conversion. Google Analytics 4 automatically analyses the data by applying machine learning algorithms on the data set and predicts the future behaviour of said user to find patterns of behaviour that are unique and the specific users likeliness to convert. 

    Re-engagement campaigns are designed for a predictive audience of users who are likely to churn. With this information, if there are a significant number of users who are likely to churn, you can consider running a re-engagement campaign to encourage these users to engage in the desired activity.

    Remarketing Audiences

    • Can be used to identify users who are likely to convert and easily convince to complete conversions.
    • Example:

    Users who have already visited the product details or added items to their carts, have given strong signals they will likely purchase the product.

    Re-engagement Campaigns

    • Use predictive audiences for users who are likely to churn.
    • Example:

    While users who are likely to churn are signalling a waning interest in your business, they have also previously demonstrated engagement with your business. Approach them again with reminders of the value you offer in terms of product variety, quality, and price, or convenient shipping and return options. Remind them of their value to you with special offers such as a promo code. 


    A Case Study

    Problem: McDonald’s Hong Kong wanted to optimise its mobile ordering journey & drive more in-app orders. They wanted to use machine learning solutions to gather customer insights and use them for their in-app ad campaigns.

    Solution: The team at McDonald’s implemented GA4 & started collecting real-time ecommerce data from their app. McDonald’s turned to predictive audiences, a tool that allows marketers to predict future purchase behaviour based on those ecommerce insights.

    McDonald’s tested a handful of GA4’s ready-to-use predictive audiences that met their prediction-modelling prerequisites. The “likely 7-day purchasers” audience ended up driving the most ROI and they decided to focus their ad investment on that segment.

    Thanks to integration between GA4 and Google Ads, those predictive audience segments could be exported to their ad campaign. McDonald’s used App Campaigns for Engagement, a fully automated Ads campaign type that prompts app users across multiple Google platforms to take action. Google Ads automatically created optimised ads for relevant audiences, and in tandem with insights on likely purchasers, it helped them drive more in-app orders. 

    Results: In just two months, the team increased their conversions 550% for “likely 7-day purchasers” and decreased cost-per-acquisition 63% among that group. On top of that, McDonald’s saw 560% increased revenue within that audience. 



    Q. Can we use historical data when creating predictive audiences? 

    1. No, the audience starts to collect data from the moment it is created, it does not apply retroactively. 

    Q. Can any type of audience be used for either a  remarketing campaign or a re-engagement campaign?

    1. Each type of campaign is to help increase the chance of a conversion, however each is designed for a specific type of audience. Remarketing campaigns are designed for users who you want to re-engage with because of their likeness to convert. You can use your predictive audience for users who are likely to convert and more easily convinced to complete the conversion. Re-engagement campaigns are designed to sway users who are likely to churn – you might use this information to change your tactics in order to engage with the audience. 

    Q. Is this completely AI or is there a human aspect involved?

    1.  This is AI trying to replace the human brain and is designed to take out the guesswork. Google’s machine learning determines who goes into the audience. A human could be involved if desired to tailor these audiences and create campaigns based on the audience, but it can also be used as is.

    Q. Would this be valuable to every type of business, whether it’s an e-commerce vendor or a content site?

    1. As predictive audiences are based on a purchase, this is most valuable to e-commerce sites as you can attribute dollar amounts and have insight into who is likely to buy, and what they are likely to buy. That said, there could be some value for non-ecommerce sites as predictive audiences can be used for form completes – who is most likely to fill out and submit a form, or most likely to read blog articles. 

    Q. What does the membership duration refer to?

    1. The number of days that users remain in the audience. Each time a user engages in behavior that meets the criteria for being included in the audience, then that user’s membership duration is reset to the full value of this option.

    Q. What are the reasons not to use predictive audiences?

    1.  As long as they are eligible to use it, I don’t see any opposition, especially if they use audiences already. It’s just as easy to use as using the audience in Google ads.

    Q. Because it’s a machine driven model, does that mean it can get smarter and smarter based on how the user uses it?

    1. Yes. Over time the predictions will get more and more accurate, and the patterns of behaviour will be easier to predict and therefore easier to market towards.  As it is now, you can say if somebody only visited 2 pages on the website, they aren’t as likely to come back, but these predictions become more accurate as predictive audiences get more data.

    Q. Is there a way to understand how much time this is saving? 

    There aren’t many companies who build their own audiences and if they do it, it’s not as accurate and would likely be a year or two old. Once you connect to google ads, the audiences will automatically appear in Google Ads. GA4 machine builds the audience and constantly upgrades the audience, and the campaign in theory should then create better ROI over time because the machine is making the audiences more and more accurate. So, the clients save time, are leveraging all of their up to date data, the machines are doing all the grunt work, and in theory, companies should be getting better results from their media spend. 

    Q.  Why should I care about this feature in GA4?

    1. One of the many reasons is that after the time you spent on setting up GA4, this easy-to-use feature, automatically connected to Google Ads will keep paying off as long as GA4 is used. 


    This is the last in our Predictive Audiences in GA4 series , download the complete Guide.

    Ready to implement and activate high-quality data to make superior business decisions with Napkyn?