Targeting solution based on a predictive model of users who are likely to take a specific action, based on advertiser's source data - pixel or S2S event.
Overview
Predictive Audiences help you discover untapped, high-converting customers at an efficient CPA. By mirroring the actions of your source data—such as past conversions—these AI-driven segments identify new users with a higher likelihood of taking action.
You also have the flexibility to optimize the audience size according to your specific scale and performance objectives.
Launch a new campaign today and pair Predictive Audiences with Maximize Conversions to drive incremental revenue and scale effortlessly.
How Does It Work?
Powered by advanced machine learning, Predictive Audiences are based on a model that analyzes your advertiser's audience, via a pixel seed or S2S event to identify key traits that drive actions and aims to find similar users across the network who are most likely to convert.
You can create a Predictive Audience from any of the following conversion types: ‘add payment info’, ‘add to cart’, ‘add to wishlist’, ‘app install’, ‘complete registration’, ‘contact’, ‘lead’, ‘make purchase’, ‘sign up’.
It’s therefore not recommended for advertisers optimizing towards page views or search click-out conversions.
Note: If you created a conversion event labeled as ‘other’, it will not be possible to create a Predictive Audience from it.
Step by step Guide
Navigate to the Audiences tab in Realize and click on + New Audience
Select Predictive Audiences as the audience type and click Continue
In the audience creation screen, select Pixel Audience under Seed Selection
Select the seed audience you wish to use to build the predictive audience from the drop-down menu, then click on Create Audience to save the audience segment
Pixel-based predictive audiences will take up to 48 hours to populate. Once the status of the audience segment changes to “Ready” in the Predictive Audiences tab, you can target it in your test campaign under My Audiences.
When Should I Use Predictive Audiences?
Predictive Audiences should be used to reach new audiences at a stronger CPA. A Predictive audience can be created as soon as 100 conversions have been triggered. These conversions must be generated from your activity on Realize, not any other platform.
It’s also important to continue to fuel the Predictive Audience with conversion data, otherwise the audience may start to fatigue and you may see a decline in performance. It is therefore recommended to always run a Predictive Audience campaign alongside other prospecting or evergreen campaigns that continue to generate conversions.
FAQs
What is the difference between lookalike audiences based on CRM files and Predictive Audiences?
A: Predictive audiences focus on leveraging real-time user behavior to predict who is most likely to convert, while CRM lookalike targeting focuses on expanding the customer base by finding new users who resemble a company’s existing customers based on historical CRM data. Note: CRM lookalike is only available in the US
How do Predictive Audiences work with Max Conversions. Do they work together? What’s the difference between users/cohorts/outcomes they target?
A: Maximize Conversions is still the recommended bidding strategy when it comes to targeting Predictive Audiences. Predictive Audiences is a targeting solution that finds users who have a high likelihood of taking a specific action, and Maximize Conversions ensures you are paying the right price for that user based on existing conversion data.
What tier should I use when targeting a Predictive Audience?
A: You can select the size of the audience you want to target after you have created the Predictive Audience from tier 1 to 5. The lower tier will result in strongest performance as you are reaching the highest percentile of the predictive audience. It is recommended to start with this tier and expand to a higher tier if you require further scale.
What should I do if I notice my Predictive Audiences campaign isn’t scaling or spending as expected?
A: 1. Check the daily budget relative to the CPA goal. Low budgets with high CPAs may limit the algorithm's ability to learn.
2. Consider targeting a higher tier unless higher precision is needed.