This article provides an overview of the Tealium Predict ML workflow, basic Predict implementation, and best practices for readying your data and creating models.

In this article:

How it Works

If you have the Tealium AudienceStream CDP product with data collection enabled, you have the ability to begin using Tealium Predict. We recommend reviewing our Prerequisites to ensure you have enough event volume to create and train a model for the outcome you are trying to drive.


Use the following steps to get started with Predict:

  1. Select
    Select the target attribute that represents the visitor behavior you want to predict. The target attribute can be any Badge or Boolean attribute in Tealium AudienceStream CDP.  If your target attribute is not ready for modeling, see strategies for getting your data ready for training.
  2. Train
    After your target attribute is ready for modeling, create a model and start training. While creating your model, an output attribute (visit-scoped number attribute) is created for you that will store prediction values for your model after you deploy it. You define the time window within which the user must return and complete the target attribute action for model training.
  3. Review
    Predict model scores and visualizations help you review how well your model is performing so you can determine whether you need to retrain your model.
  4. Deploy
    After you are satisfied with how the model is performing, you can confidently deploy it. Your deployed model stores the Prediction Value (the likelihood that this visitor will return and complete the target) in the output attribute for the model. 
  5. Create
    With your deployed model, you can use the Prediction Value to help you create audiences to better target your visitors and create personalized experiences to drive ROI.

Best Practices

The following best practices can help you get started selecting your target attribute, readying your data, and using the models that you create with Predict:

Getting Started

  • Define the problem you want to solve and goals to achieve resolution.

Working with your data

  • Ensure that your AudienceStream profile has high-quality attributes to signal target behaviors, such as Badges or Booleans. If none are present, create the attributes as soon as possible to allow more time for data to accumulate.
  • Establish best practices regarding collecting and cleaning your data before you begin. For more information on preparing your data before creating a model, see Data Readiness.
  • During your data readiness stage, join siloed datasets and consider other characteristics of your organization's data that can be refined.

Training and Deploying Models

  • Where possible, use longer date ranges for training.
  • Deploy your model in a production environment or real-world application.
  • Evaluate how well your model is working in production and return to the 'training and testing' stages as needed for improvements.

Creating Audiences

  • Consider ways to use Tealium Predict to improve or augment existing audiences and make them more efficient.