Create and Set Up a CRM Analytics Template Configuration for Predictions

Create and set up a template configuration to build customizable CRM Analytics apps that generate predictions for your scoring use case.

• Create a CRM Analytics Template Configuration
  Create a CRM Analytics template configuration to generate prediction scores for your use case. The template configuration creates a CRM Analytics app, an Einstein Discovery model, and preconfigured recipes based on the settings defined in the template configuration.
• Select an Object for Training and Scoring
  Select a standard or custom object with data to train your model and get predictions.
• Select a Prediction Duration and Historical Datasets
  You can generate prediction of an event happening during a specific time period. The model determines historical trends based on the datasets that you provide.
• Add a Dataset to Fine-Tune Your Predictions
  To analyze your data better, you can optionally include input features that are in a CRM Analytics dataset.
• Define the Target for the Prediction
  Define the variable to use as your model’s primary focus for analysis and predictions. The scoring model uncovers relationships between the input features and your target variable, and provides insights about maximizing or minimizing your target variable.
• Select Input Features to Get Accurate Predictions
  An input feature is a field value that can influence prediction. You can select input features from a CRM Analytics dataset, a predefined set of fields, or from the object that you selected for training and scoring.
• Save Assets to Debug Analytics App Installation Failures
  Save your CRM Analytics app assets within the app to maintain access to the assets even if the analytics app installation fails. Use the saved assets to easily debug app installation failures.
• Filter Your Training and Scoring Datasets
  Focus your predictions on a specific subset of the object data by defining filter conditions. Get more useful predictions by excluding records irrelevant to your prediction. Ensure that you have sufficient records in your training and scoring datasets by counting the records. Filtering is optional. You can use the complete dataset for training and scoring.
• Store Predictions in Records
  View your generated predictions contextually by storing them in records. To write back a prediction, select a preconfigured output connector and then select an object and field to store the prediction. You can write back predictions only if you’ve trained and deployed the model for a template configuration.
• Show Real-Time Predictions and Next Best Action Recommendations by Using AI Accelerator
  Show prediction scores, suggestions and insights about prediction scores, and Next Best Action recommendations for your batch and real-time use cases by creating an AI Accelerator use case. Customize the configuration of the AI Accelerator use case by defining your own feature extraction settings and Einstein Discovery model. You can create an AI Accelerator use case only if you’ve trained and deployed the model for a template configuration.