A Predictive Insights model is a model that has been trained by our machine-learning engine. The algorithm detects patterns and relationships between attributes and customer or prospect labels in a dataset. This allows the algorithm to generate predictions or scores for new accounts.

Before you create a trained model, you need a Predictive Insight dataset. This dataset must meet the standard dataset requirements, as well additional ones for the training algorithm, namely:

  • The list of accounts in the dataset must contain the names of at least 100 existing and 500 prospective customers, but no more than 500,000. These names must be unique.
  • The module must contain a Boolean line item named Is Customer, where existing customers are flagged as TRUE, and prospective customers are flagged as FALSE.

The model assigns a score to each of your matched accounts.

  1. In Home > Predictive Insights, select Predictive Insights models in the left-side panel, then New model.
  2. Enter a Predictive Insights model name and select the Predictive Insights Dataset.
  3. Select Create model.

When the model is trained, it displays as Trained in the list of Predictive Insights models.