Create a predictive model

Creating a Predictive Model in Demandbase. Multiple predictive scores are recommended when an organization has multiple products, different business units, or is exploring new markets.

To create a new predictive model, navigate to the Predictive Score Setup section and click the Create New button. This walkthrough covers creating a Pipeline Predict score, which measures the propensity of an account to become an opportunity soon.

Default settings are sufficient for most first-time setups, but the configuration includes several sections. The opportunity selection section requires choosing a date range with enough historical opportunities for the model to assess. For product-specific models, filter to only the opportunities relevant to that product using an opportunity-specific field, so the model trains on the most relevant data.

The Engagement Activities section is for advanced use cases. By default, no customization is needed here. However, activities can be excluded from the machine learning model using filters. For example, when building a score for Product A, website traffic to pages dedicated to Product B can be excluded to prevent unrelated activity from influencing that score. This filtering can be applied to any activity type, not just website traffic.

The final step is reviewing the Score Checklist. If there are enough accounts with preceding engagement based on the selected historical opportunities, the system enables saving and training the model.

After saving, Demandbase processes available data overnight using AI and machine learning to build the predictive model. Once the overnight sync completes, the new predictive scores can be used for marketing segmentation and sales prioritization.