Planning for the Predictor Resource
You use the Predictor resource primarily to augment analysis of some form of data, such as transaction data or Web log data. The Predictor resource is also used to recommend products to users online and to fill in missing user profile properties.
Before you use the Predictor resource, you must verify that the Analysis reports are providing the information you require. You should collect data on at least a few thousand distinct users before you build models with the Predictor resource.
The following table lists some of the questions that you need to answer in planning for the Predictor resource.
| Planning question | Recommendation |
| Do you plan to offer users product recommendations and enhanced targeting? | Implement the Predictor resource on your site by building models of user behavior after you have collected enough initial data in the Commerce Server Data Warehouse. |
| What are the key user segments that you must identify in order to make meaningful recommendations? | Use the Predictor resource to identify meaningful segments, and then view them with the Segment Viewer. |
| Will you use inference to augment user profile data, or will you make recommendations based only on data explicitly provided by users? | You can use the Predictor resource to infer data about users for whom you do not have complete information (for example, guest users). If you plan to infer user properties, use the Segment Viewer in Business Desk together with the Dependency Network view in Commerce Server Manager to focus on the key user attributes you will use to make predictions. You can then use these attributes to target content to users who have similar characteristics. |