Thursday, December 20, 2007

Implement Custom Process Types in Process Chains for CRM Analytics

Create up to four custom process types to automate and schedule your Analysis Process Designer models and Customer Relationship Management Analytics activities.

Key Concept

Analysis Process Designer models, data mining models, and Customer Relationship Management Analytics models (such as Customer Life Time Value and Recency, Frequency, Monetary) are not incorporated into BI process chains, and must be scheduled manually today. This article describes how to create a new custom process type for each of these items. A custom process type dictates the tasks and properties of a process. The result will be seamless automation and scheduling capabilities of these activities through process chains.

If you plan to start taking advantage of Customer Relationship Management (CRM) Analytics, data mining, or Analysis Process Designer (APD), you will find that you need to deal with automation and scheduling of your applications. Unfortunately, CRM Analytics and APD capabilities are not directly incorporated into process chains, so this can make scheduling a little complex. However, process chains offer a framework to create custom process types. I’ll show you how to create four separate custom process types for APD, data mining, Customer Life Time Value (CLTV), and Recency, Frequency, Monetary (RFM) models. I’ll explain each in detail.

APD is a graphical, drag-and-drop tool that allows you to create complex models to analyze and manipulate your data. The data mining capabilities of SAP NetWeaver BI (such as decision trees, clustering, and regression) are also exposed through APD.

Some of the data mining models (e.g., decision trees and clustering) need to go through a training process in which the system analyzes historical data to determine any relationships or patterns. The result of this training process is then used in the subsequent prediction phase. However, you want to make sure your models are as accurate as possible, and therefore, you need to retrain them with the most recent data from time to time. Before you can retrain your model, you need to remove the previously learned training results. This process type allows you to reset training results of a data mining model.

No comments:

Blog Archive