Optimal CRM process: Measure, predict, act
Alex Berson, Stephen Smith, Kurt Thearling
CRM is new enough that there are still a lot of theories bouncing around. This tip from Alex Berson, Stephen Smith and Kurt Thearling's book Building Data Mining Applications for CRM, McGraw-Hill, discusses some of those theories and urges CRM managers to prove the theories by collecting data for every CRM decision made in the organization.
Optimizing the CRM process requires a business best practice that consists of three main steps and an architecture for implementing and supporting those steps. The steps are formally named: "measure," "predict," and "act." They represent the steps on a cycle of customer relationship management that is continuously improving. The methodology is general enough that it is applicable to the vast majority of customer management functions such as the following:
- Cross-selling. Selling a new product to an existing customer
- Acquisition. Acquiring new valuable customers
- Retention. Retaining existing valuable customers
To achieve these goals of better customer management, there are a variety of processes that are used over and over again by the marketing organization:
- Targeted Marketing
- Lifetime value
Requires Free Membership to View
When you register, you'll begin receiving targeted emails from my team of award-winning editorial writers on the latest customer relationship management (CRM)and call center technology issues today. Our goal is to keep you informed on the hottest issues facing this fast-changing industry.
Hannah Smalltree, Editorial Director
- prediction
- Channel management. Matching the channel to the customer in the most profitable way
In marketing, there typically exists a certain set of interventions that can be preformed against the customer or prospect. Often, this is a fixed list that has been handed down from marketing personnel. Usually, these are new creative offers that have been created based on feedback from customers, either directly through phone interviews or focus groups, or just from the experience of particular marketing managers. These marketing products, campaigns or anything else represent the available arsenal with which to motivate the customer to do more business. If this arsenal of interventions already exists, then the remaining questions are relatively simple: "What do I do to whom, and when?" Assuming that there is some regularly scheduled launching or marketing programs (let's say monthly), the question can become: "What should I do to whom?" Not doing anything should be considered also to be a viable intervention/catalyst, as well as anything else that might be used.
If you want to do a good job of answering the question: What should I do to whom? It pays to create a model of what the likely outcome would be if the variety of different interventions were to be applied. This is the predict step (based on past experience). This is what the technology of data mining is used for.
For a complete marketing optimization system, whatever is predicted must also be measured. For instance, it doesn't help to correctly predict customer attrition if you don't know what the value of a saved customer is. Also, it is important that the step of action is with the infrastructure of running the complete marketing optimization system. For instance, many of the classical decision support or business intelligence systems will leave off on the action step. These systems will often provide useful information about how to improve the marketing process to the customer, but the results are delivered in the forms of graphs or reports that a senior manager can look at and better understand their business. In some way, the CEO can act based on the report of the graph, but typically it happens outside of the systems that should be recording the purpose of the action and the deployment of the action. The CEO may be influenced by what they've seen in the results of the decision support system, but their action is "non-recordable" and non-measurable. For marketing optimization to realize the full potential, this action step must be recorded within the same system that provides the measurement and produces the predicted future behavior.
To learn move about Building Data Mining Applications for CRM, click here.
What did you think of this tip? Love it? Hate it? Email and let us know.
This was first published in April 2001
Join the conversationComment
Share
Comments
Results
Contribute to the conversation