Next Best Action
Over the last few months I have been introducing some of the basic topics of decision management — a focus on decisions, the use of data mining and predictive analytics, and adaptive control. I have not spent much time on business rules as plenty of other authors do that around here. This month I thought I would discuss a growth area in the use of decision management — next best action.
For those of you who might not have heard the phrase, 'next best action' (or best next action) is an approach that brings together marketing and customer service by taking the most appropriate action each time there is a customer interaction. This might be a cross-sell, a discount on existing services, no action at all, or any number of other actions. The idea is to determine the action for each specific customer that will most benefit the company, typically in the long term, while building a customer relationship that the customer values. Next best action also involves identifying the right time for the action — now, next time we send them something, after their next paycheck — and the right channel for it.
To make next best action work, companies need to adopt the core principles of decision management. First they need to look at their processes and the interactions they have with customers, and determine when and how they make decisions about those interactions. Managing and improving these decisions is what next best action is all about. Delivering consistency across channels and ensuring that the decisions being made are current and timely means putting these decisions into coherent decision services that can be used by multiple systems across multiple channels.
Business rules play a key role in next best action. Not only must the company define the rules for eligibility of customers for specific offers, it must also define the rules that constrain interactions, such as those limiting what might be said to someone in collections, for example. Regulations and other restrictions on the allowed actions are likewise coded in business rules and the preferences of individual customers can (and should) also be represented as business rules. For instance, a customer might express a preference for using a particular channel and that rule should be used when offers are being made or other actions considered.
While business rules form the foundation for a next best action approach, analytics are critical for determining which actions are best. Data mining is typically used to determine the most effective way to segment customers into different groups, and many companies are pushing to more and more fine-grained segmentation. Representing these segmentation models as business rules — often a decision tree — allows them to be easily integrated into the core rules-based decision. Predictive analytics that turn uncertainty about the behavior or value or response of customers into probabilities are also widely used. Using predictive analytic techniques to develop models that will score the likelihood that a particular customer is a retention risk, or that a particular offer will be accepted, helps put the data companies have about the past behavior of their customers to work. Rules can then be written that take advantage of these predictions and they can even be used in segmentation models — using the retention or churn risk, for instance, as one of the criteria for membership in some customer segments.
Adaptive control — the final element of our triad of techniques — is also widely used. While some offers and actions have been used enough to build up a pretty good sense of what will or won’t work, others must be tried to see what response they provoke. Using adaptive control to ensure that new approaches are constantly tried against a minority of the population helps build more robust models of customer behavior while limiting the risk of any new approach to a small group. Instead of marketing, customer support, and sales arguing about what might work, adaptive control allows different approaches to be tried to see what does, in fact, work.Next best action is an interesting area and one for which decision management is exactly what you need. Don’t forget that there is a lot more information on this in Smart (Enough) Systems, and you can use RSS or e-mail to subscribe to the Smart (enough) Systems blog.
# # #