Getting Maximum Value Out of Analytics
As businesses add better analytics, they increase the number of decision points for differentiating between customers and making more targeted decisions. But just as having lots of data can be overwhelming and of little value in and of itself, so it is with the analytic predictions drawn from this data. The business value of the predictions depends on how they are operationalized.
All of the relevant analytic insights need to be brought together into day-to-day decisions. To do that, businesses need a powerful business rules management system (BRMS) and optimization engine. Best-in-class systems incorporating these technologies can take the analytic output of thousands of models and deploy them in decisions that are operationalized across millions of customers.
A BRMS is a fundamental capability since customer behavioral predictions are usually linked with actions through business rules. For example: “If a customer has a high propensity for purchasing new kitchen cabinets in the next 90 days, and is 20 percent more likely to act within the next 30 days if they receive Coupon A, then include them in this email distribution.”
Related rules like these make up a decision strategy that will generally be tested on a small population segment, and the results analyzed. Businesses can make this process faster and more efficient by using technology solutions that allow models to be deployed directly (i.e., without any kind of recoding) into the BRMS powering operational decisions.
Businesses can further compress test-and-learn cycles—accelerating performance gains—by using experimental design. Also called “multivariate testing,” this is a methodology with which large numbers of decision strategies are tested simultaneously on smaller population subsets. Because this approach enables testers to infer what the results would have been on untested populations, it yields more learning from fewer tests.
Every decision, even one based on a single “If ..., then ...” business rule, could be described as a model, since it is a representation of how a decision is being made. But when large numbers of analytic predictions are used to differentiate and treat customers,individually, the number of rules can explode. The decision process can become difficult to manage and even to fully comprehend. A decision model simplifies such complex decisions by mathematically mapping the relationships between all the factors and outputting an actionable result, such as a recommended customer treatment.
Moreover, explicit modeling of customer reactions to a range of business actions (often called “action-effect modeling”) clarifies complex decisions and exposes key performance drivers. Consider our previous example of a campaign aimed at accelerating purchases by customers with a high propensity to buy kitchen cabinets. An action-effect decision model could determine on a customer-by-customer basis what the net impact would likely be on revenue, costs and profit.
Such decision modeling would almost certainly be used with an optimization engine to identify the best treatment for each customer given the real-world constraints (mailing volumes, store locations, program spend limits, etc.) of the business and its suppliers. In this way, businesses can execute portfolio-level business strategies with precision at the level of individual customers. They can find the “sweet spot” where what the customer wants and what the business and its partners want intersect. It’s the realization of the very definition of successful marketing—”Find a need and fill it.”
Success, of course, must be measurable, and thus businesses need systematic testing practices and rigorous measurement. They must able to determine how much of a result is due to analytics as opposed to other decision elements and operational factors.
Additional information on this topic is available in our Insights white paper: "Which Retail Analytics Do You Need?" (registration required).