Cox Associates: Superior Business Decisions
through Better Data Analysis
|
|
|
We have developed a powerful approach to forecasting how different customers will respond to different product offers, bundling strategies, campaigns, and competitor offers. The key idea is to use the data-mining technique of classification tree analysis to identify relatively homogeneous groups of customers who have similar probablities for making changes ("transitions") in their purchasing behaviors, given the mix of variables offered to them. Once the transition rates have been quantified as functions of these stimuli, they can be used to forecast product-specific penetration rates over time, customer- and product-specific attrition rates, and optimal (lifetime value-maximizing) product offers. Our product attrition, forecasting, and "next optimal product" models have broken new ground in showing how to use corporate data to improve targeting of products to customers. Two unique technologies have resulted from our applied work in this area: 1. We have developed and validated a method for using a few months of cross-sectional data on individual customers to prepare product sales forecasts that compare favorably to those previously obtained using over 5 years of time series data. 2. Unlike almost all other data-mining companies, we have perfected methods for causal prediction of what customers are likely to do next in response to your decisions. This improves on methods that only match customers who have not yet bought a product to statistically "similar"-appearing customers who have already bought. The idea that products should be targeted to customers who look like past profitable customers is good. Matching products to customers based on who is most likely to buy what next is an even better idea. This is our unique contribution. For more detailed information, click here or read our paper on Classification Tree based Demand Forecasting.
|
