Imagine that you are a large retailer interested in understanding the customer base. There may be several “types” of customers, such as those shopping for business with corporate accounts, those shopping for leisure, or debt-strapped grad students. Each of these customers would exhibit different behavior, and should be treated differently statistically. But how can a customer’s “type” be defined? Especially for large customer data sets in the millions, one can imagine how this problem can be challenging.

Full Article