3. Consumer Segmentation
Segmentation is one of the most fundamental skills in marketing analytics. It brings tremendous efficiencies to any marketing campaign, and to the ways in which marketers approach their consumers.
Suppose we are working with a database that contains information about more than 50,000 different customers. Perhaps an undifferentiated, mass-marketing approach would not be optimal – if there are specific groups of customers who prefer our services or products for particular reasons, a one-size-fits-all method may be inappropriate. At the same time, it would be unreasonable and unrealistic of us to expect our marketing team to concoct a unique strategy for each of the customers in the database.
What if, instead, we could come up with something far better than either of the two extremes mentioned above? What if we could place each of those 50,000 customers into one of just eight groups, based on things like demographic data, purchase history, interests, preferences, etc.? Doing so would still involve some tradeoffs – after all, none of these groups will consist of purely identical consumer personas. However, this process does not need to be perfect in order to be effective. Rather than strive for perfection, we can settle for “good enough” and we can use a process known as clustering to get there.
Clustering, a form of unsupervised machine learning, can be performed in many contexts. While this chapter will focus mainly on consumer segmentation, we could cluster any type of observations, as long as we have enough variables to meaningfully make that separation. Lobster Land could cluster types of visitor attractions, vacation packages, competitor parks, or just about anything else under the sun.
The goal of any clustering model is to place records into distinct groups. Ideally, the groups should be built in a way that maximizes intra-group similarity, along with inter-group differences, given the constraint of forming just enough groups to meet the business purpose behind the model.
Throughout this chapter, we will use the terms “clustering” and “segmentation” interchangeably.