The Problem with Fancy Segmentation
Segmenting customers into subsets that have unique needs, interests, and priorities makes a lot of sense. The more obvious the segments are, the more it makes sense, so a priori segments are best. Big companies in one segment, small companies in another. Retirees in one segment, college students in another. You don’t need fancy K-means or cluster analyses for the most useful kinds of segmentation.
What happens when we start segmenting customers in “deeper” ways, based on attitudes, motivations, preferences, behaviors, or what have you? Fancy algorithms will identify distinct clusters of customers. Intelligent research analysts will identify the differentiating attributes. Clever alliterative strategists will name the segments for you to highlight their uniqueness. But too often the differences are much ado about nothing.
Consider the segmentation data Versta is now reviewing for a fitness-club. Members were segmented on 75 variables about why they work out and how working out makes them feel. The analysis identified five stable segments, with two segments differentiated by their “negative” leaning in terms of motives and feelings (they work out because a doctor told them to, or to lose weight, even though they hate it).
But looking more carefully at the full scope of data, it becomes clear that these segments are not all that negative! Just 12% to 13% in the two segments said they hate working out, vs. 1% to 3% in the other segments. Yes, negativity is differentiating. No, it is not defining. On the dimensions that really matter all six segments are remarkably similar.
We use fancy segmentation all the time, and for sure the techniques are useful. But in the end, our advice is always this: Review the dominant characteristics of each segment as thoroughly as you review the differentiating ones, because too often the differences just don’t matter.
