Machine Learning for Customer Segmentation





Customer segmentation is the process of dividing customers into segments up based on common characteristics like demographics and behavior. It is particularly useful for marketing: instead of treating all your customers the same way, you are able to put them in groups and treat each group differently, optimizing for different customer profiles.

As a technique, customer segmentation has been around for quite a while now, but the most traditional approaches were based in simple decision rules. With machine learning entering the scene, it was possible to use different approaches to take into account new ways in which customers can be similar or different.

SML stands for “Small / Medium / Large”, and is the simplest form possible: you split your customers according to how much money they spend on your business. “Small” customers are the ones who spend the least, and so on. You could either set thresholds based on the amount spent — for instance stating that “Large” customers are the ones who spend more than $1k a year — or based on quantiles — for instance saying that “Large” customers are always the top 10% of customers in terms of spending. Either way, you are taking into account only one dimension in your segment: their revenue.