Machine learning projects come in a variety of flavors – in this case, we used an unsupervised learning method call k-means clustering to identify customer segments based on several data elements. The nice thing about these types of machine learning algorithms is that they learn without much guidance. For this project requested by a food ordering app, I developed customer groupings into four segments that reflect four very different customer behaviors that marketing and product managers could use to iterate their work. For example, the four clusters had very clear separation on recency and frequency of their orders over the app:
Combined with typical market research methods, these customer clusters provided guidance as to when marketers should send out messages and as to what level of user engagement product managers could expect.
-Ryan Morton
