Brands have traditionally understood their customers using survey data and focus groups. In an era where those customers leave behind a digital record of their interests and motivations throughout the day, machine learning offers a route to a richer understanding of those people and their relationships with the brand

Chief Data Scientist Melinda Han Williams recently joined the UX + Data Meetup to share Dstillery’s approach to customer segmentation using digital observed behavioral data, in particular web browsing behavior across millions of sites. A key challenge to working with such high dimensional data is finding a representation and visualization that allows us to effectively make use of and intuitively understand the data. She explores how Dstillery trains a neural network on billions of behavioral interactions to create a semantic embedding of web browsing behavior. The result is a map of the internet, which they use to drive unsupervised machine learning for customer segmentation. Watch the session below.

Video by Bill Prickett