Decoding the customer journey with graph node embeddings

How to use graph networks and algorithms to derive insights from the customer journey

Clustering algorithm applied to the final customer journey embeddings — but as art. By the author.

The customer journey is complex. This may be the thousandth time you have read such a line and yet a) it is still true and b) our understanding of how to decode customer journeys is still limited. Despite countless efforts to do so, very few have been able to successfully derive insights or predictions out of a sequence of behaviors. For those who do so with tabular datasets, the number of features required to capture all the combinations of journeys (i.e., the sequence, volume, type, and context of behaviors) is massive.

Graphs, or networks, offer a new approach to modeling the customer journey and translating this data into insights. Likewise, they have the flexibility — and proficiency — to enable this to occur for different cross-sections and granularities. The suite of graph algorithms is an important component to doing this at scale (i.e., for millions of customers with millions of different micro journeys).

In this article, I explain how I’ve applied graphs and graph algorithms to achieve two goals:

• Find customers who would benefit the most from a new Microsoft product.
• Calculate the similarity among…