Knowledge graphs might seem daunting due to their esoteric use cases, the newness of the concept, and their adjacency to many technical concepts. It can be difficult to know where to start. Much of the literature on the topic is targeted toward data scientists and is often heavy on jargon. Non-technical people like myself still need to wade through all of this to access what are actually very intuitive and increasingly approachable tools. Adoption among the product teams I talk to is still limited, but that is certainly changing. Here I try to describe what a knowledge graph is and how it can be used for product communication.
The type of graphs we are discussing are based on Graph theory, which was developed by Leonhard Euler in the 1700s. He was trying to figure out a walking route around his home city of Königsberg for a visiting king. He wanted to determine if it was possible to walk through the city, crossing each of its seven bridges only once (it wasn’t). To come to this conclusion, Euler abstracted the problem into nodes (landmarks) and relationships (bridges).
A knowledge graph is a purposeful arrangement of data such that information is put in context and insight is readily available.
Barrasa, J., & Webber, J. (2023). Building knowledge graphs: A practitioner’s guide. O’Reilly Media.
Your product ecosystem is Königsberg, the nodes are its products, features, use cases, orders, documentation, individual customers, and more, and the edges are the relationships between them. Your graph lets you visualize and explore these relationships and the underlying data in incredibly useful ways. More importantly, the knowledge graph is very intuitive.
A knowledge graph and its underlying datasets can be queried using natural language commands. "Show me all the digital assets that describe Product X and tell me who created them." The query traverses the graph, plotting a route like Euler through Königsberg, and retrieves the relevant data. The data might be stored in the graph or in connected databases. Without a knowledge graph you may have to query each database individually and combine the reports in an excel spreadsheet.
