Graph Data Science Streamlines Complex Medical Supply Chain Analysis
Boston Scientific is highly vertically integrated in the development, design, manufacturing and sales of its medical products. This means that the company manufactures its complex devices, starting with raw materials such as resin and metal. Starting with raw materials requires a significant amount of batch processing in addition to discrete manufacturing of finished products.
With its complex manufacturing supply chains, Boston Scientific has numerous engineering teams aligned based on the process technology they’re using.
Many devices use multiple technologies, so multiple teams are involved. With multiple teams, often in different countries, working on the same problems in parallel, engineers resort to analyzing their data in spreadsheets.
Decentralized ad-hoc analytics led to inconsistencies and, most importantly, an inability to find root causes of defects.
Boston Scientific needed a far more effective method for analyzing, coordinating and improving their manufacturing processes across all its locations.
Eric Wespi, Data Scientist at Boston Scientific, and his team understood the complexity of their business problem and their inability to address it using traditional tools. Wespi attended GraphConnect, where he met Eric Spiegelberg, a Senior Consultant at GraphAware, and decided to work with him on Boston Scientific’s supply chain problems.
Boston Scientific moved from considering graphs to vetting the technology. “This is where the quantum leap happens, where an organization goes from merely being a passive observer of the graph community to being an active participant,” said GraphAware’s Eric Spiegelberg. “It is typically embodied by an organization getting their actual data into Neo4j, which is a critical step because it forces the organization to evaluate the design decisions and build out their model.”
The heart of Boston Scientific’s graph data model consists of three nodes representing a finished product, a part and a failure, with relationships that trace failures to parts and connect those to finished products. This simple model effectively represents a complex medical device composed of dozens of parts, manufactured from raw materials.
The data model is simple, but Boston Scientific’s graph is massive. Using Neo4j graph algorithms from the Graph Data Science Library, the Boston Scientific team analyzes its graph and computes scores that rank nodes based on their proximity to failures, enriching its models with insights derived from the graph.
“We’re using Neo4j not as just a data store, but as a place to analyze data and store those new characteristics of the data back in the graph and then extract it for traditional analysis,” said Wespi.