Streamlining Data Collection and Entity Management for Amazon's Buyer Abuse Prevention Team

Session description

The Buyer Abuse Prevention team is responsible for preventing returns abuse on Amazon.com. We do this by leveraging multiple streams of data to help make accurate decisions that minimize friction to our good customers. Enrichment of entities, for example, orders, is often distributed across multiple APIs, which makes collecting and organizing large sets of data inefficient and inflexible. To simplify development, we built a GraphQL API to consolidate the collection and storage of data that allowed us to break the dependency on API results and design our storage around entities in a way that was optimal for our business. Chaining API calls now only takes place within a single API without need for code replication. Swapping out the underlying API for specific fields no longer requires code refactoring as the shape of the entity remained the same. The schema is well-connected which allows for different entry points but ultimately arrives at the same data without needing to reinvent the wheel. We can now focus on developing a schema and set of entities that match our business needs, without risk of major refactoring when a dependent API changes.