Yeah for me the understanding really came when working in a federated GraphQL API. Each team had us own little slice of overall object graph, and overlap / duplication / confusing objects across the whole domain were a lot easier to see in that environment.