Inverse Conway Maneuver, with GraphQL
summary
Zillow builds a canonical customer and move schema on a federated GraphQL graph that keeps writable subgraphs owned by domains, uses identity/accounts as the canonical source of truth, and publishes asynchronous writes over Kafka. Teams use a provided Move SDK to emit standardized events (loanId, customerId, participantIds) and subgraphs index by entity ID so queries work from loan or customer identifiers. Result: clearer data for analytics and AI and unblocked the integrated super app without a company reorg via GraphQL federation.
AI-written summary · about
chapters (10)
- 0:00Introduction and speaker
- 1:05Conway’s law explained
- 3:34Data silos and integration challenges
- 8:46Why not reorganize the company
- 10:50Rebuilding the data structure
- 13:57Shelling point and domain modeling
- 15:02Governance in GraphQL rights
- 16:49Developer experience with SDK
- 18:36Results and lessons learned
- 20:23Trade-offs and next steps
related talks
06mEscape the GraphQL Monolith
34mRealtor.com's Technological & Cultural Transformation with GraphQL
35mGraphQL in the Enterprise
46mMoving to GraphQL
08mGraphQL: everything you want in one query
26mKeynote: Apollo's Journey with GraphQL: Transforming Enterprise APIs for the Future
38mGraphQL Federation in the Wild
31mDesign Principles of Federated GraphQL