Why I switched from REST to GraphQL Federation (and when I didn't)

I have a complicated relationship with GraphQL. I've watched it accelerate teams to delivery and I've watched it become a black hole of N+1 queries. The deciding factor — every single time — was whether the org was willing to invest in the schema as a first-class artifact.

Where federation actually shines

• Multiple teams owning slices of a domain that compose into one product surface.
• Frontends that stitch heterogeneous data with low patience for waterfalls.
• Schema-driven codegen across web, mobile, and BFFs.

Where I still pick REST

Internal service-to-service calls with simple shapes. File uploads. Streaming endpoints. Anywhere the operational surface area of a federated graph is more cost than the schema is value.

graphql
# subgraph A — owns the user
type User @key(fields: "id") {
  id: ID!
  name: String!
}

# subgraph B — extends with billing
extend type User @key(fields: "id") {
  id: ID! @external
  plan: Plan!
}