API capacity planning: start with RPM, TPM and peaks
A practical checklist for a more useful enterprise API capacity conversation.
Two measurements, different questions
RPM describes requests per minute; TPM describes tokens per minute. Provider limits may distinguish different token types and scopes, so verify the rules for your actual project.
Write down the workload
Share your use case, average and peak request rate, typical input and output size, concurrency and target deployment region. Monthly usage alone does not describe a short burst of traffic.
Define the commercial boundary
Ask what capacity can be committed, how usage will be metered and what happens if demand exceeds the agreement. Confirm the service source and authorization before discussing supply.
Bring a realistic sample
A representative workload makes the planning discussion more concrete. Use synthetic or redacted examples, never API credentials or confidential customer records.
Further reading: upstream rate-limit documentation. Always verify current limits against your own project.