Every market-data vendor quotes a latency number. Almost none tell you what it includes, where it's measured, or what the tail looks like. We think that's backwards: latency is a product surface, and it deserves the same honesty as an API contract.
The budget, itemized
Our end-to-end budget is 50 milliseconds from source publication to your socket: ingest, parse, score, publish. Each stage has its own budget and its own telemetry, and every event carries its actual timings — so you can audit us on your own traffic, not our marketing page.
Averages hide the events that matter
The headlines that move markets are precisely the ones that arrive in bursts, and bursts are where averages lie. P50 on a quiet Tuesday is trivia; P99 during a Fed statement is the product. That's the number we optimize, and the one we publish.