The Latency Budget: Where Milliseconds Go From Tick to Screen

The 100ms budget, itemized for a real-time quote path:

• Exchange → ingress: ~5-15ms (co-located feeds are the floor)
• Normalize/parse: ~1-2ms (binary fix engines are fast; JSON is slow)
• Fan-out/pubsub: ~5-10ms (Valkey/Redis cluster hop)
• Websocket encode+send: ~2-5ms per connection
• Network to device: ~30-60ms (mobile, variable)
• Render/paint: ~8-16ms (the frame budget)

Total: ~50-110ms on a good path. The render is the honest part — everything before it can be tuned.

Stock Picks's latency dashboard publishes the live budget — the philosophy is "show the milliseconds, don't promise microseconds."