Why does Figma's prototype preview latency increase when multiple users interact concurrently?

Answering the question

When several collaborators trigger the same Interactive Component in a Figma prototype, the preview often feels noticeably slower than with a single user. The delay is caused by the higher volume of real‑time collaboration traffic that Figma must process, rather than by changes in static assets.

Confirmed facts

• Figma relies on a persistent WebSocket connection for live collaboration; each interaction is sent as a JSON message to the server and broadcast to all connected clients.
• Prototype rendering is performed locally in the browser, but the client still has to consume the server‑broadcast updates to keep the view in sync.
• Static assets (images, fonts, etc.) are served from a CDN, so increased latency is almost always linked to dynamic event traffic.

Likely explanations for the latency spike

• WebSocket event queueing. More users mean more messages per second, which can saturate the server’s event loop or the client’s message queue, adding round‑trip delay.
• Rendering pipeline contention. Each incoming event forces the browser to re‑render the affected component. Multiple events queued at once can overwhelm the GPU or CPU, causing frame drops.
• Network bandwidth contention. Users on the same ISP or shared Wi‑Fi may compete for upload bandwidth, especially when many large interaction payloads are sent.

How to isolate the source of the delay

1. Measure WebSocket traffic. Open the browser’s Network panel, filter for ws:// frames, and note the number of messages per second while adding or removing collaborators.
2. Monitor client CPU/GPU usage. Use the browser’s Performance tab or system monitor to see if the rendering thread peaks when multiple users are active.
3. Test with a controlled number of users. Create a test file, invite a small group (2–3 users), then add more users incrementally while recording latency via the prototype’s built‑in performance overlay.
4. Throttle network bandwidth. Use a tool like tc (Linux) or a browser extension to limit upload speed and observe if latency improves, indicating a network bottleneck.

What to ask next

If the steps above don’t pinpoint the issue, the most useful missing detail is the typical number of concurrent users and the interaction pattern (e.g., rapid taps vs. sustained holds). Knowing whether the spike occurs at 5 users versus 20, and whether the interactions are high‑frequency or low‑frequency, will help decide whether to focus on server‑side scaling or client‑side optimizations.