Voice & Video · SLA sizing
Estimate MOS from loss, jitter, and latency — and decide the SLA thresholds you need from your carrier and within your LAN. Outputs IP SLA config you can paste into Cisco routers.
"Good voice quality" isn't subjective — it's measurable. MOS 4.2 = excellent, 3.5 = acceptable, 2.5 = users call help desk. This tool calculates expected MOS from network conditions and flips the problem around: what loss/jitter/latency do you need to hit your MOS target?
MOS is computed with a simplified E-model variant: R = R0 − Id − Ie_eff − A, where Id accounts for delay and Ie for loss. Codec constants set the baseline. A jitter buffer compensates for jitter up to a limit; beyond that, the buffer drops packets and effective loss rises.
Industry rules of thumb: one-way latency ≤ 150 ms for transparent, ≤ 300 ms acceptable with echo cancellation. Jitter ≤ 30 ms (G.711) or ≤ 50 ms with adaptive jitter buffers. Packet loss ≤ 1%. Anything worse and MOS drops fast.
Cisco IP SLA jitter probes (UDP jitter operation) measure all three metrics in one probe. Schedule every minute, keep 14 days of history, alert on MOS < 3.8 sustained for 5 minutes.
Within ±0.2, yes. It's a good planning tool. For real MOS, use passive monitors (Cisco Collaboration Analyzer, Vantage DX, Martello) that ingest real RTCP-XR reports.
MOS > 4.0 for users to stop noticing. Webex Control Hub shows MOS per call. Teams shows stream health in CQD.
Video is more loss-tolerant (up to ~2% with FEC) but latency-sensitive (< 200 ms one-way for interactive). Same IP SLA jitter probes work.
Loss 0.1%, jitter 8ms, latency 30ms. G.711 MOS ≈ 4.35. Excellent.
Loss 1.2%, jitter 25ms, latency 70ms. Opus MOS ≈ 3.8. Marginal — add FEC or move to MPLS underlay.
Loss 3%, jitter 40ms, latency 60ms. G.711 MOS ≈ 3.1. Users will complain. Don't run voice here without dedicated SSID and CAC.