MOS (Mean Opinion Score) is the standard measure of VoIP call quality on a scale of 1 to 5. A MOS of 4.0 is toll quality, 3.5 is acceptable, below 3.0 users notice significant degradation. Here's what drives MOS, how to measure it, and how to improve it.
Mean Opinion Score (MOS) is a measure of perceived voice quality defined in ITU-T P.800. It was originally based on human listener panels rating call quality from 1 (bad) to 5 (excellent). Modern VoIP systems use algorithmic MOS estimation — E-Model (ITU-T G.107) and PESQ/POLQA — to calculate MOS from measurable network parameters without requiring human listeners.
MOS is the most widely used single-number metric for VoIP call quality. Most SBC and PBX systems report per-call MOS in CDRs, and it's the primary metric used in SLAs for managed VoIP services.
| MOS Score | Quality | User Experience |
|---|---|---|
| 4.3 – 5.0 | Excellent | Wideband HD audio, better than traditional PSTN |
| 4.0 – 4.3 | Good | Toll quality — equivalent to traditional PSTN |
| 3.5 – 4.0 | Acceptable | Slight degradation, acceptable for business use |
| 3.0 – 3.5 | Fair | Noticeable quality issues, some user complaints |
| 2.5 – 3.0 | Poor | Significant degradation, many complaints |
| < 2.5 | Bad | Unacceptable, calls frequently unusable |
G.711 (PCMU/PCMA) achieves MOS ~4.1 on a clean network. G.729 achieves ~3.92. Opus at full quality achieves MOS ~4.5. Codec choice sets the ceiling — network conditions determine how far below that ceiling you actually operate.
The E-Model (ITU-T G.107) calculates an R-factor from network impairment values which is then mapped to MOS. The primary inputs are:
The formula: R = 93.2 - Id - Ie-eff, then MOS = 1 + 0.035R + R(R-60)(100-R) × 7×10⁻⁶
In practice: 1% packet loss costs roughly 0.3-0.5 MOS points depending on codec. 100ms one-way latency costs about 0.1 MOS points. Codec choice costs nothing on a clean network but G.729 starts at MOS 3.92 rather than 4.1.
RTCP Extended Reports include VoIP Metrics blocks with MOS-LQ (listening quality) and MOS-CQ (conversational quality). Check your PCAP for RTCP XR packets — not all endpoints send them but SBCs and enterprise IP phones often do.
Most modern SBCs (AudioCodes, Ribbon, Oracle) calculate and log per-call MOS. Check your CDR schema for mos_lq or r_factor fields. Asterisk with the res_rtp_asterisk module logs RTCP stats including estimated MOS.
Even without RTCP XR, you can estimate MOS from RTCP Receiver Report values: extract packet loss fraction, jitter, and delay, then apply the E-Model formula. SIPSymposium does this automatically from uploaded PCAPs.
Eliminate packet loss first — it has the highest impact. Implement QoS DSCP EF marking, separate voice VLANs, and ensure adequate WAN bandwidth. Target under 0.5% packet loss.
Switch codecs — if using G.729 and quality matters, switch to G.711 (PCMU/PCMA) or Opus. G.711 gives you 0.2 MOS points back immediately. Opus gives you 0.4+ and adds FEC for loss resilience.
Tune the jitter buffer — use adaptive jitter buffer with 40ms target. Don't exceed 120ms maximum or you sacrifice too much latency budget.
Minimize hops and transcoding — each codec transcoding point degrades quality. Maintain G.711 end-to-end where possible. Keep the call path direct — minimize the number of B2BUAs and media proxies between endpoints.
A good MOS score for VoIP is 4.0 or above, which is considered toll quality equivalent to traditional PSTN. A MOS of 3.5-4.0 is acceptable for business use. Below 3.0 users experience significant quality degradation and call complaints increase substantially.
Low VoIP MOS scores are primarily caused by packet loss (highest impact — 1% loss costs ~0.4 MOS points), codec selection (G.729 ceiling is 3.92 vs G.711 at 4.1), excessive end-to-end latency above 150ms, and jitter buffer overflow causing late packet discard.
VoIP MOS is calculated using the E-Model (ITU-T G.107). It converts network impairment values — codec degradation, packet loss, and delay — into an R-factor, then maps R to MOS using: MOS = 1 + 0.035R + R(R-60)(100-R) × 7×10-6. Most SBCs and PBX systems calculate and log MOS per call automatically.
Upload your PCAP to SIPSymposium. The analyzer extracts RTCP statistics, calculates estimated MOS using the E-Model, and identifies which impairment — packet loss, jitter, or codec — is driving your score down.