Mean Opinion Score

MOS (Mean Opinion Score) is a numerical measurement of voice call quality, rated on a scale from 1 (unintelligible) to 5 (excellent, indistinguishable from in-person speech). Originally derived from human listeners rating call quality in laboratory settings, MOS is now calculated algorithmically by analyzing network metrics like latency, jitter, packet loss, and codec characteristics. A MOS score of 4.0+ indicates toll-quality audio, 3.5-4.0 is acceptable for business calls, and anything below 3.5 represents noticeably degraded quality.

How It Works in VICIdial

MOS scores in VICIdial environments are typically calculated using the E-model (ITU-T G.107), which takes network impairment factors as inputs and produces an R-factor that maps to a MOS score. The key inputs are: latency (one-way delay, ideally under 150ms), jitter (packet timing variation, ideally under 30ms), packet loss (percentage of RTP packets that never arrive, ideally under 1%), and codec (each codec has an inherent quality ceiling — G.711 maxes at MOS 4.4, G.729 maxes at MOS 3.9).

Asterisk tracks RTP statistics for each call channel, including received packets, lost packets, and jitter measurements. These metrics can be extracted from Asterisk’s call detail records (CDRs) and channel variables to calculate per-call MOS scores. Third-party monitoring tools can also passively capture RTP traffic on the VICIdial server and compute real-time MOS scores across all active calls.

The maximum achievable MOS depends on the codec. G.711 (ulaw/alaw) provides the highest quality baseline with a theoretical MOS of 4.4 on a perfect network. G.729’s compression introduces some quality loss, capping at approximately 3.9 MOS even under ideal conditions. Network impairments reduce the score from these baselines.

Why It Matters

MOS score is the single most important metric for evaluating voice quality in a VICIdial call center. Call quality directly affects every operational metric: agents on low-MOS calls spend more time repeating information (increasing average handle time), customers on poor-quality calls are less likely to convert (reducing conversion rates), and persistent quality issues increase agent frustration and turnover.

A call center should target a minimum MOS of 3.5 across all calls, with 4.0+ as the goal. Monitoring MOS scores over time helps identify degradation trends — a carrier route that slowly worsens, a network switch that drops packets during peak hours, or a remote agent location with inadequate connectivity. Correlating MOS scores with carrier routes, time of day, and agent locations enables targeted troubleshooting to maintain voice quality standards.