1.0–5.0 MOS Explained for IT Teams: Measure, Alert, Fix VoIP QualityA MOS score is a rating of voice call clarity on a scale from 1.0 to 5.0, with higher values indicating better quality; enterprises often aim for scores considered good for their operations. The three biggest drivers are packet loss, one-way latency, and jitter, and in practice you won't have a human rate every call. Monitoring platforms estimate MOS continuously from network metrics using the same math carriers rely on.

TL;DR:

Within enterprise networks, packet loss, bursty loss, latency, and jitter have the most significant impact on MOS, especially when bandwidth and codec limitations are considered.

Monitoring should focus on sustained MOS drops over minutes and correlated packet loss spikes, rather than momentary fluctuations or average loss percentages.

Fixing packet loss is the priority, followed by QoS policies, appropriate codec choices, adaptive jitter buffers, and minimizing transcoding to improve overall voice quality.

Using continuous synthetic probes, Real-Time Control Protocol data, and call detail records provides the most reliable way to track MOS and pinpoint issues across multiple locations.

Employing a managed service provider can simplify MOS maintenance, offering constant monitoring, traffic engineering, and rapid triage across diverse sites and carriers.

What Does MOS Score Mean for VoIP Quality?
How Is MOS Calculated From Network Metrics?
Subjective vs. Objective vs. Estimated MOS: Which One Should You Use?
What Network Factors Have the Biggest Impact on MOS?
How Do You Measure MOS Score in Practice?
How Do You Improve MOS Score on a VoIP Network?
What MOS Thresholds Should Trigger an Alert?
What Do Most Teams Get Wrong About MOS Monitoring?
Get Help Maintaining Voice Quality Across Every Location
Sources

What Does MOS Score Mean for VoIP Quality?

Mean Opinion Score started as a purely subjective measurement. Telecom engineers gathered panels of listeners, played them audio samples, and averaged their ratings from 1 (bad) to 5 (excellent). That method still exists, but it's slow and expensive, so almost nobody runs live listener panels for day-to-day network monitoring anymore. Modern VoIP tools estimate MOS from measurable network conditions instead, which is why you'll see it called "estimated MOS" in most dashboards.

The scale breaks into rough perceptual bands:

Around 4.3 to 5.0: Excellent, close to landline call quality
Around 4.0 to 4.3: Good, typical practical enterprise floor
Approximately 3.6 to 4.0: Fair, occasional user-noticeable distortion
Roughly 3.1 to 3.6: Poor, noticeable call quality issues
Below about 3.1: Bad, calls may be frustrating or unusable

Context matters more than the raw number suggests. A MOS of 3.8 measured on a wideband Opus call means something different than 3.8 on narrowband G.729, so comparing scores across different codecs or test conditions without noting that context is misleading.

How Is MOS Calculated From Network Metrics?

Most estimated MOS values trace back to the E-Model, the transmission-planning standard that converts network impairments into a single transmission rating called the R-factor. R-factor runs from 0 to 100, and that number maps directly to a MOS value.

The E-Model weighs several inputs:

One-way delay: Delay above roughly 150 milliseconds starts degrading conversational flow and lowers the R-factor.
Codec impairment (Ie): Every codec has a baseline distortion penalty baked into its compression method.
Packet loss and burstiness: Loss doesn't just subtract points evenly. Bursty loss punishes the score harder than the same average loss spread randomly, because packet loss concealment can only mask so much consecutive silence.
Jitter buffer effects: A buffer that's too small drops late packets; one that's too large adds delay that acts like extra latency in the R-factor math.

The practical takeaway: your codec sets a ceiling, and network conditions determine how far below that ceiling you actually land.

Subjective vs. Objective vs. Estimated MOS: Which One Should You Use?

ITU-T P.800.1 formalizes three distinct MOS types so engineers stop comparing apples to oranges. MOS-S is the original subjective listener-panel score. MOS-O is objective, produced by algorithms like POLQA or PESQ that analyze actual audio waveforms against a reference signal. MOS-E is estimated, calculated from network parameters through the E-Model without touching any audio at all.

Each has a place. Run subjective panels or POLQA analysis when you're validating a new codec, certifying a carrier route, or investigating a customer complaint that needs forensic-level proof. Rely on estimated MOS for continuous, real-time monitoring across hundreds of simultaneous calls, since it's cheap to compute and doesn't require capturing audio streams. Whichever method you use, document the audio bandwidth and test method in your reports. A number without that context isn't comparable to anyone else's number.

What Network Factors Have the Biggest Impact on MOS?

Packet loss does the most damage, and not all loss is equal. Random loss of just 1% typically costs 0.3 to 0.5 MOS points depending on codec, but the same percentage delivered in bursts hits far harder, since concealment algorithms run out of room to hide missing audio. Research comparing loss patterns confirms burst loss degrades perceived quality well beyond what the average loss rate would predict, which is exactly why chasing a low average packet-loss number can still leave users complaining about choppy calls.

Latency and jitter compound the problem. One-way delay past 150 milliseconds disrupts natural conversation rhythm, and jitter forces the playout buffer to either drop late packets or add delay to compensate. Codec ceilings matter too: G.711 and Opus support high fidelity, while compressed, low-bandwidth codecs cap out lower no matter how clean your network is. Add transcoding between codecs, and every conversion introduces its own quality tax. Congested Wi-Fi, consumer VPNs, and oversubscribed WAN links are the usual suspects behind all of this in real deployments.

Pro Tip: Before touching jitter buffer settings, check whether your packet loss is bursty or random. Tuning a buffer to fix what's actually a congestion problem just adds delay on top of the original issue.

What Network Factors Have the Biggest Impact on MOS? — overview diagram

How Do You Measure MOS Score in Practice?

Reliable MOS measurement combines several data sources rather than trusting one number from one tool.

Collect RTCP XR data from endpoints and SBCs, which reports loss, jitter, and round-trip time per call in real time.
Pull SBC call detail records for historical trending, since most session border controllers log codec, duration, and quality metrics for every call that crosses them.
Run active synthetic probes that place test calls over the same media path and codec as real user traffic, giving you a continuous estimated MOS baseline independent of actual call volume.
Reserve packet capture and POLQA analysis for forensic work, pulling PCAPs only when an incident needs root-cause proof beyond what RTCP summaries show.
Correlate everything on one dashboard so a MOS dip automatically shows you whether packet loss, jitter, or a codec change caused it.

A reasonable measurement cadence runs continuous path probes around the clock, samples real call quality daily, and triggers PCAP forensic analysis only when an alert fires. That keeps overhead low without leaving blind spots. If you're unclear on where SBCs sit in the call path, it helps to understand how VoIP phone systems route calls before deciding where to place your probes.

How Do You Improve MOS Score on a VoIP Network?

Fix packet loss first. Everything else is secondary until loss is under control, because no jitter buffer setting or codec choice compensates for a network that's actively dropping traffic. Apply QoS policies that prioritize voice with DSCP EF marking, reserve dedicated bandwidth for voice VLANs, and move calls off congested Wi-Fi onto wired connections wherever the deployment allows it.

Codec choice sets your ceiling, so use Opus or G.711 where bandwidth supports them rather than defaulting to low-bitrate codecs that cap quality regardless of network health. Tune jitter buffers to adapt within a typical range around tens of milliseconds, while avoiding excessive maximum delays near or above 120 milliseconds, as larger buffers add perceivable latency. Cut unnecessary transcoding hops. Every codec conversion in the media path degrades audio a little more, so a shorter, simpler path beats a clever one.

Eliminate packet loss before adjusting anything else.
Prioritize voice traffic with QoS and DSCP EF marking.
Choose Opus or G.711 when bandwidth allows.
Keep jitter buffers adaptive, capped near 120 ms.
Minimize transcoding and shorten the media path.

Pro Tip: If you're running SD-WAN, confirm your voice traffic policy actually forces calls onto the lowest-latency path during failover. A misconfigured policy can silently reroute voice onto a backup circuit with double the delay.

What MOS Thresholds Should Trigger an Alert?

Set your baseline service level objective (SLO) around a MOS score typically regarded as good for general enterprise voice, with higher targets for contact centers where call clarity critically affects performance. Below those floors, treat it as a service degradation worth investigating, not just a number to note.

Build alerting around sustained drops rather than momentary blips, since a single bad packet burst shouldn't page anyone at 2 a.m. Watch for path-level MOS drops that persist for multiple minutes, packet loss spikes that correlate with the drop, and bursty loss patterns specifically, since those slip past simple average-loss thresholds.

Alert on sustained MOS drops below your SLO, not single-sample dips.
Flag correlated packet loss spikes alongside MOS degradation.
Detect burst loss patterns separately from average loss percentage.
Triage in order: packet loss, then jitter and late packets, then codec or transcoding, then routing and media relay.

That triage order matters. Chasing codec settings before ruling out packet loss wastes time and usually fixes nothing.

What Do Most Teams Get Wrong About MOS Monitoring?

The most common mistake is tuning jitter buffers as a first response to complaints, when the real culprit is usually bursty packet loss that a bigger buffer can't fix. A close second: nobody checks whether the codec itself caps achievable quality before blaming the network for a ceiling that was never reachable in the first place.

A workable checklist stays simple: measure continuously with synthetic probes, isolate whether loss, jitter, or codec is driving the drop, fix the specific cause, then validate with a follow-up measurement instead of assuming the fix worked. Teams running multi-location voice deployments often find this easier to sustain with a managed provider handling the monitoring layer, since consistent MOS tracking across dozens of sites takes dedicated attention most internal IT teams can't spare full time.

— Jim

Get Help Maintaining Voice Quality Across Every Location

Chasing MOS drops across a dozen offices with a dozen different carriers is where most internal IT teams lose the thread. A managed services provider removes that complexity by sourcing connectivity from multiple carriers and engineering each site's voice path directly, backed by a 24/7 U.S.-based NOC watching for the packet loss, jitter, and latency spikes that drag MOS below your target.Californiatelecom's managed UCaaS platform runs on a 99.999% uptime SLA for voice, so the alerting and triage steps covered above happen on the provider side before users ever notice a drop. For sites still fighting quality issues on underlying WAN links, managed LAN/WAN services address the routing and bandwidth contention that usually sit behind persistent MOS problems. If your business runs contact-center-grade voice and can't afford to guess at root cause, teams handling high call volumes may also want to look at how voice agent deployments are tuning quality thresholds on their end. Request a free network assessment from a managed network services provider to get a real read on where your current voice quality stands and what's holding it back.

1.0–5.0 MOS Explained for IT Teams: Measure, Alert, Fix VoIP Quality

Table of Contents