A Mean Opinion Score, often shortened to MOS, is a way of quantifying the qualitative experience of a phone call. Since the first telephone call was placed by Alexander Graham Bell, in 1876, it’s been a fundamental challenge of scientists and engineers to ensure that the sound coming out of the receiver is as close as possible, in terms of quality and timing, to the sound going in at the other end.
The limiting factor is often physics and it’s a situation that has only become worse as the size and complexity of the PSTN, the world’s aggregated copper-wire telephone network, has grown.
Anyone who’s made an international call will know that the more hops between networks a call needs to make, the more the perceptible quality will deteriorate with the introduction of lost packets leading to jitter and longer distances causing greater latency.
The key word, bolded above, is ‘perceptible’. Because Mean Opinion Scoring is a way of turning what is ultimately the subjective experience of listening to voice signals received by way of a phone call into something objective: a number rating from 1-5, with 1 being bad and 5 being excellent.
What MOS Should You Be Aiming For?
On such a scale, an absolute value of 5 is unlikely as there is always realistically room for improvement.
Some degree of quality degradation is unavoidable because of the latency introduced by the distance that a voice signal has to travel between start and end points, as well as the packet loss and jitter that come into play as the amount of network infrastructure through which the signal must traverse increases.
In real terms, a MOS of 4.4 is excellent and anything below 3.5 will likely start harming the overall call experience for both parties.
How is MOS Calculated?
To understand the value that MOS provides, it is good to know how it is calculated. Today, you can objectively measure voice network quality with metrics such as latency and packet loss to incredibly high degrees of accuracy.
These ‘scientific’ measurements might appear to make MOS, which was originally calculated by way of surveys, seem antiquated in comparison.
However, MOS serves the important function of putting a number on an actual human experience in a way that knowing the number of packets dropped or the amount of jitter on a call just doesn’t really capture. It provides a holistic way of quickly understanding the ‘overall quality’ of the call.
This is one reason why MOS is primarily used to judge things involving compression and transmission, where the end result can often be a pale imitation of the original. A copy of a copy, etc. How good is the experience of listening to that compressed music file or streaming that 4k video? Because it is in the compression, encryption, transmission and decryption of signals that they begin to deteriorate.
Of course, MOS today is much more scientific than simply asking the opinion of someone listening to a phone call in a quiet room, as was the case for many decades.
In telecoms today, it can be calculated using algorithms designed to estimate the experience as a human would perceive it. And to standardize this process, the ITU has created recommendations for these so-called ‘objective quality models.
How Important is MOS?
Many argue that there are still so many vagaries in the formulation of a mean opinion score, or that it cannot be used to compare two different experiments. Others claim that boiling the experience down to a score of 1-5 is too simplistic given what is being measured.
But the strength of MOS is in measuring the same thing over time, and the number is ultimately a shorthand designed to give you a clear idea of where your network might need improving or where you might need to find a better provider. If the MOS for a call placed between France and Moscow is 4.4 one day and then 3.6 a week later, you know something is probably up. Remember the numerical range may be 1-5, but in reality, you’re looking at the difference between 4.4 and 3.5 being the difference between truly exceptional and problematically bad. And that’s just 0.9 on the scale.
Ultimately, MOS is one of several tools at the disposal of modern voice and network engineers in assessing the quality of the communications they are providing to their end users.
‘A Satisfied Customer is the Best Business Strategy of All’
That quote, by business author Michael LeBoeuf, sums up why MOS matters. Because quality of service, much more than price, is what keeps customers sticky. And MOS is one way that you can choose to measure the quality of your service. Whether you’re selling voice services to customers or providing the means for contact center agents and sales reps to talk to customers and prospects, the experiences you enable are what you will be judged upon.
This is why CX has become so ubiquitous. Because in increasingly commoditized industries, it’s the customer experience that becomes the defining point of differentiation between competitors.
And if you’re spending big to ensure your teams have the training, tools and resources they need to provide good CX across all your touchpoints, there’s no point undermining it all with poor-quality telephony infrastructure during what remain some of the most important touchpoints of all – real-time phone calls.
MOS and other quality metrics help you to ensure that your network and that of your provider (not to mention the method of interconnectivity between them) are all working in harmony to provide the best possible environment for your marketing, sales and support functions to flourish.
What Causes a Low MOS?
We’ve mentioned the three major culprits already:
This is basically the lag between someone saying something and you hearing it on the other end. The lower the latency, the more your call will replicate the experience of two people talking as if they are in the same room.
As latency increases, you’re likely to be left with uncomfortable interruptions and pauses; a problem that gets worse the more people are on the call. Latency is a huge issue in telecoms and is predominantly caused by three factors:
- Distance – The further the distance between the points A and B, the longer it will take the signal to travel between them. That’s just science. Anything under 100ms should be fairly imperceptible, but as you reach latencies in excess 300ms, there will be noticeable lag on the call.
- Complexity – By which we mean how complicated the routing of the call is. If it needs to hop between multiple networks and travel through a point-of-presence on the other side of the world before coming back, that will all add to the latency. The combination of these two factors is why latency can become so high with international calls made on the PSTN.
- Encryption – Adding VPN and other forms of encryption to a calls signalling and media further increases latency because of the time needed for the encryption and decryption processes
In effect, jitter is the change in latency over time. Again, as the complexity of the routing increases and more variables come into play, it becomes ever more likely that the latency on a call will change from moment to moment. Jitter will be especially apparent if your call needs to traverse the public internet, where routing is so unpredictable. It usually sounds like interference on the line.
It can be better to have a higher latency call with no jitter than a call with a lower average latency but very high jitter as, to some degree, the brain is able to accommodate low levels of latency as long as the delay is predictable. Jitter removes that predictability.
When voice signals are digitized and transmitted, they are split into packets that can be routed asynchronously and even via different paths, before being reconstituted and the end-point of the call. But if some of these packets fail to reach the end point – usually as a result of network congestion or failed hardware – small pieces of the audio signal will be missing, resulting in audible distortion on the call.
How to Improve Your MOS?
Ultimately the way to improve your MOS is to rid your calls of as much latency, jitter and packet loss as possible.
At Voxbone, we are able to provide customers with the highest quality calls for the following reasons:
- We’re a national operator in most countries where we operate meaning we are interconnected directly with the national network in that country. As a result we don’t aggregate service, so there are no hidden operators in the picture, meaning that your call will be completed in the minimum possible number of hops.
- Where this is not possible, we work only with Tier-1 providers to provide the best possible call experience.
- We own and operate our own resilient, meshed global VoIP network and local infrastructure meaning we are able to control the routing of your call from point A to point B and ensure geo-redundant routes are available.
- We offer a range of dedicated, private connectivity options including software-defined WAN and direct physical cross-connects, so your calls never have to traverse the public internet to get from your network to the Voxbone network.
- In countries where we provide full PSTN replacement, we are able to route outbound calls entirely in-country as opposed to most cloud-based outbound services, which rely on international calls. The result is local outbound calling that is imperceptible from legacy offerings in terms of quality, dialtone and reach.
How does Voxbone Measure Voice Quality?
Usually, measuring the quality of your calls requires expensive third-party tools that only provide sample ‘snapshots’ rather than ongoing and extensive monitoring, or expensive equipment to be installed directly into your network infrastructure.
At Voxbone, we provide all the quality metrics you need out of the box via our new Insights platform, which is currently in beta and being rolled out in phases this year. With Insights, you can get metrics such as MOS, jitter and packet loss on every single call you pass through our network.
Taking things a step further, we are even able to split these metrics in two – providing a granular breakdown of the quality of calls as they pass through our network and as they transit through yours.
This lets you see at a glance if quality issues are the result of an error on our network or due to external factors such as misconfiguration on your own network or internet congestion.
As a result, you’ll be able to more rapidly troubleshoot quality issues and quickly understand where problems are arising, to know the appropriate actions to take to resolve them.