While operators focus on multimedia and IP-based content as future revenue generators, it is important that they do not forget the quality of their core product. Alun Lewis explains
The last decade has been a hectic time for the telecommunications industry. A combination of new technologies and radical regulatory change – combined with the financial quagmire of recent years – has challenged the ability of operators to stay focused and on track. It's now becoming clear that these distractions have impacted heavily on the core application that our industry has grown rich on: traditional voice services.
Even the familiar acronym of POTS – Plain Old Telephone Services – illustrates the relatively low regard shown for voice services when compared with the much-hyped potential of multimedia content and IP-based applications. Despite this perception, the global market for voice – according to international research consultancy Ovum – is expected to continue to grow from $784 billion last year to around $1000 billion by 2007, so there's still considerable market share to fight for.
One of the key problems facing operators, irrespective of whether they're running fixed, mobile or IP networks, is making sure that the quality of the voice services that they offer is acceptable to the markets that they're targeting. While voice generally used to be seen as a 'one size fits all' service – sensible enough when everything was being carried pretty much by one operator over one network – the current cat's cradle of complexity involving interconnections between different carrier technologies and service providers creates major headaches for all involved.
A business, for example, may be happy to pay a lower charge for lower quality voice services for internal company use – but more than happy to pay a premium for external calls to customers. Similarly, the youth market for mobile services may be attracted away from SMS communication to speech interaction at slightly higher tariffs – and be happy for a lower grade of service than the majority of other mobile users.
In particular, the impact of VoIP and the steady growth of supporting IP-based access technologies such as DSL and WLAN are already forcing service providers of all sizes and types into a re-examination of the whole voice services market. Additionally, for operators with more traditional networks, it's also essential that they understand exactly where in the infrastructure to make the right investments that will generate ARPU or market differentiators through enhanced or managed voice quality.
But how exactly do you measure voice quality? Unlike the hard technical parameters used to measure the efficiency of data networks – packet loss, latency and so on – perceived voice quality also depends on the hardware and software of that most complex of mechanisms, the human brain and nervous system. The telecoms value chain is complex enough when it's just based on copper and silicon – add in the human mouth and ear and that complexity increases by an order of magnitude.
The good news is that part of the ITU has been working on this problem for a number of years, coordinating research to develop standardised methods that can be applied across multiple networks and different technologies. Their approach is based on a concept known as PESQ – Perceptual Evaluation of Speech Quality – which measures end-to-end voice quality based on a database of subjective listeners' experiences of call quality defined by Mean Opinion Scores (MOS):
Listening quality MOS scale
Score Quality of the speech
While it's obviously possible for an operator to carry out regular customer surveys of voice quality, clearly the optimum solution is to automate the whole process through the use of software algorithms, so that network performance can be monitored on a regular basis. This means that they're effectively 'listening in' to calls and then reporting back to other management systems.
At the end of 2003, the ITU announced one of the first of these recommended standards, based on a non-intrusive algorithm for use in PSTN and mobile networks, following intensive evaluation of different solutions from a number of different companies.
Iain Wood of BT spin-off Psytechnics, one of the winning solution providers and a specialist in voice quality metrics, takes up the story: "The mobile market in particular is going to have to start addressing the issue of voice quality very seriously. If they're to keep pushing ARPU it's vital that they have some systems in place to be able to monitor the customer's end experience of voice services. Was a call ended prematurely because of voice quality problems, or is a customer churning because of poor reception? The kinds of data that you get from the usual QoS systems doesn't take account of this human-centric issue, even though it's vital to understanding the overall customer experience.
"Psytechnics has recently been looking at the performance of a number of mobile networks both in the UK and abroad and has found some interesting results. One of the most significant was that there was no discernable difference in speech quality between the most expensive and the cheapest handsets - despite a price differential of hundreds of pounds. Given that for many adult users voice easily remains their primary mobile service, does this mean that operators and handset manufacturers are putting too much effort and investment into unused – and ultimately unprofitable – features and capabilities?
"Similar issues apply when you consider the attempts that some mobile operators around the world are making to compete directly with the fixed line market through homezone tariffing. With a typical UK PSTN line providing a MOS of 4.3, UK GSM network typically operate at between 2.9 and 4.1, and this issue of perceived voice quality will make it unlikely that high spending business users will migrate unless service levels are improved.
"Even if the customer's getting a full signal strength bar, there are a host of other factors that degrade speech quality and an understanding of how voice quality is performing is an invaluable tool in how to best engineer the value chain and all its underlying components.
"This technology also has an equally valuable role to play for the end customer, especially the large corporate thats unsure whether it's receiving service value for its money. While it will certainly have strict Service Level Agreements in place to monitor its data networks – along with stringent penalties if these aren't met – both businesses are effectively flying in the dark when it comes to measuring speech quality. Because of this, we're currently working on an application that can be downloaded onto the mobile device itself. This can then use SMS to report back to the operator – or another party – to provide MOS feedback, as well as displaying voice quality metrics to the actual user."
Important role to play
While voice quality metrics have an important role to play with mobile service providers, they have an absolutely vital one when it comes to supporting the wide-scale roll out of VoIP services in both public and enterprise networks. The price and performance benefits of moving to an all-IP environment might be well understood, but customers are often rightly sceptical of its suitability for carrying mission critical voice services. The whole industry is full of anecdotes about disastrous VoIP implementations that have failed to deliver and caused nightmares for customer and vendor alike.
For companies like Psytechnics, this presents another market opportunity, as Iain Wood explains.
"IP was designed originally for computer-to-computer interaction and unfortunately humans are a lot more critical than machines when it comes to carrying on conversations. In trying to replicate the behaviour of the PSTN in a packet environment, there are all sorts of subtle perceptual cues that have to be taken into account if we are to provide a truly satisfying speech experience.
"Unfortunately, the legacy management tools that came with the IP world are incapable of supporting voice services – as the industry found to its cost in the past. Fortunately, we now have a range of techniques – similar to those now being deployed in the circuit switched environment – that can be deployed to ensure voice quality remains appropriate for the context that it's used in.
"What's important to remember, however, is that delivering voice quality in an IP environment must be an iterative process. You might start by testing a LAN for its suitability even before a VoIP solution is deployed. That can be done by running simulated VoIP traffic over it and monitoring performance at strategic locations. Then, during commissioning, it's useful to carry out end-to-end testing to fine tune the configurations and get it ready for live use. After thats done, regular monitoring remains essential, particularly given the constantly changing topology of most IT networks and the natural variations in traffic flow that affect any real world environment. In some of these scenarios, it's appropriate to use testing tools; in others, constant monitoring is best.
"We've found two ways into this rapidly growing market. Firstly through supporting system vendors by allowing them to incorporate our technology into their management systems, and secondly through providing a consultancy service for service providers, system integrators and end customers."
With the ITU about to announce similar standardised approaches for VoIP as they have already done for mobile and PSTN services, approaches to voice quality finally look like becoming definable – rather than infinitely debateable. Maybe it's time to update the old saying of the billing industry that 'if you can't bill for it, don't offer it,' to 'if you can't measure it, don't try to sell it'...
Alun Lewis is a telecommunications writer and consultant. firstname.lastname@example.org