I founded McIntosh & Associates to help businesses improve their customer experience by improving the performance of their contact center. We apply a data-driven approach to identify opportunities for improving contact center performance, and customer satisfaction.
One of the thorniest issues we encounter is voice quality. It’s obvious that good voice quality is an important factor in delivering a great customer experience. Good voice quality is expected by your customers and its absence will result in significant dissatisfaction and negatively impact your NPS results.
But consistently delivering on good voice quality is easier said than done for two main reasons. First, delivering good voice quality depends on many factors working correctly. Some of these factors are under your control, and some are not. The second issue is that — for most organizations — it is very difficult to consistently know with any confidence how your voice quality is performing.
What Drives Poor Voice Quality?
First, let’s discuss some of the factors that impact voice quality. Any point on the chain from customer to agent can cause a drop in Mean Opinion Scores (MOS), the measure of voice quality. Anything from a caller having a hoarse throat, to bad audio equipment can degrade voice quality. Network and equipment effects are the most apparent and measurable and include:
- Codec incompatibility, the software that allows us to efficiently move voice traffic by compressing and decompressing the voice signal
- Excessive transfers which may result in a loss of volume with each transfer
- The caller’s voice network provider — the most obvious example is a poor cell signal based on location
- Bandwidth constraints which can result in packet loss leading to jitter (voice signal breaking up), echoing, low volume, and other conditions that make it difficult for the customer and the agent to communicate effectively
- Hardware issues such as headset or softphone incompatibilities
How Do You Measure Voice Quality: Mean Opinion Score (MOS)
MOS is the most common measure for voice quality, and historically derived from the feedback of listeners who would sit in a "quiet room" and score a telephone call quality as they perceived it. Today, MOS is derived via algorithms that accurately measure our VoIP network performance and can categorize the quality of the voice signal that callers experience when on a call.
MOS is a five-point scale running from a score of 1 (inferior voice quality that cannot support effective communications) to 5 (excellent voice quality, comparable to a face-to-face conversation). MOS of 3.5 or higher allow us to communicate effectively while scores below 3.5 reflect a degradation in voice quality and make it difficult to understand the caller or achieve call resolution.
The expected range for VoIP calls is 3.5-4.2. Anything in that range is considered satisfactory; anything below that range calls for action.
How Do You Know When Voice Quality Is Bad?
Most organizations only learn of a voice quality problem when there are escalating complaints from customers or agents. However, 96% of customers won’t report when they’ve had problems. Often, the customer and agent will struggle through, talk louder, or repeat themselves. Not only is this a bad experience for your customer, but it drives up average handle times. And, if the voice quality is bad enough, the customer may simply hang up, and call back with the hopes of getting a better line on the next call.
Even if a customer or agent reports an issue, it’s tough to replicate. You likely won’t know details about times, locations, equipment, etc. This is all critical information you need in order to troubleshoot and resolve the issues.
One of the things that McIntosh & Associates helps businesses with is voice quality issues, and I want to share with you one particular example. We were hired by a US insurance provider to measure the voice quality across internal and vendor end-points, identify the impact of transfers, and identify site-specific issues for root cause analysis of voice quality issues.
To answer these questions, we:
- Called and recorded test 800 numbers using pre-recorded industry standard scripts where the MOS equaled 5
- Created and ran test call scenarios ranging from simple to complex with multiple calls per scenario placed to each location
- Measured voice quality for both the agent and caller
Our analysis showed some interesting results. First, we confirmed that transferring calls in the client’s environment negatively impacted MOS. On average, MOS degraded by 6% for each call transfer leg. Secondly, the MOS for internal sites was higher than for vendor sites, and by 15%, so a substantial difference. Finally, part of our scope was to compare MOS for their current state infrastructure and their future state. Our research showed that MOS for current state infrastructure was 2.67 (unacceptable) and for future state increased to 3.43 (marginal). And, while that was a significantly higher MOS, it was still below the target range — largely due to the issue with transfers.
To get them from a marginal to an acceptable MOS, we recommended that they revisit their transfer policy and agent training on appropriate transfer protocols. We also suggested ongoing MOS testing, using Cyara Pulse, to identify process, site-specific, and network issues in order to develop plans to remediate the issues.
Contact McIntosh & Associates to learn how we can help you improve your contact center performance, and customer satisfaction.
Contact Cyara today to learn more about how Pulse can help with ongoing monitoring.