Speech or voice analytics have been growing in popularity and use for many years. Possibly the first encounter that you will have had with speech analytics was the Microsoft program, Encarta. Encarta was an encyclopedia program developed and put into production in 1993. However, the use of voice recognition in this program did not outlast the competition. Before we get into the article too much, the terms voice analysis and speech analytics are synonymous and used interchangeably throughout the article.
Voice analytics is a common feature in today’s society, with many of the phone calls that you make asking you to say your name or other information. Along with the time came some extreme advancements in technology, too. In this article, we will look at some of the critical aspects of speech or voice analysis within call centers. Some of those points are:
While speech analytics have been available to companies for a long time, there are still questions raised about it, how it works, and if it would benefit your call center. Therefore, we will try to answer those questions here.
Speech analysis is no different from speech recognition or voice analytics. However, speech recognition is often found during the beginning of a call to identify users’ identity, intentions, and security. Speech analysis, on the other hand, is more commonly found useful in the quality assurance realms of call centers.
In the past, if agents had to track what was said, they had to write it down. While that is still common practice for many agents with large call volumes, the requirement to record everything said is a thing of the past. Almost every phone call that you make to a company today has an initial statement of “this call may be recorded for quality and training purposes.”
It is at that point in which analytics come into play. That will analyze what the caller and agent said and transform it into an easy to understand, understandable format, most of the time. We will go into how it does what it does and how good it is during this article.
So, the basis of speech analysis is picking up on certain aspects of a call and making notes for you. There are two critical ways in which software does that:
The software will convert your words into phonemes, small, recognizable sounds in speech, of which there are only 44 in the English language. They are the sounds that you make to talk, such as “oo” in “food, or “ea” in easy. One of the main issues with speech to text is that it must be a very sophisticated software to understand your speech.
The main issues with speech to text software are accents and words that are pronounced the same for different meanings:
Then you have the issues of accent where words sound very similar, such as:
Many words sound the same when broken down into phonemes, even more, when there are different accents involved. There are 42 recognized accents in the USA alone, with even more in places like the UK. That fact can lead to many issues within the speech to text software and make it sometimes unreadable after completion.
LVCSR or Large Vocabulary Conversational Speech Recognition is a newer technology that matches whole words instead of the phonemes. It requires a vast database compared to the phonetic alternative, and it is slower to process the data. However, it is more accurate. The primary consideration with LVCSR is that it needs to have hundreds of thousands of matches for words instead of sounds.
“Call centers use voice technology.”
Phonetics will see that sentence with each sound; “ca” “ll” “cen” “ters,” etc.
However, LVCSR uses direct word correlation and puts them into a sentence with an “n” value. The “n” value is how many words the software can process at one time; for example, “n-2” would see the sentence like this:
If all of the words fit together, it will produce a complete sentence. Therefore, it is a much slower voice analytics method, but it is a lot more precise as developers enter different accented words into the database.
There are many uses for voice technology. However, we are mainly concerned with call centers. Even so, there are still many uses for the technology in call centers, such as:
We will look at those sections and how the design of voice technology is to help them.
Compliance is the most significant issue faced by some of the newest regulations, such as GDPR. However, other compliance issues like Do Not Call (DNC) and sensitive personal data such as bank detail recording also fall under compliance. The regulations that face call centers grow almost daily, and some industries like finance and banking face it even more.Speech analysis should ensure enforcement of guidelines set out by the TCPA (Telephone Consumer Protection Act) and prove compliance should someone make a complaint. Therefore, protecting both the consumer and the company alike. Some of the critical aspects of compliance are:
Seeing a process in the form of words and numbers allows you to streamline a specific process to maximize agent efficiency. Some of the methods that you can improve via voice technology are:
Constant monitoring of a high percentage of calls allows you to identify issues with pinpoint accuracy. When you identify problems that you face with some regularity, you can coach agents in those areas, or for those points with a better understanding between the training staff and the agent.
We have many articles about coaching and how it can help your business. Here are a select few that you may find useful:
When you have all of the above items working correctly, you will find that your customer satisfaction will improve. Therefore, increasing your retention and acquisition rates.
The only way that you can genuinely measure customer service levels and standards is through customer satisfaction. Speech analysis allows you to see various aspects of customer satisfaction by data. For example:
Marketing is the key to business growth. Some large companies can spend millions of dollars on marketing with little to no return on investment. However, when you have access to thousands of hours worth of “the customer’s voice,” you can analyze and use it to your advantage.
Using voice recordings can save you money and increase marketing efficiency without the extra cost. Not only that, though, it will also help you to gauge how well a campaign is working. For example, if you have 100 calls a day about a specific product, then advertise it, you will see if the number of requests increases by an amount suitable for the money spent on the marketing.
In our opinion, the expectations of voice analytics are often too high. While there are some valid expectations, some of them are unfulfilled. Here are some of the categories that people expect from speech or voice analytics:
Let’s have a look at what those expectations are in a little more detail:
Complete compliance is a comprehensive statement and often overused. Gaining full compliance takes much more than voice analytics alone. While you will have better visibility of compliance issues that you face, it will not resolve them for you. For example, if you have an agent that never states, “this call is recorded,” do you think they will when you have voice technology? It isn’t very likely.
Therefore, you still need to take control of individual circumstances that are causing you to fail.
Again, as with compliance, voice analysis will not directly impact your call to conversion rate. It will, however, increase your visibility of where you may be failing. However, that is only possible through a high percentage of accuracy and proper coaching.
The accuracy expectation is, by far, the most overrated of them all. There is no possible way of gaining 100% accuracy through voice analytics, yet. As you saw earlier, there are many accents in just one language—forty-two of those in the USA, 7 in Canada, and 56 in the UK. Furthermore, there are numerous accents for each of the 56 recognized British accents, maybe even hundreds for each.
Some of those accents do not use the same word for the same meaning, either. Therefore, relying on computer-generated analytics alone is NOT going to give you anywhere near 100% accuracy.
So, what are the real benefits of voice technology? We have written a similar article about voice technology here.
There are many benefits to any technology, and VT is no different. Before we get into the downfalls of the technology, let’s take a look at what some of those benefits are:
Any technology could be the greatest thing in the world at the time of release. So, let’s say that you buy that latest and greatest voice tech. That will come with consequences:
A simple answer to that question is no. There is no possible way that a machine can replace human hearing. There are various pros to using voice analysis, of course. However, there are even more cons to relying on computers alone. For example, if you have a small call center, you would be in a position where you can almost solely rely on humans for QA instead of voice analytics.
When you get to more extensive and more substantial call centers, with say 1,000+ agents, you may require some technological assistance. Although, you will still need to bear in mind that there are some severe downfalls to committing to speech analysis alone:
Validation of anything requires a second check. Companies that rely on voice analytics alone are not in a position to check all of the points within scorecards to validate the false positives and false negatives. While speech or voice analysis is excellent for some aspects of testing for compliance, it is not always the best fit. Validation is a very time-consuming process but also a required one for higher accuracy rates.
Many companies fall under an umbrella of stringent guidelines and regulations where words in a specific statement cannot change. As I have already said, finance, banking, and insurance are some of the more common industries that require such statements. Voice analytics may incorrectly score agents on terms like that because it does not pick us the exact wording due to complications. They can include accents, etc.
The only way to get around that would be to “train” your technology for individual agents who work for you. However, with such a high attrition rate in the call center, that would be completely unacceptable.
Call centers invest a lot of time and money into phone equipment. Customers, on the other hand, are a lot less reliable in their methods of calling. That can be a real issue, especially for machines, to pick up specific phonemes or whole words. However, humans have evolved to understand many word strings and how they sound even if they cannot hear them properly.
As we have already spoken about, another one of the most common issues is accents and dialects. Some of the more pronounced accents you find will be unable to be picked up by speech analysis. However, humans, on the other hand, can understand much more.
Voice or speech analytics certainly have a place in call centers, and we would not try to say otherwise. However, relying solely on technology is undoubtedly a wrong decision. If you would like to find out more about what Call Criteria and our technology and analysts can do to increase your accuracy, contact us here.
We look forward to hearing from you.