The automated speech recognition (ASR) market is booming and is expected to be worth $31.82 billion by 2025. While much of ASR development is being used to power consumer products and virtual assistants such as Amazon’s Alexa, the boom in speech recognition will have a dramatic impact on contact centers as well. As ASR technology becomes more sophisticated, it will play a larger role in specialized markets such as finance, medicine, and travel, where industry-specific terms and language could present a problem.

As ASR technology becomes more sophisticated, it will play a larger role in specialized markets such as finance, medicine, and travel, where industry-specific terms and language could present a problem.

Using ASR as part of IVR open-end surveys provides call centers with powerful new tools to assess and improve customer experience (CX). Customer responses can be captured, automatically transcribed, and used for follow-up, call center training, and analytics. However, there are a variety of factors that can affect the accuracy of ASR transcripts, such as voice quality, accents, background noise, and industry-specific terminology. With voice analytics becoming more important for call center interactions, transcription accuracy is critical.

Applying Industry-specific IVRs

ASR technology has to be tuned to meet the unique needs of different markets. Machine learning and artificial intelligence help refine ASR software to improve accuracy, but if you start with an industry-specific solution then you are ahead of the game.

Pinpoint Research has introduced three industry-specific IVR solutions that include ASR transcription:

• Pharmaceuticals and Healthcare – Our Healthcare ASR model has been tuned to recognized drug names and medical terminology and has already been used by health insurance and pharmaceutical companies. The model offers rich training data covering issues such as pre-authorization for procedures, insurance claim disputes, disease-specific queries, and other medical terms and conditions. Although specifically tuned for healthcare and pharmaceutical applications, it can easily be extended for hospitals, care facilities, and other healthcare applications.
• Banking and Financial Services – The financial services market has its own lexicon, so we developed our Financial ASR model with lenders in mind. We developed the model using call center data from financial institutions to cover loan-related interactions, but it can be readily expanded to cover additional services with additional audio training data.
• Airlines and Travel – Accurately capturing interactions about travel can be especially tricky because of destinations and airport names. That’s why we developed an Airline model with airline operators in mind. Our Airline model is loaded with common industry terminology, including city and airport names, and is ideal for generating transcripts for call center training and data for analytics.

Accurately capturing industry-specific terminology as part of IVR surveys provides a rich repository of data for assessing CX, analytics, call center training, and other applications. You need the right ASR model if you want rich voice data that has real value. If you don’t have the best quality transcripts to begin with, you can’t expect meaningful results.

How many times have you contacted a call center only to be greeted with an automated voice that says, “This call may be recorded for training purposes”? Recording inbound calls have become routine. However, does anyone really use these recordings? Do they really have sufficient value for training purposes to sift through every call, or is there more to be gained from customer response transcripts?

Smart call centers are using call recording to identify problems with customers and improve their products and operations. You can only gain so much value from recorded calls for customer service training, but there is a wealth of data to be mined from call recording combined with customer survey data. Most call centers don’t bother to sift through call recording data because it is too time consuming and expensive.

Speech to Text Yields a Wealth of Data

A study by the Customer Contact Council showed that customer loyalty is built by providing seamless interactions and making the customer work less to get what they want. Trying to exceed customer expectations could be an impossible task, but the research shows that if you can meet the expectations of the customer experience you are well on your way.\

When a customer reaches out to the call center, it means they need help so the level of customer satisfaction is already dropping. Call center interactions are when you talk to customer when they have an issue. Rather than focusing on using the exchange to train reps to deal with problems that call center exchange could help identify issues you can eliminate in the future.

What is surprising is how few brands actually use recorded calls to learn more about their customers. By converting calls into transcripts it’s a relatively simple matter to apply text analytics application to identify key themes, and sentiment to identify opportunities to improve the customer experience in your call center interaction

When you take customer call transcripts, operational data, survey data, and other data sources, such as sales reports or call volume, you can learn a lot about what your customers expect from your brand. Once you start transcribing and archiving customer call center conversations you have a rich data repository that can be integrated with text analytics to power CX reporting.

The first step is work with a partner who can help you understand the quality of your voice recordings, and pilot a process to transcribe voice data to convert them into a data format that is searchable and machine readable for artificial intelligence. The faster you can convert speech to text, the sooner it has tangible value.

Making Call Recordings Actionable

Speech-to-text technology has advanced to the point where audio recordings can be transcribed automatically making them easy to archive and easier to search and analyze quickly to highlight customer issues.

Call transcriptions yield even more insight when coupled with IVR survey open ends. Research shows that consumers prefer to respond to IVR surveys delivered using the same channel. Using open-ended voice surveys that can be transcribed along with the call center exchange even more about customer perception and CX. Combining best of breed ASR transcription and analytics resulting in data presented in a form that can support customer experience policy and products and tools to enhance agent performance.

By combining IVR voice survey data resulting with call recording analytics, you have the opportunity to understand the root causes for low NPS and brand churn.   The combination of survey experiential data and recording analysis also can become a predictor or NPS.

It’s time that all brands started using those call center recordings to actually listen to customers and hear what they have to say. An archive of transcribed customer calls can be invaluable for guiding company strategy and learning more about brand perception. When coupled with analytics, call center conversations can be an incredibly powerful tool that improves CX and promotes customer loyalty by making the call center experience seamless.

Customer satisfaction surveys have become an essential part of marketing. Many online sales, telephone support call, and customer exchange is followed by a customer satisfaction survey, and most of those surveys are absolutely useless. When the customer satisfaction ratings become targets, then the insight from customer feedback disappears. To be effective, customer surveys need to be open, unbiased, and untargeted to get authentic customer views; insights that are actionable. That’s why we are seeing more call center clients adopting voice driven research™, leveraging voice open ends increasing the quality and quantity of qualitative open end response data.

The value of customer satisfaction research has been demonstrated again and again. According to a recent survey by The Temkin Group, 87 percent of customers surveys indicated that their 2016 investment in customer experience (CX) research has a positive impact on business, as opposed to 79 percent in 2015. As a result, 66 percent of companies are increasing spending on CX studies, including a nine-fold increase in headcount. Most of that spending is going into voice of the customer (VoC) software, predictive analytics, and experience design.

Customer satisfaction surveys can help fuel CX research, but only if they gather unbiased results that reflect the true voice of the customer.

Stacking the Deck with Scored Surveys

What’s wrong with customer satisfaction surveys is they become self-fulfilling prophecy. Too often, a company uses customer surveys as a means to gauge call center or service performance, using a score as a target threshold. Surveys become a “shop, rate, reward” system integrated into customer interaction and designed to generate a score, or target.

This type of metric is flawed when gauging true customer experience. Goodhart’s Law, developed by British economist Charles Goodhart, states that once a measure becomes a target, it can no longer be used for valid measurement. By way of example, Goodhart points to Soviet factories producing nails according to a target. If the target is to produce a specific number of nails, the factory produces more tiny nails to meet the target. If the target is based on weight, the same factory retools to producer fewer, heavier nails. In both cases, the target doesn’t reflect true value.

Now consider the same truth as it relates to customer satisfaction surveys. If a survey follows a support call or customer call, the call center representative will ask for a favorable rating which will skew the results:

Rep: “Did I address all of your concerns, Mr Jones?”
Customer: “Yes, thank you.”
Rep: “Great, then can I count on you for a 10 out of 10 score on our customer satisfaction survey?
Customer: “Yeah, okay.”

In general, everyone wants to be helpful so you will consistently get survey results that skew positive. In fact, many companies assess call center performance using post-call surveys and anything less than a 10 out of 10 is considered unsatisfactory.

To get true customer sentiment and honest customer feedback, you have to separate the survey from employee scoring.

The Value of Self Expression

The best approach for accurate, actionable insight is to avoid scoring altogether and opt for open ends, where customers get to share their true opinion.

Open-ended questions offer a number of advantages, especially if you are seeking to gather actionable data. For example:

1. You get answers you don’t expect. Open ends allow customers to offer observations, not just responses, so you are likely to uncover hidden problems, such as long wait times, problems with your phone menu, and other annoyances. Removing impediments to better customer service means better returns
2. Complete responses to complex questions. Open ends allow customer to explain. This has at least two benefits: 1) it provides a more comprehensive response to the questions and 2) it cements customer relations by giving the caller a chance to be heard.
3. It tells you how customers think.

How you administer open end questions also has an impact on the quality of response. You can use written surveys, which will give you genuine feedback that is easy to add to your database for data search. However, given the rising degree of survey fatigue, consumers are losing patience with answering surveys, and written surveys may not generate the response rate you want.

Voice capture adds another layer of authenticity to feedback surveys. Mobile phones in particular have become a popular tool for IVR surveys. Two-thirds of Americans own smartphones and use them for online data access as well as telephone calls, so targeting smartphone users gives them options on how to take the survey. We have found that most consumers prefer voice driven surveys because they are faster and easier, which is an advantage for call centers since voice responses capture the true voice of the customer.

With new speech-to-text transcription technologies, companies are seeing an error rate of 7 percent or less for transcriptions. This makes transcribed voice driven research accurate and incredibly useful for call centers. Transcriptions can be generated in near real-time, making it easy to search algorithms that can flag customer issues that need immediate attention. And the original voice file can be preserved for tone and deeper insight. The result is customer satisfaction response that is easy to capture, more accurate, and most importantly, immediately actionable.

Targeting customer satisfaction provides a false positive that will lead your company in the wrong direction. If you truly want to gauge satisfaction, don’t lead your customers, just ask them what they think. If you provide a forthright and easy way to respond, such as voice response, then you will get more accurate results that will lead to better business decisions.

With the evolution of consumer marketing and customer research terms change their meaning. One term that continues to be bandied about to the point it is no longer clearly defines is Voice of the Customer (VoC). VoC is an essential part of customer research, but the term has become a catch-all for any type of customer experience or interaction. There are parameters around how to define and think about VoC that are important to appreciate to develop an effective customer research program.

The Voice of the Customer is an old concept. Organizations have been taking the pulse of the people for generations. Steve Allen started the concept of the Man on the Street interview for the old Tonight Show back in the 1950s, doing impromptu interviews to hear the vox populi. Getting people’s opinions can be powerful and provide valuable insight into the perception of your brand or product.

VoC actually began as a means to gather information as part of product development. The goal was to gather information from target consumers to determine their wants and needs. VoC was a means of capturing insight about customer requirements in a hierarchical manner in order to make the information actionable, i.e. give the product team an understanding of how to develop the right features and functions.

What has happened over time is that the definition of “voice of the customer” has expanded beyond its original, narrow definition to support market research. VoC assessment has become part of consumer research methodology and is used to identify customer needs, trends, behaviors, and patterns.

What’s important to remember is that VoC research is designed to address the customer experience BEFORE there is physical interaction with the product or brand. Once the customer has had the product experience, then you move into a different phase or research based around customer feedback.

So what are the objectives of a solid VoC program? Consider these:

1.  Strive to understand the customers’ expectations from the brand as part of your research.
   2. Understand how VoC helps shape a brand’s product or service to align with the customers’ needs.
   3. Understand how the brand can address the demands of the customers, as opposed to how the customer perceives the product or service.
   4. How can VoC expectations be applied to bring about strategic changes across the organization.

VoC research is designed specifically to learn more about customers’ likes, dislikes, preferences, and patterns. By its nature, VoC research is anticipatory, and requires interviewees to project themselves into the brand experience and voice their expectations. An IVR surveys can be an invaluable tool for VoC research if it is applied properly. This is especially true of newer VoC systems that integrate voice driven IVR surveys that actually capture and auto transcribe the open-end responses of customers, giving researchers detailed qualitative information about customer attitudes and emotions.

Not all customer research touchpoints fall into the category of VoC, but if you are using customer research to assess customer expectations or acceptance of a new product or brand experience then you should plan your VoC research accordingly.