Those of you who are music nerds, as I am, probably recognize the sentences below as the first two lines of the song “Sentimental Journey.” The song was published in 1944 by bandleader Les Brown; he of Band of Renown fame. Les Brown is known for a lot of things besides being renowned. For example, he traveled with his big band from 1938 to 2000, including 50 years that Les and the Renown were the band behind Bob Hope. Together, they did 18 USO tours for American troops around the world.
Gonna take a sentimental journey, Gonna set my heart at ease.
The song “Sentimental Journey” tells the story of someone who is about to take a train ride to a place where they have a strong emotional attachment. Doris Day had her first No. 1 hit with the song in 1945. Since then, the song has been covered by dozens of artists including Bob Dylan, who recorded the song for his 2017 album Triplicate.
The contact center industry has been taking a sentimental journey of its own for the past few years. Sentiment analysis has become a high-profile solution for contact centers in this era of customer experience (CX), as customer service professionals seek to evaluate and better understand the language used by customers. Sentiment analysis provides CX professionals greater insight into words and expressions used during a customer contact in order to determine if the expressions used are favorable, unfavorable or neutral. More than discovering what customers are talking about, sentiment analysis lets users know how the customer feels.
The roots of sentiment analysis go back to the early 2000s and the increasing popularity of online reviews. Companies whose products or businesses were being reviewed were looking for as much business intelligence as possible in the online reviews of their products, beyond just the words left in the review. At the same time, social media was also becoming a force in communications of all kinds, including customer communications. Once again, companies were interested in more than just the words posted. They wanted to know more about the emotions behind the posted text.
With advancement in natural language understanding and processing, combined with innovations in speech analytics, it was only a matter of time before sentiment analysis would be applied to the spoken word, and that happened about 10 years ago. Although an inexact science at the time, voice sentiment analytics paved the way for greater understanding of unstructured data beyond the application of speech analytics in recorded transactions.
The value of sentiment analysis to the written word is obvious. Understanding the emotion behind the text of a social media post or blog adds a degree of value to whatever has been written and would otherwise be left to subjective interpretation. But voice? Isn’t it pretty clear what the caller’s emotional state might be during a call simply by listening to the tone of voice? I’m pretty confident in my ability to discern a sarcastic tone of voice (I have two daughters), angry words, shouting, growling, etc. Besides, isn’t that what quality management is for? We listen to the recorded interaction and determine the emotional state of the customer. And if listening isn’t enough, we have speech analytics to help us sort through the noise for words and expressions that are meaningful to our CX objectives.
Tammy Marinac, Calabrio To help clarify some of this uncertainty around sentiment analysis, I called on TAMMY MARINAC who is the Product Marketing Manager for Analytics and Advanced Reporting at Calabrio. Calabrio has been pushing the sentiment analysis envelope recently so I asked her to weigh in on the use of sentiment analysis in the customer experience.
“First of all,” Marinac told me, “Sentiment analysis is an automated process so it doesn’t require the human intervention that both quality management and speech analytics does. It goes beyond word-spotting and searching out predefined key phrases to determine sentiment by utterance, and looks at those utterances in the context of the words around them. For example, you might use speech analytics to find instances where callers said ‘cancel my service’ and assume that all those customers feel negatively about your company. But if the context around that phrase tells us that they’re canceling their service because they’re moving out of the area, it becomes more of a neutral sentiment.
“Speech analytics in this case can help you identify all your customers who are at risk of attrition,” Marinac said. “But it’s the sentiment analysis that can be applied to those speech results that helps you decide what to do with that information. A customer who wants to cancel because she’s moving out of the area, or recently lost a job, might still be a promoter of your product. There might still be positive sentiment there that you can leverage in referral programs or online reviews. I recommend that you use speech analytics to find all customers at risk of attrition. And then use sentiment analysis to filter the attrition risks by positive, negative and neutral sentiment scores.”
Back to my quality management argument. Why couldn’t sentiment qualification also be determined by listening to the call? Or, if random call selection means risking missing any particular call, you could apply speech analytics to 100% of the calls.
“There’s so much more to sentiment analysis than simply listening to the call,” Calabrio’s Marinac continued. “Sentiment analysis provides an automated score easily interpreted and usable by supervisors, managers and others throughout the enterprise.”
So, if we’re capturing the sentiment behind the words on the customer side, why isn’t sentiment analysis being used as a tool to improve employee engagement? How could sentiment analysis impact the employee experience as well as the customer experience?
“Sentiment analysis is most definitely being used to improve the employee experience and monitor performance,” Marinac explained. “You can view sentiment scores by agent, by team, by group and so on. You can also correlate the sentiment score with other contact center data points, like average handle time. Sentiment analysis and scores can provide guidance to supervisors and managers in terms of identifying agents with the greatest need of coaching. It can also be used to identify the agents with the highest sentiment scores and use their customer care skills as a template to improve performance across the board. Sentiment analysis provides a much more holistic view of the customer experience.”
Radial Inc. is a business process outsourcer (BPO) running customer care centers in North America, Central America and Europe. In the highly competitive outsourcer market, Radial is always looking for a competitive edge, and with their clients demanding more data and usable customer information, Radial looked toward an advanced analytics platform for a solution.
Working with Calabrio, Radial implemented a comprehensive CX platform solution including sentiment analysis. Initial sentiment analysis results revealed high rates of neutral sentiment. Radial analysts evaluated the calls with the less positive sentiment and identified key topics that correlated with that sentiment. They then built training materials that guided agents into using more positive phrasing and implemented highly focused training sessions.
In just one week, Radial saw positive sentiment jump from 17% to 93%. Positive improvement in sentiment analysis also drove improvement in other key performance indicators.
Sentiment analysis is now hitting its stride in the contact center. Able to positively impact both the customer experience and the agent experience, sentiment analysis is ideally positioned to meet the evolving needs of the contact center industry in today’s customer experience economy.