How improved sentiment analysis in voice-based communications can help increase sales

Understanding how customers feel about your company and its products is an ongoing challenge, especially as more transactions are conducted through e-commerce instead of face-to-face interactions. Nowhere is this challenge more keenly felt than in call centers.

SEE: Special report: Managing AI and ML in the enterprise (free PDF) (TechRepublic)

“Historically, call centers have been inbound calling centers, where agents answer questions from customers but don’t really get involved in selling or upselling,” said Matt Matsui, chief product officer at Calabrio, which provides customer experience intelligence software. 

Over the past few years, call center agent roles have begun to change; agents are getting better pay and are being moved into selling, too. To make this transition, call center agents need real-time intelligence about how customers feel about products and the company during a live conversation. Analytics applied to voice-based communications while the communications are happening helps deliver this intelligence.

“What companies like ours are focused on addressing is what I call the art of humanity,” Matsui said. “In other words, can technology understand people? Which customers are likely to cancel an order or their subscriptions, and which are likely to buy more?”

To aid this understanding, customer sentiment analytics aggregates information on the customer and produces an assessment of customer sentiment that a call-center agent sees while still on the phone or in a chat. This gives the agent an understanding of customer intent and can be used as a signpost to the agent about whether to try to solve an issue, calm a customer, or to take an opportunity to upsell.

At its most rudimentary level, sentiment analysis is a scientific data analytics technique that subtracts the number of negative words from the number of positive words in a conversation to come up with either a negative or a positive sentiment score. Research indicates that 80-85% of the time, the sentiment analysis generated by analytics conforms with what a human would deduce from a conversation.

Companies like Calabrio want to take analytics sentiment analysis one step further.

“We are taking our linguistics analytics down to the level of what we call ‘functional’ words,” Matsui said.

Functional words include pronouns (“us,” “we,” “you”), possessives (“yours,” “mine”), and articles (“a,” “they”). Functional words are not the context-type words that have a more direct bearing on how a conversation is going. But companies like Calabrio believe that functional words also play an important role in sentiment analysis.

“If a customer uses words like ‘you’ and ‘your’ instead of ‘us’ and ‘we'” in a conversation, it could be a clue to a call-center agent that the customer views the relationship as a more distant and possibly as an adversarial one,” Matsui said. “Similarly, if a customer says something like ‘The book is on a table,’ instead of ‘The book is on the table,’ it might signify a more distant relationship. In both cases, there are sentiment indicators that show whether the relationship is a closer or a more distant one.”

SEE: How AI and voice analytics can improve the call center experience (TechRepublic)

Matsui said that some of his company’s clients are using this additional layer of sentiment analytics to assist them with building customer advocacy skills in their call-center agents that in turn were leading to greater customer loyalty and sales.

“By encouraging customer advocacy, a company was encouraging its call-center agents to place themselves in the shoes of their customers,” Matsui said.

Transforming call-center agents who have historically been order takers and complaint handlers, isn’t an overnight process. Matsui suggested that companies interested in transforming their call centers and improving call center performance take small steps initially, perhaps by trying a single project.

“Sentiment analysis is still a relatively new technology in the call center, and including functional as well as contextual words in sentiment analytics is also relatively new,” Matsui said. “But it promises to be an indispensable tool for call-center agents to get to know their customers and to respond to these customers in positive ways that build loyalty and company success.”

Data, Analytics and AI Newsletter

Learn the latest news and best practices about data science, big data analytics, and artificial intelligence.
Delivered Mondays

Sign up today

Also see