Sentiment Grows in Significance as COVID Surges
Companies have been collecting customer feedback for decades, but it has taken on a new importance as they try to maintain shaky customer relationships in the face of COVID-19 lockdowns. That feedback is only useful, however, if companies can understand the full meaning behind it.
Key to that level of understanding is the sentiment that underpins any feedback received. Sentiment is the feeling and emotion that customers have about the company as a whole, its products and services, its personnel, and the interactions they have with the company. Customer sentiment is more than just a list of keywords; it’s a complex set of subjective targets reflected in the tone of their posts on social media, their online reviews, their voice on phone calls, or the words used during chat sessions or any other interaction that brings them in contact with the company.
Hopefully, most customer sentiment is positive, but that isn’t always the case, of course. Inevitably, some customers will have negative or neutral feelings, but that information is just as valuable.
Customer sentiment goes far beyond basic emotion, too. It is scored much like a star rating on Amazon or Yelp, according to Rick Britt, vice president of artificial intelligence at CallMiner. “You can think of a sentiment score as a line on a single axis. When you talk about emotion, we are adding a second axis to that.”
In other words, sentiment is basically the positive or negative feeling about an experience. Emotion is the intensity of that feeling, Britt explains. Using Amazon as an example, someone might provide a one-star review—sentiment—but if they explain why they gave a one-star review, that provides the emotion component.
“Everybody is very good at reading the positive and negative because it’s so well researched,” Britt says.
Sentiment analysis considers the linguistic elements and the acoustic (if a voice call rather than an email or chat exchange), says Daniel Ziv, Verint’s vice president of global product strategy for speech and text analytics. “We look at the words and how they are being said in the flow of the conversation. For sentiment, words are usually a little more accurate than tone. Tone has impact, especially in lab settings. But in noisy contact center settings, tone can easily be missed and misinterpreted. Tone alone tends to be much less accurate, even though there are acoustic elements that tend to indicate emotions.”
For that reason, Verint’s contact analytics technology, as well as that from other vendors, will also consider gaps in the conversation, for example, which Ziv says could signal a lack of knowledge on the agent’s side.
Though some analytics vendors have launched solutions with more comprehensive capabilities, the basic technology itself isn’t really new. Some of the technology has been around for a decade or more.
“The technology has become more commoditized,” says Christian Wettre, senior vice president and general manager of Sugar Platform at SugarCRM. “They used to be more specialized companies built around just delivering the voice [sentiment]. Now that it’s become a little more commoditized, it can be used as a service by just about anybody. Increasingly, vendors are able to do more with the voice itself, instead of just the infrastructure delivering the voice. They’re able to turn the voice into information. You can say words or you can shout words. The words say the same thing, but the meanings are very different.”
Whether looking at the meaning of the words themselves or also including acoustic data, customer sentiment analysis technology provides companies with much better understanding of customer opinions than basic Net Promoter Score surveys ever could, Wettre says. He points out that many of those surveys are never completed, whereas customer sentiment analytics can be used for every interaction.
The technology’s algorithms have advanced markedly in the past year, according to Britt. “We use a particular algorithmic model that is very powerful. It works very well for us.”
SENTIMENT’S IMPORTANCE EXPANDS BEYOND CALL CENTERS
Companies are now seeing customer sentiment analysis as much more important in providing good customer experiences and in customer retention, according to several experts.
“In the past, a lot of interactions weren’t that emotional,” Ziv says. “Now, every interaction has moments of emotion or sentiment. That becomes a focus, particularly when you are talking about human-to-human interactions.”
So Verint has extended its analytics offerings from speech and text analytics to also include emotion detection, Ziv says. “We’ve added positive and negative emotion to every single interaction. Now we can share [a customer sentiment report] across an entire enterprise, not just within the contact center.”
The marketing, web app, and other teams within the organization need to know what’s driving positive and negative sentiment, Ziv adds. “It’s almost everyone’s job in a lot of organizations, and even for back-office operations, to make interfacing with the customer part of their jobs.”
That responsibility goes all the way from customer-facing contact center agents to C-suite executives, who now have shifted focus more than ever on customer sentiment and NPS scores, Ziv says.
Natalie Petouhoff, chief business strategist at Genesys, agrees: “Businesses are moving away from being very business focused to actually observing their customer and employee experiences.”
Historically, business have been built to maximize efficiency, with the assembly line as a prime example, according to Petouhoff. Now there is a shift to empathy and putting oneself in the customer’s shoes.
She points to Airbnb, Apple, and Tesla as companies that are doing a good job of incorporating sentiment in building customer journeys.
The increased organizational focus on sentiment and emotion is due, at least in part, to the COVID-19 pandemic, which heightened the emotions of nearly everyone as their lives drastically changed in a short span of time, Ziv says.
“Customer expectations have been raised,” Ziv adds. “If Amazon does a good job with a problem—if I need something and it’s here within an hour—that’s now my expectation from everybody.”
So a much higher percentage of interactions that Verint analyzes have some form of sentiment or emotion than just a couple of years ago, according to Ziv.
Though some providers promote the ability of their solutions to perform real-time sentiment analysis, for as many as 95 percent of interactions, post-call analysis is sufficient for most needs, according to Rachel Lane, a solution principal at Medallia.
Similarly, most companies rely primarily on sentiment scores and analysis, according to Britt, because accurately measuring and analyzing emotion is more complex and more expensive. With words alone, companies can be about 75 percent accurate in judging emotion (i.e., someone is slightly angry, very angry, extremely angry, etc.), which is good enough for most situations.
It can cost perhaps 10 times more to accurately determine emotion than sentiment, according to Britt. So companies should determine whether they stand to receive a positive return for the additional investment. That’s at least for the time being, Britt adds, noting that the cost of the technology is continuing to decrease.
Many companies are also investing more in customer satisfaction analysis as they look to better home in on meeting customer expectations.
For example, a large rideshare company is spending aggressively on the technology, according to Britt. “They want to keep their drivers happy, and they want to keep their riders happy, because they compete with the other large rideshare company for both.”
Ziv agrees with Lane that the analysis of sentiment and emotion is shifting to real time, crediting artificial intelligence for the ability to meet much of that demand.
“We’ve added sentiment analysis in real time,” Ziv says proudly. Providing that capability to agents helps them better handle stressful situations, he says.
In mid-November, Verint launched AI-driven, real-time agent assist (RTAA) capabilities, including real-time sentiment analysis and work assist functionality, in the Verint Customer Engagement Cloud Platform. The new functionality is designed to help organizations connect with customers on a more empathic and human level, as well as to provide agents with real-time context for customer sentiment and intent. Verint’s technology provides notifications, knowledge, and reminders on the desktop to guide agents on next best actions.
“While nearly all businesses know the importance of customer empathy, most struggle to deliver it on a consistent basis, especially given the work-from-anywhere contact center environment,” Ziv says. “Our latest innovation supports the delivery of exceptional experiences aligned with the customer’s emotional state and intent for more impactful and meaningful interactions, guiding agents in real time toward the best possible outcome.”
Real-time sentiment analysis is going to grow over the next 12 to 24 months, Lane predicts. “Up until now, it’s been a real nice-to-have, but I think that we’re going to see that use case expand fairly dramatically.”
AIDING HUMAN INTERACTIONS
Though chatbots continue to handle an increasing percentage of customer interactions, human agents are still needed to handle many interactions, most experts agree. But because many agents today are unseasoned, sentiment analysis can help them better recognize when a customer is a little upset or so upset that he is on the verge of ending the relationship with the company. Sentiment analysis will also work with other systems to provide agents with immediate advice on how to handle customer complaints and to pop up forms or other information that a customer might need, Lane adds. “It seems like a very seamless engagement, but we’re actually connecting a series of potentially different vendors and connecting them to create that customer journey,” she states.
Though customer sentiment analytics providers tout their solutions as enabling companies to react more quickly to preserve customer relationships, most are analyzing only single interactions at a single point of time, which is a huge weakness, according to Puneet Mehta, founder and CEO of Netomi. “Business relationships, like personal relationships, are not built on a single moment of time.”
For example, a customer calling about an error on a shipment or billing might be noticeably upset. But that customer will be much more upset if it’s a repeat occurrence or if she has had to communicate with the contact center multiple times about the same issue.
“At Netomi, we look at a series of interactions,” says Mehta. “We also look at explicit versus implied sentiment. Not everyone is expressive in the same way.”
One person might use all caps in a text message, and another might be shouting on a voice call, but they could be doing those things just to get attention, even if their issue is relatively minor. Historical analytics can help determine if the anger is simply to get attention or if the issue has escalated from something minor to something much more pressing.
Petouhoff says incorporating that type of sentiment over a customer’s history of interactions can help companies build better customer journeys. She points out that there’s a close correlation between the length of time a customer has been interacting with a company (including time on the website, with its interactive voice response system, or with an agent) and negative sentiment.
VIDEO LENDS A HAND
One of the newest developments in customer sentiment analysis is video, according to Lane. Her company, Medallia, acquired LivingLens in early 2020, with the following goals in mind:
- add video to feedback strategies to obtain much richer insights;
- capture six times more information than from an equivalent open-end text response;
- seamlessly collect video feedback from mobile app, surveys or Zoom meetings; and
- better understand customers and employees with AI that interprets expression and emotion.
LivingLens also offers users automated multilingual speech-to-text transcriptions for each video.
“The whole premise is to engage with the customer in the way that they engage with you, in the channel that they engage with you,” Lane explains. “They just give you that verbatim feedback there. With LivingLens, we’re able to add the emotion detection (from facial recognition detecting frowns, smiles, etc.) from the video as well.
“The great thing about it is that over the past two years customers have gotten really used to using video [apps] like FaceTime and Zoom, so it’s a much more natural way of giving feedback. And that’s why the feedback is so rich,” she adds.
And while most of the technology is undergoing an early renaissance, “in a year from now, you’re going to see a lot more success stories and use cases,” Ziv predicts.
Phillip Britt is a freelance writer based in the Chicago area. He can be reached at email@example.com.