Friday, August 15, 2008
Contact Professional Magazine: Harnessing the Value of Emotion Detection to Improve Customer Satisfaction
If you have had overlapping friendships, you will immediately “get” why harnessing the value of emotion detection is a crucial component of speech analytics.
For example, a conversation between “friends” might
go something like this:
Joey: How you doin'?
Rachel: Fine.
Phoebe: Fine.
Monica: Fine.
Chandler: Fine.
Ross: Fine.
What kind of conclusion can you draw from this conversation? Without emotion detection, not a very complete one. You might be tempted to conclude “all is right among these “friends”.
If however your speech analytics solution captures the speech, meta data and voice data of 100% of calls
and you are able to detect, capture, analyze, track and report on what is being said and how it is being said,
you have an infinitely richer way to postulate an assumption or a conclusion.
The conversation again, this time with an emotion detection layer:
The ability to detect emotion and to objectively analyze it provides critical insight into the context of a conversation.
Harnessing Emotion Detection for Customer Satisfaction
Before the advent of speech analytics, contact centers struggled with how to measure emotion objectively. They also struggled with how to use the information they collected. What did it mean? How could it be used
to help the organization?
According to CallMiner's Chief Technology Officer Jeff Gallino, “Emotion is inextricably linked to customer satisfaction. Unless you understand HOW a customer says something, WHAT they say provides you with only
a small portion of the customer's true intent for calling. Not having an emotion detection layer in your speech analytics solution, is like eating cake without the icing or soup without the crackers; it's okay, but certainly not great!”
Customer satisfaction is a critical metric for any enterprise. Keeping customers happy increases loyalty,
builds brand recognition and generates solid, dependable revenue.
And then there's the other side of the coin: customer dissatisfaction.
Today a disgruntled customer can bring an entire organization to its knees. A single customer can morph into an entire community overnight.
Measures of Emotion in Speech Analytics
As customer calls are being captured in the contact center, three basic buckets of data are collected: speech data (what is being said), meta data (where the call is emanating, which agent is handling the call and basic CRM or intelligence data) and voice data (percentage of silence and other acoustic data).
A sophisticated emotion model takes over from there. Breaking the key component of emotion – percentage of silence in the conversation - into two parts high and low – “CallMiner automatically measures silence, tempo and gain and integrates those results into a composite index of overall customer satisfaction,” explains Gallino.
For example, more silence on a call may mean the customer is less engaged.
A large block of silence may mean an agent has placed a customer on hold; there may be a process problem.
Agitation, an infinitely more complex indicator of emotion, that includes tempo, stress and gain, may mean the customer is highly dissatisfied.
Emotion indicators are specifically customized by client as an exercise to bring objectivity to a subjective measure.
The following charts illustrate how the percentage of silence, gain and tempo can help identify different emotions.
From the “friends” conversation, it is easy to see how one indicator – percentage of silence – can be integrated into a customer satisfaction score.
Rachel: “Fine!” (low silence, fast tempo, high gain = excited).
Monica: “F.I.N.E.” (low silence, slow tempo, high gain = demanding).
In this example, an increase in tempo totally changes the resulting categorization. Rachel might have been excited/upbeat because she just bought a great new pair of shoes. Monica might have been demanding/angry because she just broke up with yet another boyfriend.
For companies seeking comprehensive insight into customer satisfaction so they can make sound business decisions, additional layers of analysis is always better.
• Layer in gain and tempo and results are more revealing.
• Layer in words and phrases and results are more demonstrative.
• Layer in context and you will get full results and have a complete view of a conversation.
Contact Center Agents and Emotion Detection
Sometimes the customer is not the source of the problem.
Sometimes a call center agent is rude and a customer reacts by demanding “I want to speak to a supervisor.”
Typically the result of the conversation is a dissatisfied customer and the risk of churn is high.
Sometimes a call center agent will empathize with a customer by saying “I understand” or “let me help.”
Typically the result of the conversation is a satisfied or “saved” customer and loyalty is increased.
Emotion Detection in the Marketplace
Gallino provided insight for a large Midwestern hospital in a competitive healthcare market, “Our client was specifically looking to measure agent politeness. Not only was it a key ingredient in keeping patients happy (especially before surgical procedures), but it was also critical to the hospital's success in improving market share.”
In addition to customer satisfaction and agent performance (productivity, life cycle and compliance), hospitals and other savvy health care marketers are using speech analytics to get a complete picture of what their customers are saying and how they are saying it to:
• understand customers on an individual basis (customer centricity).
• understand why customers are defecting to competitors (customer churn).
• test new products and services (customer preferences).
• measure their brand equity (customer loyalty).
Yet another benefit of emotion detection is the ability to gain a better understanding of upsell and cross-sell opportunities. When CallMiner analyzed the emotional context of calls relating to upselling and cross-selling offers for one prospective large financial services client, it found that the offers were negatively affecting customers — even though they were increasing sales.
Gallino finds that “sometimes customers are so stressed by an upsell offer, that you are actually decreasing the customer satisfaction experience and taking away from the value of the additional revenue. Having insight into the emotional aspect of a conversation will allow a company to delve deeper into the data and discover how different customer demographics respond to different upsell and cross-sell opportunities, and then determine how to optimally offer additional products and services without upsetting its customers.”
Emotion Detection and its Impact on OPEX and ROI
WHAT! Is there an actual business case for using a speech analytics solution that integrates emotion detection? In French, the answer is “mais oui!” The English translation is “yes, of course!”
A large Northeastern telecommunications provider recently engaged CallMiner to help them with their costly operational expense of truck rolls.
Every time an agent was unable to resolve a problem over the telephone – for either telephone service or DSL – a service call was logged. By gaining insight into the emotional component of a customer interaction, the telecom was able to discover which customers were likely to be saved (because of problem resolution) and which customers were likely candidates for defection (due to dissatisfaction).
By listening to what their customers said and how they said it, the telecom provider was able to devise an agent training program to improve first call resolution. Most importantly they prioritized service calls to customers who were likely to remain customers and reduced truck rolls in the process.
Conclusions
Clearly emotion detection is vital to fully understanding a customer’s intent.
What is said is just as important as how it is said.
Gallino tells prospective clients, “If your speech analytics solution relies on word spotting or search, then you are going to miss critical data that will ultimately skew customer satisfaction scores. Emotion as expressed in voice data is absolutely essential to understanding customers.”
Article as appeared in Contact Professional Magazine by Michelle Craft, VP of Marketing of CallMiner
http://www.contactprofessional.com/issues/article.asp?ID=528