Home > Columns

Voicesense Combines Predictive Analytics with Voice Analytics to Foresee Customer Behavior


Presented By: Voicesense

Imagine how much easier it would be to prevent customer churn and maintain loyalty if your business could go into each interaction with real-time indicators showing when customers are dissatisfied. Or how much more efficiently sales could be conducted if outbound agents could get a go/no-go indication regarding each customer’s probability of purchase, enabling them to focus on interactions with high revenue-generating potential.  

Voicesense, a forward-thinking Israeli provider of voice-based predictive analytics solutions, now provides contact center operators with an automated framework for predicting the behaviors of customers during live interactions. The recently released iteration of their solution incorporates predictive analytics, enabling the technology to build an AI-driven personal profile for each individual and a predictive score for the customer’s potential behaviors. The solution creates this profile by analyzing more than 200 prosodic parameters of a person’s speech—the non-content characteristics, such as intonation, pace and emphasis.

The product combines the two worlds of interests of the company’s Founder and CEO, Yoav Degani—signal processing and clinical psychology. For more than 25 years, Degani worked in the defense industry on intelligence systems to gain his expertise in the former discipline. But he also studied and worked to become a clinical psychologist and at a certain point, he realized he could use speech signal processing to thematically measure personal tendencies, personality traits and interpersonal communications. He started the company nearly two decades ago with what he calls a “first-generation call center product” which measured excitement in the voice and gave companies a better idea of the call atmosphere. “We used it as a tool to determine customer dissatisfaction in real time,” recalls Degani. “While we believed we had a great product, it took a very long time for the market to mature to realize the need for such a solution. It took over a decade for speech analytics to capture a share of the market Even today, there are relatively little real-time speech analytics solutions in general use.”

While the company was initially able to raise funds just before dot.com stock bubble burst, enabling them to get off the ground and register patents, it quickly became nearly impossible to attract investors in that era. “We knew we had a good technology so for several years, we returned to just providing software services. By 2007, we were able to start raising money again and were able to complete the call center product, which we found we could sell. We also created a partnership with NICE,” he said. “At that point, we were focused on measuring the few last sentences of a conversation. But after extensive research, we noticed that people tend to have typical, repeating speech patterns. I then began to postulate that this has something to do with behavior patterns and personality.”

After additional intensive research, Voicesense was able to plan and perfect a second- generation product, one that Degani describes as being unique, not only for the call center field but for speech analytics as well. “Rather than classifying sentiment or interpreting current state of mind, it analyzes people’s typical speech patterns and links them to behavioral patterns. We were able to build a complete personality profile based only on people’s speech,” he said. “We focus on specific speech patterns that are related to specific consuming behaviors, such as the personality tendency of people to buy online or those who would be susceptible to future loan default or employees who might burn out.  These are related to tendencies to take risk in financial markets or the level of personal integrity, impulsiveness and so on.”

According to Degani, these unique speech patterns allow Voicesense to perform generic analysis that is language-independent, learning over the years to identify traits that have to do with the speech production mechanism as opposed to speech content or culture. The company now has experience in numerous Western languages besides English—not just French and German, but Polish, Czech and Hungarian. What Degani considers more impressive is the ability to interpret behavioral traits from Asian languages such as Chinese, Japanese and Korean, which he describes as tonal languages where speakers use intonation to express meaning. “No matter what the language being spoken, we’re able to measure more than 200 parameters per second. Within every sentence, we have thousands of measurements which enables us to compensate for the intonation which might be used on one word within the sentence.” As they widened the scope of the company was also able to incorporate machine learning and AI into the solution.

Voicesense’s current focus goes beyond contact centers. It is being used in healthcare with a mobile app to track such conditions as depression, schizophrenia and ADHD in patients, all with very successful clinical trials, according to Degani. It is being put to work in the enterprise for big data and marketing use as well as for loan application review and business development. Voicesense is also proving to be a valuable tool in HR applications, providing profiles of candidates for screening and recruitment.

Degani maintains that unlike other customer experience monitoring tools, companies will be able to achieve tangible bottom-line results from Voicesense.  “We’re analyzing activity in real time and are able to provide accurate predictions of whether the customer will buy or who is dissatisfied and at risk of churning. We provide businesses with clear-cut conclusions and we are willing to be compensated on that basis. All the data we generate can be stored in the CRM and thus the customer will not need to be analyzed on subsequent calls. The Voicesense application can be fully integrated on premise or through an API to the cloud with an organization’s other systems used in its call center operations. In addition to CRM, it also can integrate business intelligence and other systems.

 “Until now, the speech analytics technologies used in call centers environments revolved around emotion detection and had limited applications to support sales activities, while most predictive analytics approaches were not applicable to call center operations as they typically rely on historical data and offline analysis,” explained Degani. “We have strong expectations that the initial PoC trials that are currently underway at a number of call centers in the telco and financial service sectors will lead to impressive results and wider demand.”