VERA algorithm measures customer satisfaction

Anger, sadness, impatience, joy, anxiety – these emotions may all be heard in the human voice. Pension management organizations APG and PGGM have argued that it’s helpful to be able to recognize these emotions in their telephone contact with clients. They contacted Brightlands Smart Services Campus in Heerlen to find out if these emotions could be translated in an algorithm. The response was yes, and the first pilots are already running in various call centers.

Companies and organizations are very interested in getting the opinion of (potential) customers and business associates who have been in contact with their call centers by telephone. However, people’s willingness to leave reviews online or elsewhere is extremely low. The response rate to requests to fill in an email form is barely five percent, and people’s desire to participate in telephone surveys conducted immediately after a call is even lower. “Except when people are very dissatisfied or very happy,” Pieter Custers, director of the Techruption department at Brightlands Smart Services Campus in Heerlen, confirms. “This isn’t very helpful, even though we live in a time where reviews are very important. As an organization, you want to know what customers think of you, what can be improved or done differently. You certainly don’t want to provoke any anger or irritation. It’s also very nice to know when a caller ends the call with a satisfied feeling. This is the data that’s important in the formulation of business strategies. We here at this campus are well aware of this; data is, after all, our core business.”


PGGM and APG’s research question to find out whether emotions may be captured in an algorithm was tackled energetically. A project group was formed in the Techruption innovation program including researchers from BISS and CAROU, knowledge institutes affiliated with Maastricht University and Open University respectively, and the clients’ own data specialists. “This is how we work,” Pieter Custers explains. “We bring the experts together at this campus’ ecosystem and get to work as fast as we can.

Our mission is to use data to innovate, initially for and with our partners, and in this case, with these four parties. This might sound rather vague to many people, but we are happy to report an increasing number of concrete applications that have come out of this approach. The recognition of emotion in the human voice is one of them.”

Together with people from APG and PGGM, the researchers from the knowledge institutes CAROU and BISS analyzed thousands of recorded telephone calls. The experts ultimately filtered eight emotions out that were incorporated into an algorithm called Voice Emotion Recognition Assistant, or VERA for short. “This algorithm recognizes emotions,” Pieter Custers explains. “The call center agent can immediately see the caller’s state of mind on their screen and respond accordingly. This can really improve the quality of conversations.”

Customer satisfaction

René Rateischak, head of the Groeifabriek (“growth factory”) at pension management organization APG in Heerlen and closely involved in the development of VERA, nods. “We haven’t reached this conclusion yet, but we can measure customer satisfaction much better now with VERA. Each contact event is evaluated by VERA based on the emotions measured. This means a 100% response rate, which is fantastic. This is important for us but particularly for the pension funds since these are our clients. They want satisfied and well-informed customers after all. Based on customer satisfaction, we can implement improvement programs, evaluate conversations and see what went well and what didn’t. This can serve as the foundation for training our call center agents. This can help call centers work much more efficiently.”


There is another not insignificant benefit, René Rateischak reports. “We can see how the call center agent deals with the caller’s emotions. If there’s too much stress during the call, it might be time for a break or a meeting about the call afterward. Van Lanschot Bank will start a pilot soon in which employees’ blood pressure will be measured via a tracker, on a voluntary basis of course. It’s important that employees don’t feel pressured or rushed. By measuring stress and blood pressure, this can be prevented.”
In the meantime, several Heerlen-based parties have gotten in touch about the VERA algorithm. This number will undoubtedly increase once the research, documented by two of the researchers involved who are currently affiliated with the Open University, is published in the Journal of Service Management. One of the OU researchers is also receiving a grant from Netspar, a knowledge institute specialized in pensions.


“This project is a great example of how we want things to go here at Techruption,” according to Pieter Custers. “Organizations working closely with knowledge institutes. All of the parties have benefited from this cooperation. The project has led to a new smart solution for the companies participating, a scientific publication for the knowledge institutes, and the start-up Contexta360 is involved in further implementation. VERA is also initiating new research, including a project where we’re studying how voice recognition software can be used by the police.”