Zendesk Harnesses Big Data to Predict Unhappy Customers‘Satisfaction prediction’ feature analyzes customer service calls to determine if a customer is annoyed enough to form a brand-damaging opinion, company says
October 07, 2015
Wall Street Journal
by Shira Ovide
Zendesk Inc., which sells software that companies use to manage customer service call centers and websites, on Wednesday unveiled a feature that analyzes billions of data points to head off grumbling customers before they reach the breaking point.
Just as Google chooses which Web ads to display based on a user’s surfing habits, and Netflix recommends movies based on a subscriber’s Web-video history, Zendesk believes its software can predict which customers are annoyed enough to hurt a company’s reputation.
For example, when a mobile-phone subscriber calls the customer service center with a problem, the software will crunch the caller’s account information, compare words in the conversation to similar customer service interactions in the past, and measure the staff’s response time to assess the risk that the subscriber will quit the service or form a brand-damaging opinion of the carrier.
Armed with that information, the idea goes, a representative can give the caller special attention before it is too late.
It doesn’t take a supercomputer to figure out that a screaming caller is unhappy, but Zendesk Co-Founder and CEO Mikkel Svanesaid it isn’t always clear which customers are most at risk, especially when they’re posting on Twitter or Facebook. He said his company is trying to train software to help where human intuition can fail.
“In the volume of pure transactions, sometimes you can easily misunderstand each other,” Mr. Svane said. “These signals can be hard to identify.”