When you call some firm’s customer service number, do you really care whom you talk to? I mean, beyond basic competence in addressing your request or walking through how to solve a problem, do you really care? Would it matter if there were some way of matching you with a better agent? “Better” here is not about skill level, per se, but rather about someone who matches your personality type. That is, you and I may call with the exact same issue but have different agents recommend as best for each of us based on just how we behave on the phone.
This is the kind of service offered by a Chicago-based firm called Mattersight, which has been featured in recent articles in both Crain’s Chicago Business (Why you might not hate calling customer service next time, Feb 12) and InformationWeek (Big Data: Matching Personalities In The Call Center, Feb 17). Here is how Crain’s describes what they do:
Your call is automatically routed to a like-minded agent who’s been matched to you according to factors such as communication style and personality type. It sounds a little like science fiction, but it works. Clients such as pharmacist CVS Health and online insurers Progressive and Esurance (an Allstate subsidiary) say Mattersight’s software speeds up calls, boosts sales or raises customer satisfaction by 10 percent or more. …
Mattersight’s product, based on more than 10 million algorithms developed by an in-house team of behavioral scientists, is overseen by David Gustafson, Mattersight’s product chief and executive vice president. The algorithms are if-then statements that analyze callers according to speech patterns and cadence in order to gauge their personality type and mood and route them toward a simpatico customer-service rep.
People’s speech patterns constitute “an emotional syntax,” says Gustafson, 37, one that can quickly demonstrate whether a caller “is someone who values order and logic, or if they’re fun, spontaneous and creative.” The best customer service reps are adept at working with all personality groups but still do better with one type or another; Mattersight’s tech aims to play to that strong suit as often as possible.
This is obviously an intriguing possibility, but there are some obvious challenges. First, call centers work best when capacity can be treated as a generic pool — that is, every agent is equally good. Now that obvious doesn’t hold in reality. Some are going to be better at some tasks while the weakest are just mediocre at everything. But service — in the sense of the shortest wait — will be best when any available agent can take any call.
Mattersight’s algorithms can muck with that routing. Thus, they are creating a tradeoff between good service (matching with the ideal agent) and a short wait. This can obviously be done smartly. If an agent that would be a good match is four minutes into a call at a call center where 90% of calls finish in under four minutes and five seconds, leaving the caller on hold for a second or two seems worth trying. On the other hand, if 90% of calls go for more than five minutes, it may not be worth waiting for that ideal match.
The second problem here is how you actually build a profile of customers. I have been a customer of the local gas utility for over a decade but I don’t remember the last time a called them. I would argue that this is not unusual. So many basic transactions can now be done on-line (e.g., balancing inquiries or account updates), that there are very few reasons to call. Some services like CVS’s pharmacy benefit management business may have more frequent interactions with a large slice of customers than other firms, but even then interactions may be only, say, quarterly. That is a limited amount of data on which to build a profile.
So how does Mattersight do this? Here is what InfoWeek says.
The company has a second product offering, where it comes in and looks at its client’s historical call center data.
“We’ll cross-reference it with our database of personalities and phone numbers,” Wesbecher said. The goal is to pair two personalities — caller and rep — a match designed to generate shorter, more effective support calls.
“In this case, we don’t actually present any information to the call center employee,” said Wesbecher. “We just create more magical connections by pairing, right when the caller calls in, that [customer] with the service rep.”
So linking all of my customer service calls could provide a much richer picture of my personality and that would give the gas company a running start at routing my call. But let’s face it: this is really creepy. I realize that in many ways this is no different from what Google or other Internet firms do in tracking my movement across the web, but those firms ostensibly have public policies that govern how they use the data they collect. I don’t recall a firm ever telling me that a third party may use my phone call to build a profile of me.