Forget about social security numbers, addresses, passwords or maiden names, call center employees at Virginia Credit Union are now authenticating members via a voice-based security system in as little as six seconds.
“I needed a quicker way to verify our members because they hated our procedure – it was laborious requiring a lot of transactional information,” said Kate Hopson, AVP and contact center department manager at the Richmond, Va.-based credit union. “It just took a long time.”
In the hopes of solving that problem, the credit union signed on with Burlington, Mass.-based Nuance Communications’ Voice ID. Prior to signing, it took an average of 92 seconds for call center employees to authenticate members, but since rolling out Voice ID, overall call times have reduced by 24 percent.
“You don’t have to know what’s going on with your account, you just need the sound of your voice,” said Hopson. “In call center environment, shaving off a minute and a half is huge.”
And it’s having a positive impact on more than just the membership, she added. “The staff likes it because they can help somebody opposed to asking all these [authentication] questions.”
Two-fold benefit
While fraud deterrence is a leading selling point for biometric voice authentication, Nuance’s Director of Product Strategy, Brett Beranek, said big banks and credit union clients are approaching this service for diverse reasons.
“It’s a very different conversation we are having with credit unions than we are having with the top five financial institutions — they are getting hammered when it comes to fraud,” said Beranek. He added that the balance between security and the customer experience at these large financial institutions is “heavily tilted” toward security.
“The conversations we are having with credit unions are much more focused on improving the member experience,” said Beranek who said the firm caters to a “handful” of CU clients. “I’m not saying that fraud is nonexistent because it is there, but we do not [yet] see the [same] systemic professional fraud attacks [against credit unions].”
Since rolling out the platform to all membership in October of 2016, VACU has flagged more than 6,200 calls. On average, the CU answers 60,000 call center calls from members each month. Hopson said the “vast majority” of these calls were not fraud-based. The balance of stopped calls was due to “voice prints” not matching, explained Hopson, such as a member calling behalf of a family member.
Voice control
The $3.8 billion-asset credit union supports 266,000 members, 720 employees and 18 branches. Hopson explained that once the software was installed in March 2016, her team began “blind beta testing” segments of membership. As always, all calls were recorded and voice prints were made to ensure that the member’s voice could be authenticated the next time they called in. These recorded calls were later deleted.
During the blind testing, Hopson said, there was only one failed authentication, and a subsequent beta test of 100 members resulted in a 100 percent success rate. Beranek cautioned that no technology is 100 percent “fool proof,” and said Nuance’s Voice ID has between a 98 and 99 percent success rate.
Beranek said Nuance has deployed hundreds of systems globally since rolling out its first voice biometric solution in 2001, and Nuance’s partnership with VACU even resulted in a
“By using biometrics, security is significantly more consistent and accurate than asking security questions or passwords,” said Beranek. “From time to time it will fail for all sorts of reasons. For example, if someone has a disease affecting his voice or laryngitis, or if you were in an extremely loud environment like a rock concert.”
To date, more than 23,777 VACU members have enrolled. Hopson said the process takes less than five minutes, and members can enroll online or with the help of a call center specialist. On average, 1,700 members enroll each month.
In order to gather enough “net audio,” members are required to have their voice recorded for roughly 40 seconds, which equates to the time it takes a member to provide his or her name and mailing address. The Nuance software can then detect the 140 unique characteristics of a member’s voice. The software can also detect whether a call is recorded known as “anti-spoofing” or live.
“With people posting videos on YouTube and other social media sites, we have anti-spoofing algorithms in place to detect recordings and synthetic voices,” said Beranek.
“There is a lot of anxiety around these attacks [and] we haven’t seen synthetic voice attacks, but have seen a handful of recording attacks — primarily these were pranksters trying to see if they would be detected,” he continued. “The fraudsters — the actual bad guys — once they see protection with biometrics, they tend to move on to another financial institution instead of trying to beat it [Voice ID].”
