How U.S. Bank Works with Voice Biometrics

By Ann Brocklehurst |  October 7, 2015

The innovation team at U.S. Bank observed a particular problem growing larger as customers began doing more of their banking on smartphones: typing complex passwords was a pain. Often, if the customer had a strong password, she had to toggle between different keyboards several times to enter special characters, numbers, and upper and lower-case letters.

The bank’s mobile app already allows customers to use spoken commands to check their balance, available interest rates, and make payments. Could it allow them to speak a phrase that would replace their password, too?

A 25-Person Innovation Team

The 25-person innovation team at U.S. Bank works with different business lines within to imagine and develop solutions that tackle just that sort of present-day challenge, as well as exploring future markets and opportunities. The team, supervised by Chief Innovation Officer Dominic Venturo, looks to do “light-weight, fairly fast pilots,” in Venturo’s words.

Venturo says, “We pitch our own ideas to our own team to begin with … and because this is a team that has been together for quite a while and has quite a bit of experience, we can be very critical about would it ever work.”

A recent voice biometrics project being rolled out in the latter part of 2015 provides a window into how the team works. U.S. Bank had been using voice biometrics for some internal uses with “varying degrees of success, in terms of its overall performance,” according to Venturo. At the same time, his team was examining different biometric technologies. They were interested in both the security aspects — for example, adding voice verification on top of regular passwords — and usability, as consumers get more comfortable talking to Siri and interacting with their mobile devices by speaking.

Among the emerging technologies his team was examining was a voice identification system from Nuance, which “looked to be really good in terms of reliability and speed,” Venturo says. Nuance was ultimately chosen as the provider for the pilot, which launched in 2014. (See the video below to get a sense for how it works.)

Establishing Success Criteria for the Pilot

During the pilot phase, the innovation group’s role was to get the business and technology units of the bank together with the vendor to establish pilot success criteria; recruit pilot participants; and then actually get the requisite technology deployed. Once everyone sees the results, they can then make business decisions. Before moving forward with commercialization, “We ask, ‘Does the technology model make sense? Does it work from a customer experience [perspective]?'” says Venturo.

After approximately six months in the design and pilot phase, and nine months of preparing the offering for commercialization, voice biometrics is now being deployed as a partial password substitute for certain scenarios. “It won’t be available to every customer on Day One,” says Venturo, noting that it can be a step-up authentication tool when additional verification is needed. At some point, it could be used together with a fingerprint to replace a PIN, for example. As it rolls out over the fourth quarter, the bank will be looking to see how well it performs, and how customers react.

Managing the Risk of New Projects

Asked how typical the voice biometrics timeline was, Venturo replied that there is no such as an average timeline. Nor does the innovation team get exempted from bank rules to speed things along. Over the years, however, the innovation team, working in partnership with the bank’s risk management and compliance groups, has developed a system that looks at each case individually to appropriately manage the level of risk. “There is healthy balance and tension that occurs between being able to say, ‘Oh, we’re going to try to do this a little bit faster, a little bit different,’ and at the same time making sure that we don’t put the bank at risk,” says Venturo.

“The governance methodology around what is involved to do [early stage research] is very different than what it takes to bring a product fully to market commercially. Similarly, if I were going to do a pilot that were internal only [instead of recruiting] actual customers to participate… each one of those would have a different approach. They are different and we have an appropriate level of risk management around the way those projects happen.”

Venturo credits top management for its “very strong support for more innovative work.”

Letting Failure Point to a Better Strategy

Which is not to say everything succeeds. Asked to give an example of a failure, Venturo cited various experiences with contact-less payments beginning in about 2008. Sticker chips were put on bank and credit cards and, among other things, a fob was inserted into medical ID bracelets to explore how consumers might pay for things in the future.

With wearable technology like Fitbit, Jawbone, and Apple Watch making major inroads, that may not seem so dazzling today, but it was almost unheard of at the time and, according to Venturo, the technologies U.S. Bank tested proved unappealing for widespread consumer use.

“We need to be able to kill [ideas] early, so too much energy isn’t invested in them,” he says. “Three iterations was enough. We learned what we needed to learn. We really didn’t feel that that was the way forward relative to payments, but we learned a ton. So it’s a great example of how we iterate so that we can focus in on what we think is the best winning strategy.”