Creating a Company Voice

If Yahoo Inc. had a voice, it would sound like Jenni McDermott.

The perky dog lover who dreams of becoming a professional painter reads news headlines, weather reports and e-mail messages to Yahoo customers who access the popular Internet service by phone each day.

But Jenni’s true role is much bigger than that. Her job is to personify the quirky dot-com with her voice--and every detail about it has been meticulously designed with Yahoo’s brand image in mind.

As automated phone systems become ubiquitous, companies are paying close attention to the voices that represent them to callers. Generic voices still abound in corporate America’s phone systems. But some firms spend tens of thousands of dollars designing custom voices to solve the difficult problem of conveying a brand image through sound alone.

For instance, the team behind Yahoo’s fictitious Jenni wrote a five-page biography that included her career as a high school soccer player and her stint playing guitar in an all-girl band. Their aim was to create a voice that sounded innovative and fun to hang out with, with the hope that callers would associate those qualities with Yahoo.

“I understand where she comes from,” said Deborah Ben-Eliezer, the voice actor whose words are mixed and matched to create the McDermott who “talks” to callers. “It affects how I stand in the recording booth and the dialect I use.”

The driving force behind this transition to brand-specific voices is the increased accuracy of speech-recognition systems. Computers can understand spoken words, even if the speaker has a thick accent or a cold, or is talking on a cell phone while driving on the freeway with the windows rolled down.

But when people speak into their phones, they expect to hear a person talking back--not some harsh synthesized voice. That’s what prompted Yahoo to craft a specialized voice for its Yahoo by Phone service.

“It’s like a conversation, and in a conversation the persona is much more amplified,” said Madhu Yarlagadda, general manager of Yahoo Everywhere, the mobile access initiative that includes Yahoo by Phone. “The moment you start talking to somebody, you get a picture in your head. You start putting a personality to the person.”

And that personality will be associated with the company--at least in the customer’s mind.

“When a customer calls a customer service agent, whatever that person is like brands the company,” said Lynda Kate Smith, chief marketing officer for Nuance Communications Inc., the Menlo Park company that created Jenni McDermott’s voice. “If you get a British accent versus someone who woke up with a headache and has to get home to fix dinner for the kids, you get a totally different perception of the company.”

It’s a far cry from the 1939 World’s Fair in New York, where AT&T; Corp. demonstrated the first machine that used computers to blend prerecorded sounds into words. The Voder, as it was called, was more of a novelty item than a useful device.

“Operators had to be trained for months on how to spell phonetically and hit keys on the device that prompted it to speak,” said Michael Dickman, a spokesman for AT&T; Labs.

Researchers kept at it. Half a century later, the technology was good enough to let brokerage clients dial up the latest stock quotes over the phone. All they had to do was spell out a ticker symbol using the phone’s touch-tone keypad, and a computer would read the price for the desired stock. The voice sounded far from human, but brokers and their clients came to rely on the technology because it saved time and money.

Now, all sorts of tasks that used to be performed by armies of customer service agents are handled by automated phone systems, said Dan Hawkins, managing analyst at New York market research firm Datamonitor.

“About two-thirds of call center costs are taken up by agent salaries and training,” Hawkins said. “Anything you can automate can save companies money.”

Companies are expected to spend $700 million this year on automated phone systems, according to Kelsey Group Inc., a market research firm based in Princeton, N.J. About $50 million of that is earmarked for the text-to-speech applications that put voices--such as Yahoo’s Jenni McDermott--on the line.

Soon, automated systems will do more than merely replace humans, said Mark Plakias, senior vice president at Kelsey Group.

“We’re looking at new applications that it just doesn’t make sense for a live agent to do,” he said. “I’m not going to have an operator call all the people on Flight 85 to say that the flight’s been delayed. But with automation, I can do that.”

For now, the cutting edge is occupied by characters such as Jenni McDermott. Yahoo customers will never hear Jenni discuss her dogs, her artistic hobbies or any other facet of her elaborate personality. But those details are critical in allowing Ben-Eliezer to strike the exact tone that embodies Yahoo’s brand.

To get to that point, Yarlagadda’s team gave Nuance a list of characteristics--including efficiency and youthfulness--they wanted to hear in the Yahoo by Phone voice. Nuance used the list to create more than 10 sample personalities. Yahoo tested them with its users, who described Jenni as “innovative” and “fun to hang out with.” That’s when the company knew it had its voice, Yarlagadda said.

Then Ben-Eliezer went to work, recording thousands of phrases such as “Got it,” “What can I get you?” and “Is that right?” For Yahoo’s weather function, she recited the names of 1,500 cities. When she records, she adopts Jenni’s imagined mannerisms, holding one hand on her hip, shrugging her shoulders up and down and bouncing her head slightly as she talks.

Jenni made her debut in February. The result is a system that sounds more lifelike--and more Yahoo--than even Yarlagadda expected. Jenni’s personal touch came through on a day when his e-mail inbox happened to be particularly full.

“I got, ‘Oh, you’re very popular,’ ” Yarlagadda said. “And I thought, ‘Oh, am I popular?’ These are the little things that make a lot of difference.”

It may seem like a lot of trouble to go to for a disembodied voice. But people easily pick up on voices that sound false or insincere, said Stanford University professor Clifford Nass, co-director of the school’s Social Responses to Communications Technologies Project.

In one of Nass’ studies, a computer read news stories in both happy and sad voices. Listeners found “a total loss of credibility” when the happy voice read a sad story and vice versa, he said.

“The human brain is built for speech,” Nass said. “We’re built to detect personality and emotion.”

All kinds of subtle clues help listeners discern the personality behind a voice, and none of them should be taken for granted, said Steve Chambers, chief marketing officer for Boston-based SpeechWorks, which also designs custom voices for automated systems.

Among the details he considers are “the phrasing of the questions, the intonations, the speed of the questions, the navigation--how you get from one place to another--and the grace with which the system recovers if it makes a mistake,” he said.

When Wells Fargo executives decided to revamp the automated phone system for the bank’s credit card customers, they thought they could evoke their Western-themed brand by adding sound effects such as horse hoofs and stagecoach wheels. But unimpressed customers said they would rather talk to a banker than to a cowboy, said Tom LaCentra, director of customer service for Wells Fargo Card Services in Portland, Ore.

The company took that advice to heart and designed a new voice based on a Nuance character named Reed Johnston. Originally conceived as a surfer dude with a literary side, Nuance aged Reed slightly, took him out of the surf culture and put him on a golf course so that he was more in keeping with the bank’s corporate image.

“The voice is trustworthy,” LaCentra said.

All that attention to detail makes the voices--and the companies they represent--seem human.

“People are intelligent,” said Ben-Eliezer, the voice of Jenni McDermott. “They know ultimately they’re talking to something automated. But I want them to be comfortable enough that they’ll vocalize rather than push buttons.”