Best Practices: Achieving Success with Speech

When deploying speech services, what best practices do successful companies follow to engage and delight callers? What separates the good speech services from the great ones? Using real company examples, based on SpeechWorks Best Practices competition, this article will describe four deployments and look at how several companies followed best practices to maximize results from speech. Selecting the ‘right’ application One way to ensure success is to carefully determine the best speech application. Take into account your high-level corporate-wide initiatives, industry and competitive drivers, and specific business requirements. While the process may involve many considerations and stakeholders and take some time, it’s well worth the investment. Teleglobe, a leading provider of global communications, selected as its first speech application, an international call routing service. The process involved analyzing industry drivers, scanning service offerings, checking infrastructure and even soliciting customers on a cross-functional speech team. Faced with increased competition in Canada and from other calling methods, Teleglobe selected its international call routing service, Canada Direct, to be its first speech-activated application. Canada Direct was seeing traffic decreasing. Working with its long-distance provider-customers to evaluate what they could do to improve the service, they concluded that introducing speech recognition functionalities to this service was the best viable solution. Canada Direct carries more than 9 million calls per year, however, 30% of these calls were routed to other long-distance operators by default because many foreign countries do not support DTMF capabilities within their networks. The result: By speech-enabling Canada Direct, Teleglobe expects to automate an additional 600,000 calls within the year resulting in approximate cost savings of $600,000 USD. Building a speech service’s personality and voice brand Successful speech services focus on building an engaging ‘personality’ and, often times, a unique voice brand. An engaging, friendly personality, which is reflected in the prompts, the call flow, the voice and its tone, is designed to guide and encourage successful caller interactions through speech. It’s natural and easy. It yields high levels of caller satisfaction because your customers can accomplish their self-service transactions efficiently. Some speech service providers take the ‘personality’ a step further to create a unique and compelling voice brand, which may include an exclusive voice talent, a special name, sound effects and more. A voice brand could build customer loyalty and affiliation, and typically supports a corporate branding strategy. An example of a successful speech service personality and voice brand is “Julie,” the voice of Amtrak at 1.800.USA.RAIL. Amtrak followed a structured and iterative user interface design process starting with a thorough understanding of its callers’ needs. Departing from its traditional, formal approach with its customer dialogs, Amtrak opted for a more casual, conversational approach, one that would put callers at ease. Today, Julie greets all callers in a warm and friendly manner and provides regular reassurance as she navigates a caller through the speech service. The results are powerful. Since Amtrak’s speech-enabled 1.800.USA.RAIL launched in September 2001, automation rates have increased by 61%. In addition, not only has Julie received nationwide recognition on National Public Radio, she has also increased Amtrak’s speech bookings by 71%. Ensuring customer satisfaction Although this point seems obvious, everyone agrees that speech recognition’s predecessor, “touch-tone,” never ranked high on this scale. One best practice in speech is to pay close attention to the caller experience and the goal of achieving high levels of customer satisfaction. Though many companies believe that this process starts with the initial design of the system and continues through iterative testing and tuning at various stages of the project, other companies achieving success with speech consider customer satisfaction even sooner. Take the National Weather Service for example. The National Weather Service supplies local weather forecasts, watches, warnings and advisories directly to the public 24 hours a day, 7 days a week. It knew that it needed to update its NOAA (National Oceanic and Atmospheric Administration) Weather Radio. The text-to-speech (TTS) technology used to convert text forecasts and warnings into audio broadcasts was outdated and customer acceptance of the voice was low. In fact, customers never quite warmed up to the Weather Radio’s brittle voice. Shortly after being launched in 1995 it was nicknamed IGOR by the public. It was monotonic and hard to understand (especially during noisy, stormy weather conditions). To ensure customer satisfaction this time, NOAA considered its listeners’ experience from the very start. The NOAA speech team thoroughly evaluated the voice quality of several TTS technologies. The evaluation criteria was based on acceptability, intelligibility, credibility, clarity, likeability and pleasantness. No particular quality could be “unsatisfactory”. For example, a child’s voice might be outstanding in all factors, but not be acceptable or credible. Going a step further, as part of the National Weather Service evaluation process, NOAA posted samples of the leading voices on its Web page and requested comments. More than 15,000 comments were received through its Web-based survey. The selected male and female TTS voices received 95% and 80% of the public votes respectively, compared to 5% for IGOR. Marketing the Speech Service The right application, designed with the right personality, achieving high levels of caller satisfaction is a solid start to strong business gains. But those gains will only come from usage. An important best practice in speech is to implement marketing programs to raise awareness of speech and drive adoption. The most successful speech services are backed by efforts that address both internal ‘buy in’ as well as external awareness and adoption programs. While marketing to external callers is critical for many speech services, the University of Texas Southwestern Medical Center at Dallas (UT Southwestern) decided it would be better served by focusing its marketing efforts on internal buy in, awareness and adoption. The UT Southwestern speech recognition system is a ‘super’ auto attendant, nicknamed “Bill”. Bill answers both internal and external calls and directs them to more than 10,000 different University phone numbers. Bill also permits the caller to speak the name of the party and send a numeric page. When considering its launch strategy, UT Southwestern focused on the fact that its internal caller base was more likely to be frequent callers of the system and it was easier to reach through existing communication channels. From the start of the project, upper management, including the president, was informed and educated on speech. During deployment, numerous multi-media presentations that included samples of call flows were given in meetings across the campus. Articles were prepared and distributed to the entire campus population in newsletters and in the university paper. During the phased roll-out, callers were provided a voice mailbox and email address to report problems and make comments. In addition, numerous staff and students took part in testing, a practice that yields multiple benefits. In addition to obtaining real-user input, involving internal users in testing phases helps gain ‘buy in’ for the new technology and service, and can provide content for quotes and testimonials that can be used for marketing purposes at a later date. And finally, on Bill’s full deployment/launch date, a public notice email was sent to all faculty and staff. Internal calls to the system have grown more than 75% in less than a year, indicating the effectiveness of UT Southwestern’s internal marketing efforts. In addition, call volume to operators has been reduced by approximately 70%. Maximizing results for your organization The deployments profiled implement four best practices -- selecting the ‘right’ application, building a speech service personality and brand, ensuring customer satisfaction, and marketing the speech service. Although all are important in deploying a successful speech service, as you can see, there’s no one right way to execute on these practices to achieve success. It’s the thorough process, the careful analysis, and the almost fanatical attention to the callers (or listeners) that makes the difference.
Carol Lawrie is the Manager of Customer Dynamics at SpeechWorks International, Inc. She can be reached at carol.lawrie@speechworks.com or 617.428.4444. Editor’s Note: The companies profiled in this article participated in the SpeechWorks Here Best Practices contest.
SpeechTek Covers
for qualified subscribers
Subscribe Now Current Issue Past Issues