In 2003, Liberty Wireless, an inPhonic company based in Washington, D.C., recognized that it needed to upgrade to speech rec to provide customers with more convenient service, especially since many customers call from their mobile phones while they're involved in other tasks.

At the time, Liberty Wireless was preparing to go public and expecting to rapidly expand its customer base, so it needed to have a fully-functional speech rec-enabled IVR system up and running within a relatively short period of time.

"We didn't have the resources in-house that we felt were necessary to build the system ourselves," says Brett Peters, senior manager of operations at Liberty Wireless. "We also didn't have the time to spend on hiring in-house resources and going through the learning process to develop a professional speech application."

To increase its time-to-market, Liberty Wireless contracted with hosting provider BeVocal (Mountain View, CA) to develop and manage its speech system.

"One of the best aspects of our hosted solution is having a company whose core competency is speech and that can devote all of its resources to making sure the system works as best as possible," says Peters.

Liberty's speech app even received an "A" rating from VocaLabs, a third-party company that tests and measures the performance of customer service operations. The VocaLabs study measured the system for account balance, minute refill and bill payment. The applications each finished in the top tier for overall customer satisfaction, call completion and call consistency.

"The sheer number of customers who continually use our speech rec system to complete their tasks is the best indicator of its success," says Wolfand. "Our quality metrics within Liberty Wireless have increased 30% over the last year."

The Hosting Advantage

Hosted speech rec solutions are attractive for many reasons. As in the example of Liberty Wireless, reduced time to market can be a huge draw for companies that are interested in deploying a speech system.

"By hosting a voice application, the customer removes the time it takes to build, test and deploy the infrastructure," says CPT International's (Marietta, GA) David Cohen, director of marketing strategy, and Mark Rayburn, senior technical advisor. CPT offers its Voice Harbor service for VoiceXML, speech, multi-modal and traditional telephony applications.

But one of the biggest incentives for choosing a hosting provider continues to be cost. Buying and maintaining an on-site system requires a huge amount of capital, anywhere between $500,000 and $1 million, according to most experts. For many smaller companies, hosted solutions provide a cost-efficient way for them to get on the speech rec bandwagon.

Outsourcing speech is also a great way to deal with seasonal business. Hosting providers can add and take away ports based on demand so you don't end up paying for needless capacity.

But hosted solutions aren't limited to just small companies. Increasingly, larger businesses are looking to outsource because they lack the internal knowledge needed not only to develop speech applications but to constantly monitor and tune the application to ensure it runs to peak performance.

It's no secret that speech isn't plug-and-play technology. While the market's continuously working on improving tools to cut down on development time and the level of expertise needed to develop and manage speech applications, it still isn't for the faint of heart.

"Speech is more art than science," says Steve Tran, BeVocal's vice president of client solutions and marketing. "It requires a lot of artistic skill to design a speech interface that is useable, friendly, efficient, and ultimately helps callers complete their task."

This is where many outsourcers believe they can provide a significant advantage.

"While some ambitious companies have embarked on their own speech development, many have found that developing speech applications requires a different skill set than developing graphical applications," says Adeeb Shanaa, CEO of Voxify (Alameda, CA). "It generally takes more than a year for company to build a custom speech application. However, Voxify is deploying very successful systems in six to eight weeks."

According to Greg Pal, senior business analyst with TellMe (Mountain View, CA), the company's customers have "reevaluated where they really want to spend their time and energy. One of our clients felt that they were spending so much time and energy on just keeping their speech system upgraded that they weren't spending any time learning how to innovate on behalf of their callers."

Companies like TellMe, BeVocal and Angel.com also feel they offer their customers an additional benefit: "The fact that we're a network-based provider is a bonus," says Pal. "When you run on a network, you get network effects ¡ª every time we fine-tune the performance of our service, all of our clients benefit."

According to Michael Zirngibl, president and founder of Angel.com (McLean, VA), "The tremendous amount of information sharing and collective learning" of its network-based service is a huge advantage for customers.

Navigating the Hosting Market

But this is where the water tends to get muddy. When you use the term "hosting," it can represent many different things to different people.

"As hosting continues to permeate the speech market, the challenge for the industry and for customers is to clearly define what the difference is between the hosting models," says Bruce Pollock, director of information technology for West (Omaha, NB). "Without a clarification of terms, the lines around who's responsible for what can get blurry."

When you begin to evaluate hosting providers, it's essential that you understand what your options are and to clearly outline what you expect your outsourcer to be responsible for.

Here are four of the most common hosting scenarios available based on a January 2004 Yankee Group report authored by Art Schoeller:

Complete hosting. The hosting provider supplies the communication lines, IVR/speech servers, applications servers, and usually professional services to develop, tune, and maintain the application.

Voice front-end hosting. The IVR/speech server resides at the hosting vendor's site while the applications server stays on the call center's premises. In this scenario, the hosting vendor still provides applications development and tuning bundled into its pricing model.

Hosted infrastructure. The call center takes responsibility for developing and maintaining its voice applications and simply uses the hosting provider to manage the communications lines, speech/IVR servers, and possibly the applications server.

Managed services/facilities management. The IVR/speech servers, applications servers, and communications lines run at the call center's site, with the hosting provider supplying staffing to install, maintain, and tune the applications and hardware. In most cases, this means that the call center actually owns all of the speech assets.

VoiceXML is responsible for these myriad of options. Because of the transition from proprietary hardware ports and software license fees to open systems based on VoiceXML, outsourced service providers have much more flexibility.

VoiceXML also helps ease some of the issues that prevented call centers from outsourcing their voice self-service applications in the past, such as loss of control and security issues; with VoiceXML companies can retain their own databases on-site.

While many of West's customers opt for the complete hosting option, Pollock is seeing an increasing number of companies that want the outsourcer to host the speech engine and telephony infrastructure while the customer retains responsibility for the actual speech applications.

"Many of the companies that choose this platform hosting option are those that have traditionally developed and managed their own speech systems and applications in-house but when they're forced to make forklift upgrades to their engines and infrastructure, realize they can't afford to or don't have the desire to," says Pollock.

"Shared management is a trend that is changing the definition of hosting," says Steve Chirokas, Convergys' (Cincinnati, OH) senior director products and channels. "It's not simply a 'take everything off my hands' requirement. Rather, organizations can leverage existing programming talent and systems for increased control and investment protection while leaving the heavy lifting of answering calls and processing immense volumes of data to the speech hosting provider."

Pollock and Chirokas expect that this demand will continue as knowledge of speech rec spreads and as VoiceXML makes it easier for companies to develop and maintain their own apps.

What the Future Holds

In addition to increasing automation rates, many companies are looking to speech rec to improve customer service. Some are even attempting to use speech rec in cross-selling and up-selling scenarios. "Companies are under intense pressure to compete based on service," says Voxify's Shanaa. "[They] now demand that speech solutions improve customer service, as well as reduce costs. Companies have been on-the-fence, waiting and watching, for speech solutions that provide high quality and comparable customer service to their live agents."

TellMe sees more customers who are looking to manage their voice applications in the same way as they manage their Web sites. To address this, TellMe offers Experience Manager, which is a Web tool that lets customers make changes to their speech apps. "There is a growing requirement to re-use existing corporate assets for multi-channel integration," says Convergys' Chirokas. For example, the ability to deploy a speech-enabled FAQ application that takes advantage of existing data, typically used for the Web-based contact channel.

"The proliferation of VoIP (and the associated ease of integration) will jumpstart a number of opportunities and breathe some life into the voice hosting market," says Angel.com's Zirngibl.

Companies are also increasingly seeking a more seamless handoff from automation to live agents through the use of CTI.

"This becomes even more critical in a geographically dispersed contact center scenario that requires intelligent call routing," says Chirokas.

Challenges Facing the Hosted Market

One of the biggest challenges for hosted speech providers is making their economic model work for clients with mid-size call volumes. Many call centers are hesitant to even consider hosting because they don't believe that they have high enough call volumes to justify the cost.

"This is absolutely not the case," says Chirokas. "This perception has been created from analysis that only considers call volumes against equipment cost and amortization. In fact, there are many other factors that need to be considered such as upgrades for standards or performance, personnel staff and training, cost for capacity to deal with spike conditions, and more."

According to Voxify's Shanaa, "In contrast to 'early adopters' [of speech], the mainstream [mid-volume] corporation will start by only automating a subset of their calls. As a result, to meet this mainstream adoption, the hosted solution provider's cost model must work even for mid-volume clients. In other words, they must greatly reduce the traditionally high application costs, without sacrificing customer satisfaction." Shanaa predcits that the hosted providers that are able to meet this challenge will be the most successful.

Final Considerations

When it comes to cost, the most common pricing scenario is based on a per minute basis. Some outsourcers also offer per port pricing but this scenario seems to be falling out of flavor.

"Pricing per port helps set capacity usage expectations for both the hosting vendor and the call center but needs to be re-factored for growth," says CPT's Cohen and Rayburn. "Pricing per minute allows for more flexibility and can be useful if growth in call volume is anticipated or seasonal."

A less common pricing scenario is based on performance but this is something that should be carefully negotiated with your provider.

"Speech ¡ª above that of IVR ¡ª has a strong need for ongoing tuning, so per minute pricing is an incentive for providers to keep the speech apps running optimally in order to increase its usage," says Schoeller.

Once you've made the decision to outsource either all or some of your speech recognition, Schoeller advises that you carefully choose a hosted provider based on your call center's needs, not price.

"When it comes to differentiating one provider from another it comes down to the good-olds," he says. He advises that you understand the level of professional services available and check references; find out who the company's established clients are and whether the outsourcer was able to grow with their clients.

"While there are some differences between the outsourcers that have multiple data centers and reliability strategies in place, I'm hard pressed to say that any company's technology stands out over another's," says Schoeller. "The key in choosing a provider is finding the company that understands your speech needs the best. You hav to find the folks with the specific set of experiences in the specific vertical and the specific set of transactions that meet your requirements."