VocaLabs CEO Peter Leppik can cite two facts that speak to the success of his company. Just about every one of its customers is on a subscription plan. And during each of the past three years, since the first year the Eden Prairie, Minn.-based firm generated revenue, the total amount of revenue has doubled or more.

"We have a significant number of clients that are the developers of automated call center systems and we have pretty much all of the major speech recognition companies as our clients as well as many end-user call centers," said Leppik, whose firm has been in existence since 2001. VocaLabs has found success by providing feedback to call center clients through the use of surveys regarding how their operations perform from the caller's perspective.

"We started VocaLabs with the idea that we wanted to … create a way for companies to gather statistically meaningful data sets that had very rich data in them that would allow them to … gather the information that they needed to understand not just how well they're doing, but also what the underlying causes would be," Leppik said. "We want to take the idea of statistically valid, scientifically meaningful, methodologically sound data collection, which gives the customer specific action items around specific problems, to as broad an audience as possible. "Customer service is the most important marketing tool that companies have in many cases. There's going to be a greater realization of that fact." He added that many organizations view customer service as a cost center that needs to be minimized as opposed to a significant component of the marketing effort.

"Whether a customer decides to continue doing business with an organization or not is determined primarily by how they're treated when they go out and do some sort of interaction with the company," he added. Leppik emphasized the importance of his firm's ability to collect survey data as close as possible to the time of a call. "In a typical study that we do right now, the survey is normally done within minutes of the end of the call so people don't have time to forget what happened in the call before they go on to provide their feedback," he said. "As a result, we get data that's got much more to do with what actually happened as opposed to trying to do the feedback [in] two or three days as is common with a follow-up survey."

The company has access to a pool of 90,000 consumers for gathering data. "The end-of-call surveys that some people are doing … there's a very clear methodological problem with that, which is it excludes from the sample pool anybody who got sufficiently frustrated or annoyed that they hung up before the call was over," he said. "I've found that a lot of people … just don't know what can be done better. With a follow up survey … it's very clear that … calling people back 72 hours after they call the call center is not a very good way to gather data because, in many cases, the people that you're calling have forgotten what's happened on the call or, in some cases, even forgotten that they called at all and the quality of the data that you're gathering is very poor." Leppik warned against being guided by "what a bunch of gurus" advise. He added that simply following a top 10 best practices checklist is not going to do the trick and that listening to a consultant "is not going to give you the information that you need."

"You need to actually go out and get your hands dirty and you need to find out what actual customers do when they interact with your system," he said. "If you do that, in pretty much every case that we've looked at, you'll discover a wealth of things that can be done to improve the quality of that system and, in most cases, improve the financial performance of it as well. "At the end of the day, it's about the quality of the data. If you can trust the data you're getting you can make actionable decisions on it. It's nice to know in a call center that you're satisfaction rate is up a tenth of a point in the last six months, but what you really want to know is why the satisfaction rate is up a tenth of a point in the last six months." Leppik also addressed the topic of speech recognition. "The things that make me cringe the most are the ones where there's a company that has a group of five people that's been developing their own IVRs internally for 10 years -- now they're starting a speech project -- because I know it's going to be bad," he said.

Leppik also cited a strategic error forced on the sector by Wall Street: presenting speech recognition as something "somehow more world-changing" than it is. "Speech recognition is a much better version of IVR, but it's not going to change the universe," he said. "I think we saw the voice portal companies come to that realization pretty quickly.

"There should not be a [new] self-service system going in today … that is not speech-based because the technology is much better." But he emphasized the importance of paying attention to "the human factors" and how callers actually utilize systems.

"It's not hard and it's not expensive to provide good quality customer service, be it an automated system or agent-based or some mixture of the two," he said. "It's not hard to provide a really great experience for the customer. It's not expensive to provide a really great experience, and, in many cases, it's actually less expensive because you can convince people to use automation where they would have refused to do so before."