IS ACCURACY STILL AN ISSUE?

In short, no. The speech engines -- the software that provides the vocabulary and grammar for speech recognition systems -- are very powerful, and most people in the industry acknowledge that accuracy hasn't been an issue for almost a decade. Speech engines today recognize more than 200,000 words, matching the size of the typical desk dictionary. That includes slang and regional variations and accents.

However, as Peter Leppik, cofounder of the customer service survey firm VocaLabs, notes, "There are still applications being built where because of bad design, accuracy becomes an issue. We still see bad implementations of good technology."

Design is key -- so key, that it can make up for a weak engine. Leppik observes, "It's possible to build an application that functions very well from the user perspective, even when the speech recognition accuracy is relatively poor. It's a question of paying attention to what the users want to do, and good design, and good error recovery. In a lot of cases, when you've got good error recovery, the callers won't even perceive that there was an error. With a bad design, accuracy can make or break an application, and with good design, you can have relatively poor accuracy and the system still works."

Good design may mean unambiguous structure. Many vendors champion the "natural language" style of speech recognition, but not Angel.com's Mike Zirngibl: "I think that's a little bit of a fad. I think it's the wrong way to go. It's not really natural language. You spend nine months inserting 20, 30, 40,000 utterances and combinations of things that people can say into the system, and then you start running it. I believe the way to go is to avoid having to do that. Use data, use customer profiles, data-driven personalized IVR systems that are concise. The future of IVR is when the caller has to say as little as possible."

Maybe, maybe not, but as Steve Pollock of TuVox points out, "The most important measure of an application typically is how successful the call completion rate is through the application as opposed to the specific speech accuracy, because sometimes you need to do things a couple of different ways to get the answer from the caller."

So while accuracy isn't necessarily an issue, design sure is. The advantage of hosting is that experts who work on a variety of types of speech applications full-time can manage and fine-tune your application indefinitely, leaving you to concentrate on your agents.