Building a Better Voice Response System
Recently, my wife and I went car shopping in Fremont, CA. The first place we went to was the Volkswagen dealership, where the salesman gave us a long pitch about the headlights--how they made light in front of the car, how you could turn them on and off, how they were replaceable in case one burned out. I don't know what he was after, but we had a good laugh as we left.
Oddly, we got the same thing at the Ford dealership. "Our headlights are already compliant with upcoming DOT 1728.B standard for headlights in automobiles," the salesperson told us. "Plus, we've added some proprietary technology so they turn on faster than other headlights. Ford just made the speed of light faster." He chuckled as he said this, clearly in love with the new tag line and technology. For some reason that he couldn't understand, my wife wasn't using headlights as a buying criterion.
So then we went to the Honda dealership. Surreal but true: The saleswoman, named Nancy, started off with a story on headlights. "With headlights, you get a greater return on investment with your vehicle, because you can drive at night. This represents an increased value of up to 2,000 percent for selected driving applications."
"Where, in Nome?" I asked her.
"Actually, our study was done in Stockholm in January. Would you like a white paper on it?" She offered me a blue glossy document with a picture of a smiling blond and a car with it headlights on.
As I flicked the headlights on and off, timing them with my watch, my wife started looking under the hood of the car and reading the engine specs. Over her shoulder, she said, "You know, Nancy, you can't buy a car that doesn't have headlights."
Nancy, knowing an educate-the-customer opportunity when she saw one, posed a rhetorical question. "Yes, but are they DOT 1728.B compliant headlights?"
My wife cut her off. "Yes, Nancy. They are." I looked up to see Nancy, slack jawed, trying to figure out what she could say next, and it dawned on me that this was exactly what is happening with VoiceXML.
You hear it everywhere: "VoiceXML is better than traditional IVR." "Use VoiceXML instead of dinosaur language." "VoiceXML makes it faster and cheaper to deploy applications." Faster and cheaper than what? Who doesn't support VoiceXML? I've looked and looked and I can't find a single voice response platform that's commercially available in North America that doesn't support VoiceXML. And as vendors in this space continue to promote VoiceXML as a key differentiator and the most important breakthrough in application development, you can't blame the customers if they start to ignore the noise and look under the hood on their own.
That said, what should buyers do beyond just kicking the metaphorical tires? Get back to basics.
Whether you're buying a $30,000 car or a $500,000 voice processing solution, how much you're going to spend is usually one of the biggest criteria. But unlike cars, with voice applications the total cost is a very hard number to get to. Costs are both hidden and unpredictable. How much will the application and the hardware really cost? How expensive will it be to keep it in-house and have the IT staff maintain it? How much will it cost to train them?
Everyone who has ever undertaken a large IT project is well aware of both the schedule slippage and the cost overruns that seem unavoidable. VoiceXML may help with the application, but it doesn't do a lot for getting the T1 lines, racks, and servers installed. And if you want redundancy, then you'll pay twice as much. It's like buying two cars so that you'll always have one that works.
But here's where the car metaphor breaks down. VoiceXML can help drive costs down, way down, for two reasons. First, VoiceXML separates the application logic from the platform. Before VoiceXML, applications were written specifically for the platform they'd run on. Because of the severe fragmentation in the platform market, this meant that packaged applications had a very limited addressable market, and never really took off. With VoiceXML, the market for any given application is huge because everyone can buy and run it. For customers, this means they can buy an off-the-shelf application and deploy it right away, rather than buy a custom application that will take months to develop.
The second way VoiceXML drives down costs is by separating the application from the platform. This means that companies can own their applications, but outsource the running and management of them. Hosting companies provide high-availability redundant platforms that are sold on a pay-per-use basis. The enterprise is freed from the platform and telephony management, but gives up nothing in terms of flexibility and control. As for costs, it's like paying for your car only when you want to drive it, rather than paying for it to sit in the garage.
It's hard to find a good mechanic, and it's hard to keep a trained IT maintenance person on staff. Even if you spend a few thousand dollars sending one of your people to a course on how to keep a VoiceXML platform up and running, that only means they'll be there to get it back up if it crashes while they're on call. Which is bad if they ever go on vacation. When they get promoted, change roles, or leave, you're back to having someone spending a week and a few thousand dollars on training. Figure a couple people a year may need to be trained just to make sure someone on site knows how to manage the platform at all times.
But it's not just the mechanic. There is also the issue of spare parts. My mechanic always seems to have to send out for the brake pads or belts that my car needs. Where are the spare parts for your platform going to be kept? Some platform vendors talk about having regional parts distribution warehouses as if having your application off-line for a day or two isn't a complete and total disaster.
The issue of uptime is a serious one, and it isn't easy for any company to keep a platform up and running 99.9 percent of the time. It requires expensive maintenance, sharp staff, and lots of spare parts.
Again, this is where the VoiceXML/car analogy breaks down. Hosting companies can manage the platform in multiple NOCs, and be multiply redundant within each one. And they can spread the maintenance costs over a much larger system. It's no surprise that even tech-heavy companies like e-Bay and Yahoo! outsource the management of their Web sites. It's just cheaper and more effective. The same reasons apply to voice, and with VoiceXML, it's a viable model
The right size car
It's hard to know how big a car to buy. For driving to work, it's just me, and I've paid for three extra seats I'm not using. But on the weekend, the car is too small to take the kids and their friends to the zoo.
Likewise, with voice applications you can't win on provisioning. Too much capacity is a waste, and too little capacity results in blocked calls, lost business, and customers who feel like my six year old who can't go to the zoo to see the gorillas. And again, buying a VoiceXML platform doesn't help you here. As soon as you make the decision to own the car, you're stuck with that number of seats. As soon as you make the decision to own the platform, you're stuck with that many ports.
If your call volumes are growing, shrinking, variable, or unpredictable, outsourcing is just about the only way to always have the number of seats you want. Hosting providers can provide as many or as few ports as you need, and work with you on the basis of dedicated capacity, or they can provide flexible capacity on a per-call, per-minute, or even per-transaction basis. Either way, you only pay for what you need, everyone gets to go to the zoo.
Oh, yeah, and the headlights
You'd have to be pretty unusual to want a car without headlights. Even if you aren't planning on driving in Stockholm next January, the flexibility is worth having. Besides, every car comes with them.
Likewise, it's unlikely that anyone making a major investment in new voice applications will go without VoiceXML. VoiceXML has moved from being a platform decision to a core part of a strategic direction. But it's only the right direction if you actually take advantage of the things that VoiceXML has to offer:
Packaged applications: They're less expensive to buy and have been tested and refined before you put them in front of your customers. They help lower deployment time to weeks rather than the months that a custom application requires.
Hosted Solutions: Hosting provides flexible capacity and high availability without the cost and hassles associated with installing and running their own platform. Flexible capacity is critical for lowering costs while preventing blocked calls and lost business.
Who we bought the car from
By now, you're probably wondering whom we bought our new car from. Did Nancy get the deal? Well, no. Not that day at least. We decided to do a little more research and to find someone who really knows about cars to talk to about our transportation needs and how best to meet them. Buying a car is a big decision, and we wanted to work with someone that would help us solve our problem rather than hype us about headlights. If you're looking at deploying a voice enabled application, doing that may be a good idea, too.
About the Author
Daniel Enthoven is director of marketing for BeVocal.