Out of the Iron Age: Redefining Success
A recent visit to a major trade show (V-World) amid continued blurry first quarter financial results from IVR platform vendors has led me to wonder if it's time to give up some long-time analytical perspectives. No, this has nothing to do with New York's ban on smoking, we're way past that. It's about Iron, hardware, ports - lots of them.
Life was simple back in the Iron Age of call automation. Whoever sells the most ports wins. Big deals equal big discounts for hundreds or thousands of IVR ports. Even in the outsourced, IVR game, the value proposition was capacity-based: HSN was in the outsourced Voice Response Units (VRU) business in a big way, with 20,000 ports, more than a match for Call Interactive (now First Data Voice Services) with something like 10,000 ports (this was 10 years ago).
Back when life was simple, application complexity was not an issue and imposed no constraints. In today's environment of personae, VUI design and conversational voice response, the real limitations on self-service are not capacity-bound but design-bound. The decision to move from touchtone to speech is often dictated by application complexity - if enough people bail on your touchtone the ROI from migrating to speech will write itself. That is, of course, assuming you know what's going on with your IVR platform(s). There may be multiple clusters, they may be different revs, and the combination of network VRU and on-premises equipment may be such that you have no clear view of what's really happening in terms of resolutions, abandons or transfers.
Which brings us back to V-World. This year's edition was a milestone in the sense that it marked Nuance's transformation from a speech company to a platform company, marked by the formal introduction of the Nuance Voice Platform (NVP) 1.0. As I've said before, as a stand-alone platform company, Nuance looks great: nobody can touch it from the perspective of speech understanding, a referencible customer base of speech deployments, deep professional services, not to mention cash on hand. Apart from smaller, pure-play VoiceXML specialists like VoiceGenie that are a quarter of its size, Nuance is currently the only major voice platform vendor with strong growth characteristics.
One of the areas where Nuance put a lot of engineering support for NVP was reporting. If the influence of the Web on call processing is finally taking root, nowhere is this more evident than in the NVP reporting stack, where "hits" on very fine-grained speech 'sites' can be brought into view and analyzed from multiple perspectives. What does that tell us, in the larger sense? That the value of speech platforms is in understanding what's happening in your infrastructure and what your callers are doing.
What's the key to unlocking that value? Being able to adapt and optimize the system in response to those caller behaviors.
In one of those conceptual ironies (pun intended) that appeals to columnists the world over, the debut of the NVP platform came at a show that was teeming with hosting providers: Convergys, First Data, MCI, Tellme, Versay/Vail and West were all there in force. Buzz at the show in this sector included Versay's encroachment on Tellme's Allstate account, Tellme's win at Verizon, the continuing suspense over Amex, and who's using what tools for app development within hosted environments (and the answer appears to be Audium).
But the heavy metal behind the burgeoning options in the hosted arena is the combination of managed visibility, tunability and flexibility. None of these have to do with capacity, but with the increased resolution on the application's performance (visibility), leverage of recognizer accuracy across multiple applications and environments (tunability) and faster time to market with greater control that comes from multiple options on application server/telephony resource outsourcing (flexibility).
If that's a bit too high-level, let's break it down into hard cases. First Data provides us with three case studies. In the first, a straightforward DTMF-based CPE-to-outsourced scenario, Customer A doubles IVR resolution from 27 percent to 52 percent of all incoming traffic. Each percentage point of lift is worth $300,000 in agent worktime savings. That translates into a delta of $7.5 million with no capital expense involved. The beautiful part of this is a pricing scheme that incents the hosting provider to move the needle on resolution, by getting fatter minutes (price-wise) as the delta opens up. This is enabled by visibility and tunability (yes, even DTMF can be optimized).
On the other hand Customer B has a set of tech support applications that include domain-specific grammars and some tap-dancing queuing and routing routines. The hosting provider (First Data) writes and hosts some of the apps in VoiceXML on its premises. The customer writes some apps and runs them on its own app servers, collocated on the First Data premises. Just to show how flexible this gets, some of those app servers are Macs (no that's not a typo).
A third case has a DTMF-to-speech migration, with the initial request not specifying VoiceXML. But, given it involves customer addresses, and there's a sweet VoiceXML-coded grammar for place names, it's coded up that way. Future apps enabled by speech are all done in VoiceXML. Time-to-market increases, transfers decrease.
None of these are about capacity. They are about the ability of a managed/hosted provider to leverage expertise from multiple tenants and engagements to achieve results faster and higher than what an inhouse IT department would. The 'five nines' is a given, it's table stakes, it's Iron Age thinking. The new definition of success is not how many calls I take in, but how many are resolved and how fast the next app goes into production.
Mark Plakias is a partner and senior consultant for The Zelos Group. He can be reached at (212) 366-0895.