Revisiting the ROI of Speech
Back in June 2002, while I was a senior vice president with The Kelsey Group, we conducted a research study on "The ROI of Speech." A select group of relatively large companies, primarily in the financial services and catalogue sales sectors provided some impressive numbers that have been baked into countless presentations delivered by the peripatetic SpeechWorks/Nuance sales force.
The top-line findings were glowing, to say the least. All of the firms under study experienced return on investment results that "met or exceeded" expectations. Using the most common yardstick for speech deployments, the "payback period," an average deployment recouped its cost in 9.5 months, with six months listed as the shortest time to positive ROI.
Symptomatic of the size of the firm under study, the "average savings per deployment" expressed on an annual basis were $1.5 million. This figure alone should tell you something. Clearly the investigation focused on firms that had sufficient call volumes and cost-per-call to qualify for the club that had $1.5 million to save in the first place.
Of course, the study probed into several "hidden factors" underlying ROI calculations. Increased opportunities for live agents to up-sell callers; reduction in the number of abandoned calls and increased agent efficiency all figure into the decision process that governs vendor selection and technical approach. In most cases, these "soft" factors were of secondary importance when measured against the hard "truth" of $1.5 million in savings.
Reasons to Revisit
Remember the summer of 2002? It was a time when strength and heroics ruled the silver screen. The blockbuster movie heroes included Tobey Maguire as SpiderMan, Matt Damon as Jason Bourne and Vin Diesel as xXx (though this doesn't explain Jar Jar Binks in the second prequel in the "Star Wars" series). We got the sense that our research subjects were using forceful tactics to bring speech into their self-service infrastructure with dramatic success.
The "Internet bubble" had largely burst, but the move to extend Web-based self-service logic to contact centers had not yet begun. To a large degree, the ROI of speech was very much tied to improvements in classic call center metrics: reduction in agent work time, shorter queuing times, and increased automation rates.
In the summer that brought us more subtle filmed fare (e.g. "The Forty-Year-Old Virgin"), we judged it time to take a more "nuanced" look at the quantitative and qualitative aspects of speech deployments. Going into the 2005 study, I expected to see results that reflected fundamental changes in the automated speech marketplace. This would include an emphasis on speed to deploy, giving rise to prepackaged applications and reusable code. A desire for self-sufficiency should be reflected in the percentage of companies developing and deploying applications with in-house staff. Finally, experience and success with early speech deployments should drive an increased percentage of respondents with multiple speech deployments.
The Results Are Extremely Positive
Nearly 80 percent of the respondents to this study achieved financial results that either "met or exceeded" their ROI objectives. Of the 20 percent who "did not meet" expectations, only five percent failed to gain cost savings from speech. The remainder merely failed to reach self-defined objectives. While this result doesn't "meet" the 100 percent positive reported in the 2002 study, we feel it is a realistic reflection of the state of the market and speaks well of the technology's adaptability to new architectures and methods of deployment.
The average payback period stretched to 11 months, but the variability around the mean was dramatic. The shortest payback was 3.5 months for a call routing and order entry application built from a prepackaged application. By contrast, the shortest payback in the 2002 study was six months. Looking at the hard-dollar savings, the range was $100,000 to $1.5 million, with an arithmetic mean of $375,000.
Although the dollar value is much lower than that reported in 2002, it is a very positive finding as it reflects speech automation's successful penetration of smaller companies as well as the proven profitability of smaller deployments. Not every implementation can be the $1.2 million "home run." Indeed, the industry will be more successful as it builds up the roster of "singles hitters" and prove that ROI objectives can be met even though the net savings generated are six-digit numbers, rather than seven digits or more.
Dan Miller is a senior analyst for Opus Research. He founded Opus Research, Inc. and published Telemedia News & Views, a monthly newsletter regarding developments in voice processing and intelligent network services. He can be reached at
dmiller@opusresearch.net .