Wednesday, October 29, 2008

Voice Recognition Projects Bring More Than Just ROI

In these belt-tightening times, any new voice recognition project must pay its way in the short term with a satisfactory near-term return on investment. It should also improve the customer experience, since customers are often alienated by ineffective customer service centers. Since IVR projects can be introduced incrementally to complement human agents, it is usually not difficult to show that, in the short term, service is improved at a lower overall cost. On cost savings alone, the ROI is often acceptable.

Even a basic IVR project can sometimes have an ROI to brag about. Just think about the system recently launched by Victrio to fight credit card fraud. A report on the new system reads:
Victrio introduces a new credit risk management services that identifies fraudsters by their voice during credit card authorisation phone calls. The service works by comparing the callers' voice against a database of known criminals. Authorisation is denied if a match between a caller's voice and a fraudster's voice is found. The service is aimed at online merchants as well as banks and other credit card issuers.

Compare that with how present systems handle the same problem. Try calling your bank branch by telephone. The agent will quite rightly ask you a series of questions to verify that you are who you say you are. Unfortunately, in most cases all the answers could easily have been provided by someone who had stolen your identity from fairly public records. So the agent and you are tied up for a little while in a meaningless exercise. Systems based on PINs and passwords are fine but are still not unbreakable. They may also frustrate customers who forget their passwords or even the answers to the security questions.

On the other hand, voice data contains so much information that it can provide a very strong and clearly unbreakable identification system. At the same time, it speeds up the authorization process for much higher customer satisfaction at a lower system cost. The customer is very happy and the agent is only involved with operations that require a human agent.

Let us assume that the authentication/authorization process with a live agent takes an average of 20 seconds, whereas IVR authentication takes approximately 5 seconds. The agent time that is saved can be channelled to assist in other tasks, meaning that wait time would be shortened. Thus, clients would wait less and have shorter calls – a good recipe for increased customer satisfaction. That having been said, this is a very simplistic way of doing ROI. This evaluation only means that we have shortened the call authorization time to ¼ of the initial length, but the question is, have the operational costs been cut accordingly?

In order to really properly perform ROI analysis for your company, you will have to evaluate the overall costs of the IVR project and quantify the benefits in dollar savings. Doing so will show how the organization can cut costs and still improve its customer service. In addition, ROI analysis can evaluate how the company can reap the benefits of fraud detection, which translates into greater financial savings.

Overall, ROI involves detailed planning in order to calculate the estimated savings in operational costs and to forecast whether or not a project is worthwhile. ROI analysis allows the organization to give a green light for a project, with concrete steps to ensure a win-win situation all around.

Tuesday, October 14, 2008

What do IVR Systems Have to Do with the Election?

A polling company has recently instated automated IVR calling to predict the outcome of the election. The concerns raised by the recent changes showcase some of the prevalent suspicions regarding IVR technology. Read an article on the move to IVR polling here.

The article cites George Bishop, a professor of political science at the University of Cincinnati, who highlights common concerns regarding IVR:
What if you don't understand the question? There's no opportunity for clarification. There's also no one there to probe you to answer the question if the person isn't sure.

Bishop’s statement reflects a skepticism regarding the usability of IVR systems. He betrays the same preference for human agents that call centers often see in their callers. The same issues with respect to usability and system intelligence often return in discussion regarding voice recognition software systems. Still, the issue is not whether or not IVR is the best method for election polling. The issue surrounds how to build a voice recognition system that fits the bill. For example, there needs to be proper testing to ensure the IVR interface is tailored to the needs of its users. A good IVR system accounts for potential deviations from the script and is tailored to the expectations of callers. This can only be accomplished with usability testing. In addition to usability testing, system monitoring is often needed. Election polling is only implemented for a matter of days; but organizations that implement ongoing IVR systems require constant post-deployment monitoring to ensure that users are continually provided the best and most efficient service. The context of how the IVR system is used must be taken into consideration, with the expectations and mental model of users being considered. Of course companies want to maximize the ROI of an IVR project, and initial ROI analysis ensures that a voice recognition software project will yield the expected outcome.

So how to respond to the concerns raised by Bishop? Again, it’s a matter of having a customized IVR system that has been properly designed, tested, piloted, and tuned. Whether the IVR uses speech recognition software or responds to touch-tone dialing, the system needs to be tailored to the needs of the respondent to ensure optimal functioning. Proper usability research and testing reduces any concern over understanding the question.