Today, more and more IVR speech recognition applications begin with an open-ended prompt supported by a large statistical language model (SLM) grammar. The prompt invites callers to speak a short phrase describing what they want. For example, “Thank you for calling State Bank. How can I help you? You can say things like, ‘What’s my balance?’ or ‘Where can I find an ATM?’ Now, tell me what you’re calling about.” At West, we find that responses fall into four general categories: Read More >
Statistically successful speech recognition is not the only indicator of a successful self-service IVR application. If an overall speech success rate is not also accompanied by rising key performance indicators — such as containment rates, completion rates, and customer satisfaction scores — then a closer examination of what is working and what is not working needs to be examined.
In order to do this, organizations need to take an honest look at their customer care and realize that not all points of customer interaction are equally important. A good raw number or bottom line is not so important if the core customer needs are falling short.
For example: What is the real value to having a stellar rate of success on a “yes” or “no” confirmation type of prompt if more of the responses are “no,” indicating that customers are unable to successfully speak the correct information the first time? If the core functions of your IVR are underperforming but are statistically padded by confirmation-type prompts, then it’s time to focus less attention on the overall speech success rate and more on the most important drivers of true success. In short, it’s time to strive for a speech success rate that is weighted more heavily not only on higher volume prompts but also on importance. In doing this, you will recognize that not all prompts are created, or valued, equally.
Conversely, it is also important to recognize that just as is true with success, not all failures are the same. For example, in a non-speech-enabled application, a miskeyed entry is rejected. There is no chance that an invalid entry of say “4” can be misinterpreted as a “3” when using touch-tone.
In speech applications, this is handled the very same way. In this case, a spoken “4” will be rejected, as it is not a valid in-grammar response. Depending on a particular organization’s reporting, this properly rejected response is not viewed as a successful interaction even though it was treated properly. It is important to make the distinction between a phrase that was misunderstood and thus not recognized versus a phrase that was properly rejected.
How an organization wants to view and report the success of its speech application will not be the same for all, and that is fine. What is important is that those doing the reporting are looking beyond the raw numbers to the true picture of what speech is providing to the core business functions. Remember that a best practice is not so because a group of experts say that it is. A best practice is only so if it results in increased productivity, greater customer satisfaction and specific bottom-line results that are measurable and provide true value.
The average U.S. household has 3.5 phone numbers.
If it’s confusing to you to keep all of those numbers straight, then you can imagine the complexity involved for big business to keep data current to identify customers. The number of phones per household continues to grow, and we now have the ability to port numbers from one person to another. Wouldn’t it be great if companies could tie all that data together and identify a customer based on any one of those household numbers?
When you call a company’s IVR from a mobile device or a new number, chances are those businesses, with which you may be a long-time customer, are having a hard time identifying who you are — and then you’re unable to quickly access the services that you need. When a business can’t identify you, it increases call time, drives up operating costs and in most cases, can reduce customer satisfaction.
Twenty-seven percent of U.S. households are now mobile-phone-only, up from 14 percent in 2007.
With the culture shifting to a mobile-driven world, these issues are creating a strain on companies that are trying to keep their consumer data updated and relevant. More and more companies are now accessing learning databases that can identify and verify callers no matter which family member’s phone you may be calling from.
Customers, like you and me, want a personalized experience. We can still have it if companies can keep up with all our phone numbers.