When speech applications do not not make adjustments for the in-call behavior of individuals, real cost efficiencies are lost.
Without any kind of adjustment based on caller behavior, everyone is treated the same way regardless of their knowledge, experience, script navigation skills, calling environment and willingness to use voice self-service.
For one thing, all audio content is delivered to the caller at the same WPM rate regardless of their demonstrated behavior during the call. These applications do not listen for signs that the caller understands what is being said and is comfortable with the pace and content of the dialogue. Without “tuning in” to a caller’s behavior during the call, it’s easy to lose your audience before they even get a chance to engage with the IVR. We have conducted some production trials to test this hypothesis.
At a client site, two independent trials were conducted to determine the effects adaptive technology would have on the performance of the IVR. The voice application used was designed to handle caller inquiries for medical insurance claims, benefits, member coverage and general information. The voice application served primarily members (generally novice users) and providers (generally expert users).
During Trial 1, Audio Playback Speeds of 100, 110, 114, 117 and 119 percent of normal were used. During Trial 2, these values were changed to 85, 88, 92, 100, 110, 114, 117 and 119. For both trials, a caller-adaptive speed changing algorithm as incorporated into the VUI Cloud Service (http://www.vuicloud.com/technologies.html) was used.
The data in Table 1 illustrates production results gathered during these trial periods. These data indicate improvements in IVR Utilization, as measured by the number of IVR turns per call. An IVR Turn is defined as a single instance of a caller encountering a Call Script Node (CSN) and entering a response.
IVR Utilization Changes due to Adaptation
As Table 1 shows, the mean number of IVR turns for adaptive and standard in Trial 1 were, respectively, 4.4 and 3.2. This represents a 36.90 percent increase in utilization of the automated system when adaptation is used.
It should be noted that while the there was a dramatic improvement in IVR Utilization here, indicating that callers were actually using the the IVR more, this would not mean a lot if those same callers experienced more input errors. The caller input error rate actually went down during these trials.
These results, and the accompanying reduction in error rates, indicate that callers are more likely to stay with the automated system provided it moves at their comfortable speaking rate – even if that rate is 10-19 percent faster than it would be without adaptation. This is consistent with the hypothesis that tuning into the natural rhythm and pace of the caller helps keep them engaged in the automated call process.
While the Trial 2 data for this same metric showed some improvement for adaptive calls at 4.7 versus 4.5 for standard calls, the change was not statistically significant. This is likely due to the callers not engaging well to the slower speeds in Trial 2. Finding the “sweet spot” for a particular voice application and it’s unique calling base is a matter of adjusting the speed parameters used.
Automated speech and touch-tone calls cost about $.75 each to answer while agent-handled calls are about $4.25 on average. Telecom costs can range from less than $.02 to $.06 or more depending on call volume, carrier service contracts and other factors. Hosted IVR services can be as high as $.40 per minute for small call volumes.
Keeping callers happily engaged in your IVR has obvious economic benefits besides being good for repeat business. Studies show that disgruntled callers will defect and it’s not like IVR systems in general have a great reputation to begin with. Your agents should be free to answer the tricky, sensitive and high value calls and it serves no one well when they are tied up listening to a customer complain about how the IVR did not serve their needs.
For additional information on the trials conducted here, white papers with additional trial data or details on the use of adaptive technologies in the IVR, please contact [email protected]
