Data from new trials show a better customer experience


Share on LinkedIn

In my last entry below, I mentioned some production tests we performed at a client site to measure optimal speaking rates (words per minute) for IVR callers.

Individual callers speed and accuracy determined whether (and by how much) the audio playback rate was increased or decreased at each IVR script Interaction Point (IP).

An IP might be as simple as answering yes/no, selecting from a menu, entering a 16 digit credit card number or anything in between.

A caller that behaved as an expert throughout the call would experience an audio playback speed adjustment profile that went from 100% to 108% and then to a final maximum rate of 112% of normal playback.

A caller that was not quite an expert, but better than average, experienced an adjustment profile of 100%, 105%, 108% and 110%.

Callers with less than average speed and accuracy experienced similar adjustment profiles in the opposite direction – 100%, 95%, 92% and so forth.

For callers of all skill levels, the audio rate was corrected when callers went from behaving like an expert to behaving like a novice (and vice-versa) during the call.

Finally, callers that navigated the call script with average skills heard what they hear today – 100% all the way.

The Adaptive Audio software performed the WPM adjustments according to its internal, proprietary algorithms after it had gathered a sufficient calibration sample for caller skill levels at each script IP during production.

While the production run referenced in the blog entry below showed results indicating that over 80 percent of all callers qualified themselves (according to the above described WPM adjustment process) to speeds other than the “Normal” speaking rate, the results here showed a different type of result.

As Figure 1 below indicates, there were a total 1913 IP’s or 1.66 IP’s per call for the adaptive calls versus 1807 IP’s or 1.52 IP’s per call for standard. This represents a 9.2 percent improvement for adaptive over standard in terms of IP’s per call (IVR Utilization).

Figure 1 also shows that there were 278 caller input errors on 605 adaptive calls resulting in .17 errors per call. On the standard calls, this rate was .19 errors per call, resulting in an overall error rate reduction of 10.5 percent, again in favor of the adaptive calls.

Figure 1 – IVR Performance Improvements with Adaptive Audio

It should be noted here that the low number of IP’s per call was due in large part to the fact that this production trial was carried out at a service provider that supports many different types of voice applications. Some of those apps simply announce a message and transfer the caller – no opportunity for adaptation here. Other apps just direct callers via one or two audio messages. The numbers collected and shown here reflect these conditions.

The benefits described above came largely from the calls with three or more IP’s.

It is also worth noting that these benefits were produced with basic adaptive functionality – only the audio playback speeds were adjusted to match caller skills and other adaptive features like Adaptive Timeout Extensions and Dynamic Application Smoothing were not implemented during this particular trial.

To learn more about how dynamically adjusting the call experience to suit individual callers can help your bottom line, contact Interactive Digital ( today.

Daniel O'Sullivan
CEO, innovator and technologist in software engineering and product development. Created and implemented Adaptive Technology and Fastrack Software products that have optimized over 1.5 Billion self-service phone calls worldwide and saved clients over $100M to date. Electrical Engineering undergrad with a Masters in Computer Science. Lucent/Bell Labs alumni. Winner of worldwide eco-design project and received several patents. Currently CEO of Software Technology Partners.Focus: Business Development, Technology Partnering, Mobile, Web and Cloud Technologies and Human-Computer Interaction.


Please use comments to add value to the discussion. Maximum one link to an educational blog post or article. We will NOT PUBLISH brief comments like "good post," comments that mainly promote links, or comments with links to companies, products, or services.

Please enter your comment!
Please enter your name here