Kaizen is the Japanese word for “a change for the better” (”kai” means “change” and “zen” means “good”), which results in “continuous improvement”. (1)
Striving for excellence, always looking for ways to improve what already exists, and believing that one can impact change, is at the heart of the Kaizen spirit.
Speech applications life-cycle management is essentially a Kaizen methodology. Tuning – continuous improvement can have a significant impact on the caller experience and adoption of speech automation for customer service. For maximum effectiveness, tuning should included detailed analysis of the application performance using live caller data. In applying Kaizen to speech applications, we need to consider:
- Areas of the speech application where improvements can be beneficial – Particular attention should be paid to the dialog states that occur early on in the application, as it is these states that receive the most traffic and therefore provide the greatest opportunity for improving the overall performance of the application. Also, a detailed analysis of the entire call flow (speech automation business process) will reveal hot spots that need attention
- Speech application statistics – including call volume, uptake rate, zero out rate (DTMF, speech), abandon rate, recognition rate, completion rate, caller satisfaction and others
- Sources for the statistics – Logging mechanisms used to record the callers’ inputs and to collect information such as the recognition results, call volume statistics, the status of barge-in and many others
Typically, Kaizen applied to customer service speech applications will include:
- Dialog Tuning – includes the changes to the call flow and to the audio recordings. The scope of dialog tuning may include wording changes, navigational restructuring, or re-recording with different pacing or intonation. For example, analysis of data from a speech application that shows too many zero-outs on the first menu itself may indicate that the callers are confused by the menu choices and may require breaking down the call flow into nested menus of questions with yes/no choices.
- Recognition tuning – includes grammar modifications as well as adjustments to the recognition parameters. Grammar tuning requires a thorough analysis of the spoken utterances, incorporating factors such as misrecognitions and out-of-grammar utterances. Recognition parameters include phonetic dictionaries, grammar probabilities and confidence thresholds
Kaizen – the methodology for continuous improvement comes naturally to all speech applications. In addition to the concrete results of improved efficiency and effectiveness, the biggest benefit from speech applications and Kaizen is the culture of continuous improvement. As the benefits of speech applications such as increased efficiency by improved automation rates and increased effectiveness by increased consistency and accuracy of responses are observed and appreciated, one can expect Kaizen to be adopted into other areas of customer service business processes such as the mid-office and back-office. With Kaizen, change management is now easier – leading to a virtuous, continuous cycle of business performance improvement.
(1) “The Meaning of Kaizen” - Kaizen Consulting Group
0 Responses to “Kaizen for Customer Service Speech Applications”
Leave a Reply