IEEE Computer Society and GBC/ACM

7:00 PM, Thursday, 19 March 2015

MIT Room E51-395

Big Data versus Big Insight - A Case Study from Healthcare Customer Service

Bernhard Suhm

In recent years, larger amounts of data have become available than ever before. While most research in 'big data' is focused on developing efficient data processing methods, equally important to gain novel insight from these large data sets are meaningful metrics and the ability to drill down to root cause within data sets. In this talk we describe how we accomplished this goal in our call center optimization work, and illustrate them with case studies in healthcare. - The metrics that matter in the call center environment are operational cost and caller experience. Agent labor cost represent 70% of the operational cost. Once we identify issues in how calls are handled, we estimate the frequency relative to the total inbound call volume, and measure the time agents spend on specific parts of the call handling process. The product of frequency and time spent per instance yields a simple model of issue impact on average agents call handling time (AHT). - Prior research suggests that effort is a better indicator of repeat business and overall customer experience than some commonly used survey-based measures of customer satisfaction, such as Net Promoter Score (NPS). To measure the impact on caller experience we developed a benchmark of 22 automated metrics of caller effort, such as time spent in the menu system, whether the caller was transferred to a second agent, or previously attempted to resolve their issue on a website. These effort-related metrics relate to both the difficulty of obtaining and the operational cost of delivering customer service, and thus represent a great criterion for call center optimization. - Having access to the complete caller experience, the complete call from dialing to hang up, has been invaluable to empower us to drill to root cause and deliver actionable recommendations. The complete experience is difficult to get by, especially when data is unstructured and fragmented like in many Big Data settings; however, the complete experience for a reasonably small subset of the data is sufficient for most problems of practical relevance. The talk will illustrate our methodology for mining end-to-end customer service calls with case studies from our work for large healthcare customer service centers, and demonstrate the cloud-based call analytics solution that captures anywhere from thousands to millions of live calls.

Bernhard Suhm joined the AVOKE Analytics team in 2006 to establish the professional services group. As Director of Professional Services he manages the delivery of data analysis and consulting services to the subscribers of the AVOKE end-to-end call analytics solution. Previously, as a Senior Scientist at BBN Technologies Bernhard co-developed many elements of the AVOKE Analytics solution. Bernhard has over 20 years of experience working with speech recognition, voice user interfaces, and data-driven methods to optimize contact center satisfaction and efficiency. He and has co-authored several patents, book chapters, and published papers on these topics. Prior to joining AVOKE, Bernhard was a Senior Consultant with the Enterprise Integration Group, a leading developer of voice user interfaces, and a Research Programmer with the Interactive Systems Laboratories at Carnegie Mellon University. Bernhard received a PhD in Computer Science specializing in speech user interfaces from Karlsruhe University in Germany.

This joint meeting of the Boston Chapter of the IEEE Computer and GBC/ACM will be held in MIT Room E51-395. E51 is the Tang Center on the corner of Wadsworth and Amherst Sts and Memorial Dr.; it's mostly used by the Sloan School.