Monday, 13 February 2017
There has been recent progress in the analysis of call-center data. Call-by-call data from a small number of sites have been obtained and analyzed, and these limited results have proven to be fascinating. In some cases, such as the characterization of the arrival process and of the delay of arriving calls to the system, conventional assumptions and models of system performance have been upheld. In others, such as the characterization of the service-time distribution and of customer patience, the data have revealed fundamental, new views of the nature of the service process. Of course, these limited studies are only the beginning, and the effort to collect and analyze call-center data can and should be expanded in every dimension in Kenya and Africa at large.
Perhaps the most pressing practical need is for improvements in the forecasting of arrival rates. For highly utilized call centers, more accurate, distributional forecasts are essential. While there exists some research that develops methods for estimating and predicting arrival rates, I strongly believe there is surely room for additional improvement to be made both here at home and the entire continent. However, further development of models for estimation and prediction will depend, in part, on access to richer data sets. Some of us believe that much of the randomness of Poisson arrival rates may be explained by covariates that are not captured in currently available data.
Procedures for predicting waiting-times are also worth pursuing. Field-based studies that characterize the performance of different statistics and methods would also be of value. More broadly, there is need for the development of a wider range of descriptive models. While a characterization of arrival rates, abandonment from queue, and service times are essential for the management of call centers, they constitute only a part of the complete picture of what goes on. For example, there exist (self ) service times and abandonment (commonly called “opt-out”) behavior that arise from customer use of IVRs. Neither of these phenomena is likely to be the same as its CSR analogue. Similarly, sojourn times and abandonment from web-based services have not been examined in multi-media centers.
Parallel, descriptive studies are also needed to validate or refute the robustness of initial findings. For example, lognormal service times have been reported in two call centers, both of which are part of retail financial services companies. Perhaps the service-time distributions of catalogue retailers or help-desk operations have different characteristics.
Similarly, one would like to test some finding that the waiting-time messages customers hear while tele-queueing promote, rather than discourage, abandonment.
It would also be interesting to put work on abandonment (Palm, Roberts, Kort, Mandelbaum with Sakov and Zeltyn) in perspective. These studies provide empirical and exploratory models for (im)patience on the phone in Sweden in the 40’s, France in the late 70’s, the U.S. in the early 80’s, Israel in the late 90’s and now Africa(Kenya in particular under this research) in the early millennium. A systematic comparison of patience across countries, for current phone services, should be a worthy, interesting undertaking.
There is the opportunity to further develop and extend the scope of explanatory models. Indeed, given the high levels of system utilization in the QED Quality and Efficiency Driven (operational) regime, a small percentage error in the forecast of the offered load can lead to significant, unanticipated changes in system performance. In particular, the state of the art in forecasting call volumes is still rudimentary. Similarly, the fact that service times are lognormally distributed enables the use of standard parametric techniques to understand the effect of covariates on the (normally distributed) natural log of service times.
In well-run QED call centers, only a small fraction of the customers abandon (around 1-3%), hence about 97% of the (millions of ) observations are censored. Based on such figures, one can hardly expect any reasonable estimate of the whole patience distribution, non-parametrically at least. Fortunately, however, theoretical analysis suggests that only the behavior of impatience near the origin is of relevance, and this is observable and analyzable.
Indeed, call-center data are challenging the state-of-the-art of statistics, and new statistical techniques seem to be needed to support their analysis. Two examples are the accurate non- parametric estimation of hazard rates, with corresponding confidence intervals, and the survival analysis of tens of thousands, or even millions, of observations, possibly correlated and highly censored.
Last but certainly not least, a broader goal should be, in fact, the analysis of integrated operational, marketing, human resources, and psychological data. That is, the analysis of these integrated data is essential if one is to understand and quantify the role of operational service quality as a driver for business success.
A prerequisite for understanding the financial effects of operational decisions is the ability to analyze an integrated data set that includes operational (ACD) automatic call distributor and marketing / business (customer information systems) data. With this information, one can attempt to tease out the longer-term, financial effects of operational policies.
My experience has been that both types of data are very difficult to access, however. One reason for this is technical. Only recently have the manufacturers of telephone equipment given customers something of an “off the shelf ” ability to capture, store, and retrieve detailed, call-by- call data. Similarly, the integration of these operational data with the business data captured in customer information systems is only now becoming widely available. Another reason stems from confidentiality concerns; most of our Kenyan companies are rightly wary of releasing customer information. Once managers recognize the great untapped value of these data, i believe they will employ mechanisms for preserving confidentiality in order to reap the benefit.
Ultimately, i envision a data-repository that is continuously fed by many call centers of varying types. The collected data would be continuously and automatically analyzed, from both operations and marketing perspectives. Then the data would be both archived and fed back to the originating call centers, who would use it (through visualization tools) to support ongoing operations, as well as tactical and strategic goals.
Little imagination is required for appreciating the value of such a data-base. As a start, its developer could become a benchmark that sets industry standards, as far as customer-service quality and call-center efficiency are concerned. As already mentioned, such a data-base would enable the identification of success-drivers of call-center business transaction.
Researched & Compiled: Samwel Kariuki
Date: 12th Feb 2017