Forecasting time-varying arrivals: Impact of direct response advertising on call center performance
Introduction
Call centers, where customers' questions, needs, and requests are addressed, are essential systems, especially for large organizations (Buist et al., 2008). For companies that organize their businesses by collecting their orders via call centers, there is a crucial link between their call centers’ effectiveness and their marketing activity. To meet the targeted service level, managers have to employ an appropriate number of skilled staff for the appropriate amount of time in call centers. The scheduling of an available pool of agents based on detailed short-term forecasts is a fundamental challenge for call center managers. As peak call moments can have short duration, the call center's staffing is not always sufficiently flexible to accommodate this demand. Therefore, modeling instability in call center arrivals and calculating staff requirements are crucial issues in call center management (Aktekin and Soyer, 2008, Chassioti and Worthington, 2004).
In practice, it can be challenging to provide the desired service level in the call center at peak call moments. Arrival rates may vary due to the hour of the day, day of the week, and month of the year and may vary across ad campaigns. (Testik, Cochran & Runger, 2004). In this study, we analyze the effectiveness of the operations of a specific call center, and we indicate routes to improvement. Our approach can be considered prototypical for companies in similar industries. Our analysis aims to investigate the extent that the results from the advertising/call relationship can be utilized in the capacity or manpower planning of the call center. The key research issue is the extent that the call center capacity can handle increased calls due to advertising. We investigate whether advertising affects the efficiency of the call center.
The outline of the paper is shown in Fig. 1. First, we develop a forecasting model for the incoming calls to the call center. Second, we construct a capacity planning model by discrete event simulation, where the forecasts feed the simulation. Last, we analyze the service level requirements and queue behavior in terms of waiting times, abandoned calls, and idle time.
The operation of the call center relates to queuing theory. For simpler, stable queuing systems, analytical tools can be employed. In our case, the average number of servers or agents varies from 1 to 48 or more each day. Moreover, a complex dynamic process characterizes the number of incoming calls. Therefore, we resort to a simulation of the system rather than to analytical methods.
In our study, two measures of the call center activity are of interest to the company. The first measure addresses relevant calls. These calls pertain to the business; refer to Kiygi-Calli et al. (2012) for more details. The second measure pertains to all incoming calls, which may be related to information requests or orders. This second measure is more relevant for the call center's operation, while the first measure more related to the actual conversion of calls to sales. We use a discrete event simulation that directly models the process, which allows integration of the forecasting system and the call center simulation. Our paper is an empirically relevant study for characterizing call center operations and performance.
The remainder of the paper is organized as follows: Section 1 describes the call center operation and the objectives that management wants to achieve in terms of effectiveness. In Section 2, we briefly review and discuss the literature related to our topic. 3 Data, 4 Calls forecasting model explain the data and modeling of the incoming calls, followed by a discussion of the model results in Section 5. Section 6 contains the simulation model, and Section 7 discusses the simulation results. In Section 8, we discuss the conclusions.
Section snippets
Literature review
Many companies communicate with their customers via internally managed or outsourced call centers. In these centers, traditional telephone service is enhanced by additional customer contact channels, for instance, by interactive voice response (IVR) (Koole and Mandelbaum, 2002).
The random variation in incoming calls of the call center is a significant feature when assessing call center performance (Betts et al., 2000). According to the need of the call center manager, the call center data might
Data
We analyze the data related to a national call center, which collects all requests from consumers. The company broadcasts direct-response radio and TV commercials on national channels. In the first part of our study, we measure the impact of direct-response commercials on the number of incoming calls. The call center operates on a 24/7 basis, and the number of incoming calls is recorded in real time (seconds). Service centers are located in two different regions: Flanders (Dutch-speaking part
Calls forecasting model
The call center operates according to a single queue/multiserver principle, and the total number of incoming calls is used as a dependent variable in our model. We compare the various models of interest by considering their forecast accuracy. Our study follows the modeling exercises of Chandy et al., 2001, Tellis et al., 2000, Tellis and Franses, 2006, and Kiygi-Calli et al., 2012, Kiygi-Calli et al., 2017), which generally recommend analysis at the most detailed level possible because
Results of the forecasting model
Table 1 shows the parameter estimates of model (1). The regions are Belgium-North (Region 1), Belgium-South (Region 2), and Undefined Calls (Region 3). The estimation results of the three regions present similarities in terms of the significant lag orders. We observe a highly significant goniometric wave for the autoregressive (AR) effects according to the estimation results. The first three lags of the AR terms show a pronounced and highly significant goniometric pattern, which are specified
Simulation
In the second part of this study, we evaluate the company's call center performance when direct-response commercials are broadcast. The call centers are dynamic systems that can only be modeled by simulation. To simulate the manpower planning system in seconds, we generate a discrete event simulation system. The simulation system is explained in Section 6.2 in detail. In the simulation, we link the week-ahead forecast of the forecasting model, which was previously discussed, with the call
Results of the simulation
The week-ahead forecasts are computed for inactive and active weeks, and the waiting times of the customers are evaluated for both weeks. Table 3 gives the percentage of the waiting times for the active and inactive weeks. For the active week, the percentage of the waiting times less than 5 s is 74.87%. The percentage of waiting times less than 12 s is 81.45%, and the percentage of waiting times less than 20 s is 100%.
In contrast, for the inactive week, the percentage of waiting times less than
Conclusions
In the first part of this study, we constructed a forecasting model for incoming calls, considering the impact of direct response commercials. We employed an ADL model and found a highly significant intraday pattern for the first three lags of the autoregressive terms. The forecasts from this model were utilized as inputs for evaluating the manpower planning of the call center. The stated call center objectives are quite ambitious, that is, for 80 percent of the incoming calls, the waiting
Funding
The authors acknowledge the financial support of the Research Foundation - Flanders (FWO), Belgium (Project number: G.0361.07).
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Meltem Kiygi-Calli has been an Assistant Professor in the Department of Business Administration at Kadir Has University, Istanbul, Turkey since September 2017. She received her PhD degree in Applied Economics/Marketing from the University of Antwerp, Belgium in 2011. Her research particularly covers innovative contributions to advertising research, forecasting and operations management based on the application of hierarchical models to high-frequency data.
References (35)
- et al.
Modeling intra-seasonal heterogeneity in hourly advertising-response models: Do forecasts improve?
International Journal of Forecasting
(2017) - et al.
The effectiveness of high-frequency direct-response commercials
International Journal of Research in Marketing
(2012) - et al.
The modern call center: A multi-disciplinary perspective on operations management research
Production and Operations Management
(2007) - Aktekin, T. and Soyer, R. (2008). Call center arrival modeling: A Bayesian state space approach. The institute for...
- et al.
Workload forecasting for a call center: Methodology and a case study
The Annals of Applied Statistics
(2009) Introduction to discrete event simulation and agent-based modeling: voting systems, health care, military, and manufacturing
(2011)- et al.
Forecasting call frequency at a financial services call centre
Journal of the Operational Research Society
(2002) - et al.
Dynamic skill based routing: A system dynamics approach to a policy definition in call center management
- et al.
Call center staffing with simulation and cutting plane methods
Annals of Operations Research
(2004) - Avramidis, A. N., Chan, W., Gendreau, M., L’ecuyer, P., & Pisacane, O. (2010). Optimizing daily agent scheduling in a...
Using simulation in call centers
Call centre capacity management
International Journal of Service Industry Management
Statistical analysis of a telephone call center: A queueing-science perspective
Journal of the American Statistical Association
Speeding up call center simulation and optimization by markov chain uniformization
Service quality and customer abandonment: A system dynamics approach to call center management
Service quality and customer abandonment: A system dynamics approach to call center management
Introduction to discrete event systems
Cited by (2)
Design and Implementation of Call Center Service System Based on CTI
2022, Proceedings - 2022 6th International Conference on Wireless Communications and Applications, ICWCAPP 2022
Meltem Kiygi-Calli has been an Assistant Professor in the Department of Business Administration at Kadir Has University, Istanbul, Turkey since September 2017. She received her PhD degree in Applied Economics/Marketing from the University of Antwerp, Belgium in 2011. Her research particularly covers innovative contributions to advertising research, forecasting and operations management based on the application of hierarchical models to high-frequency data.
Marcel Weverbergh was an Emeritus Professor at the University of Antwerp, Belgium. He passed away on 30 October 2018.
Philip Hans Franses is a professor of applied econometrics and a professor of marketing research at the Erasmus School of Economics (ESE).
- 1
Marcel Weverbergh was Professor Emeritus at the University of Antwerp; he passed away on 30 October 2018.