Forecasting time-varying arrivals: Impact of direct response advertising on call center performance

Highlights

• We analyze high-frequency advertising and call data by specifically addressing subtle seasonal cycles.

• We assess the impact of advertising on call center performance.

• We develop a forecasting model for the incoming calls.

• We create a discrete event simulation to simulate the manpower planning system.

• We demonstrate the importance of the link between marketing activities and operational management.

Abstract

This study investigates manpower planning and the performance of a national call center for scheduling car repairs and responding to road interventions. We model the impact of advertising on the required capacity and develop a forecasting model for incoming calls, where the impact of direct-response advertising is considered. With the estimation results, we forecast the number of incoming calls to the call center. Next, the forecasts are input into the capacity planning simulation module to directly simulate a service process at the highly disaggregated level. This simulation mimics the service level requirements and queue behavior and shows that the call center is operating at a high level of efficiency and performance. We illustrate that advertising may cause a temporary overload of the system and increase the number of abandoned calls, which is suboptimal for 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.