No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Don't Call Us, We'll Call You: An Empirical Study of Caller Behavior Under a Callback Option
... A call center that offers a callback service gives callers the option between waiting on hold and letting agents call them back later (often within a guaranteed maximum delay). The callback service has been both theoretically studied (Armony and Maglaras 2004a,b) and empirically investigated (Hathaway et al. 2020). The rationale for callbacks is similar to that of line-sitting: callbacks do not reduce callers' waiting time, but free them up while they are waiting. ...
... Second, because line-sitting is operated by a third party, it maintains the FIFO order, whereas callbacks present the service provider with an opportunity to prioritize online customers 29 waiting on hold over offline customers waiting for callbacks. Hathaway et al. (2020) empirically find that customers who accept callbacks are usually made to wait longer than if they do not. Thus, while waiting for a callback is less aggravating than waiting on hold, the callback service is not suited for customers with an urgent need. ...
... Moreover, when agents call back, customers may be preoccupied and unavailable to answer the call. Hathaway et al. (2020) find empirical support for these two limitations: the rejection rate of a callback in their data is as high as 62.5%, and even out of the accepted callbacks, 7% are unanswered. ...
Recent years have witnessed the rise of queue scalping in congestion-prone service systems. A queue scalper has no material interest in the primary service but proactively enters the queue in hopes of selling his spot later. This paper develops a queueing-game-theoretic model of queue scalping and generates the following insights. First, we find that queues with either a very small or very large demand volume may be immune to scalping, whereas queues with a nonextreme demand volume may attract the most scalpers. Second, in the short run, when capacity is fixed, the presence of queue scalping often increases social welfare and can increase or reduce system throughput, but it tends to reduce consumer surplus. Third, in the long run, the presence of queue scalping motivates a welfare-maximizing service provider to adjust capacity using a “pull-to-center” rule, increasing (respectively, reducing) capacity if the original capacity level is low (respectively, high). When the service provider responds by expanding capacity, the presence of queue scalping can increase social welfare, system throughput, and even consumer surplus in the long run, reversing its short-run detrimental effect on customers. Despite these potential benefits, such capacity expansion does little to mitigate scalping and may only generate more scalpers in the queue. Finally, we compare and contrast queue scalping with other common mechanisms in practice—namely, (centralized) pay-for-priority, line sitting, and callbacks.
This paper was accepted by Victor Martínez de Albéniz, operations management.
... Delay announcements are also relevant in healthcare provision; Dong et al. (2018) show that patients react to these announcements that can be used to coordinate emergency services in a network. There are few empirical studies analyzing customer behavior in outbound call centers, such as Hathaway et al. (2021), who use a structural estimation approach in a setting where customers are offered a call-back option, which has some similarities to an outbound call center. ...
... In this dimension, Dhar and Nowlis (1999) study how time pressure can lead choice overload in customers, producing higher procrastination. Furthermore, in an outbound call center, it may not be desirable to call immediately after a failed attempt as the customer may be unavailable to answer the phone (Samuelson 1999, Hathaway et al. 2021. ...
The gig economy open a new business model in which services have access to a large pool of workers that are compensated based on their actual production, which can be useful to operate at lower levels of utilization to improve response times to customers. However, having a large pool of workers with low utilization may lower their motivation and increase employee turnover, which can hurt productivity in the long run. “Balancing Agent Retention and Waiting Time in Service Platforms” looks at this tradeoff and provides a data-driven approach to manage worker capacity in on-demand service platforms, showing evidence through a real-world application of an outbound call center with freelance agents.
... Delay announcements are also relevant in healthcare provision; Dong et al. (2018) show that patients react to these announcements that can be used to coordinate emergency services in a network. There are few empirical studies analyzing customer behavior in outbound call centers, such as Hathaway et al. (2021), who use a structural estimation approach in a setting where customers are offered a call-back option, which has some similarities to an outbound call center. ...
... In this dimension, Dhar and Nowlis (1999) study how time pressure can lead choice overload in customers, producing higher procrastination. Furthermore, in an outbound call center, it may not be desirable to call immediately after a failed attempt as the customer may be unavailable to answer the phone (Samuelson 1999, Hathaway et al. 2021. ...
... The term Queueing science has been adopted in the recent queueing literature to denote datadriven mathematical models aimed to reproduce and predict the observed queuing systems data behaviour. A number of papers have contributed in this research area lately, see Brown et al. [2005], Soyer and Tarimcilar [2008], Demiriz et al. [2009], Aktekin and Soyer [2012], Ibrahim and L'Ecuyer [2013], Ding et al. [2015], Armony et al. [2015], Barrow and Kourentzes [2018], Azriel et al. [2019], Hathaway et al. [2021], Albrecht et al. [2021]. This paper also undertakes a Queueing science perspective since it deals with exploratory analysis, statistical modeling and performance estimation of the queueing system generated by the well known "Anonymous bank" call center data (already analyzed by some of the previous references). ...
... The term Queueing science has been adopted in the recent queueing literature to denote data-driven mathematical models aimed to reproduce and predict the observed queuing systems data behaviour. A number of papers have contributed in this research area lately, see Brown et al. [2005], Soyer and Tarimcilar [2008], Demiriz et al. [2009], Aktekin and Soyer [2012], Ibrahim and L'Ecuyer [2013], Ding et al. [2015], Armony et al. [2015], Barrow and Kourentzes [2018], Azriel et al. [2019], Hathaway et al. [2021], Albrecht et al. [2021]. This paper also undertakes a Queueing science perspective since it deals with exploratory analysis, statistical modeling and performance estimation of the queueing system generated by the well known "Anonymous bank" call center data (already analyzed by some of the previous references). ...
In this paper we analyze the well-known "Anonymous bank" call center dataset from a queueing science viewpoint. For this purpose, fitted distributions for both the inter-arrival and service times as well as for customers patiences are integrated in a simulator to infer quantities of interest related to call centers managerial decisions as waiting times, abandonment rates and queue lengths. In particular, it is shown how a type of Markov renewal process, the Markovian arrival process (MAP), is able to capture some of the characterizing properties of arrivals in a modern call center as overdispersion and positive correlation between arrival counts. The work provides a new inference approach for the MAP based on the count process descriptors and presents new properties concerning the dependence structure of the cumulated number of arrivals in a MAP.
... For instance, Liu et al. (2018) conduct discrete choice experiments to elicit patient willingness to pay for waiting. Hathaway et al. (2021) employ structural estimation to quantify how callers trade off between waiting for callbacks (akin to leisure waiting) and waiting in queue (similar to regular waiting). Finally, it is important to note that estimating the exact costs of waiting may be more challenging than those of idling because measuring provider productivity is generally more straightforward for management. ...
Problem definition: Multistage service is common in healthcare. One widely adopted approach to manage patient visits in multistage service is to provide patients with visit itineraries that specify personalized appointment time for each patient at each service stage. We study how to design such visit itineraries. Methodology/results: We develop the first optimization modeling framework to provide each patient with a personalized visit itinerary in a tandem (healthcare) service system. Due to interdependencies among stages, our model loses those elegant properties (e.g., L-convexity and submodularity) often utilized to solve the classic single-stage models. To address these challenges, we develop two original reformulations. One is directly amenable to off-the-shelf optimization software, and the other is a concave minimization problem over a polyhedron shown to have neat structural properties, based on which we develop efficient solution algorithms. In addition to these exact solution approaches, we propose an approximation approach with a provable optimality bound and numerically validated performance to serve as an easy-to-implement heuristic. A case study populated by data from the Dana-Farber Cancer Institute shows that our approach makes a remarkable 28% cost reduction over practice on average. Managerial implications: Common approaches used in practice are based on simple adjustments to schedules generated by single-stage models, often assuming deterministic service times. Whereas such approaches are intuitive and take advantage of existing knowledge of single-stage models, they can lead to significant loss of operational efficiency in managing multistage services. A well-designed patient visit itinerary that carefully addresses the interdependencies among stages can significantly improve patient experience and provider utilization.
History: This paper was selected for Fast Track in the M&SOM Journal from the 2022 MSOM Healthcare SIG Conference.
Funding: The work of the last two authors was supported in part by the National Natural Science Foundation of China [Grants 71931008 and 72001220].
Supplemental Material: The online appendix is available at https://doi.org/10.1287/msom.2023.0134 .
... Hathaway et al. (2023) analyze the behavior of frontline workers in service encounters, analyzing their decisions to either address customer requests themselves or escalate. Hathaway et al. (2021) use a structural model to study callback policies in a call center setting. They find that offering callbacks as a demand postponement strategy reduces average online waiting time. ...
Despite being one of the most cost-effective and sustainable modes for transporting freight, railways globally have been rapidly losing market share in the inland freight transportation sector. One of the salient reasons for this is the slow speed of freight trains in many parts of the world. For example, in Indian Railways, the world's fourth-largest in size, the average freight train speed is only around 25 kmph and has remained constant for the past few decades. The slow pace of freight trains is because passenger trains, which share the same railway infrastructure, get prioritized in dispatch by railway traffic managers (also known as section controllers). In this paper, we empirically study freight delays in the Indian railway setting by analyzing how section controllers make freight train stop and hold decisions while dispatching freight trains. Subsequently, we propose policies to reduce freight delays and, thus, increase trains' speeds through the network. We use detailed high-frequency network congestion data and estimate a structural model to estimate the key parameters underlying the controllers' decisions. The estimated parameters provide empirical evidence for (i) the priority accorded to passenger trains over freight trains, (ii) the push effects in the freight train queue, and (iii) the strategic behavior of section controllers in holding trains at larger stations. Using the estimated model, we conduct a set of counterfactual analyses to address the problem of slow freight train speeds. First, we evaluate the impact of constructing Freight Only Corridors (FOCs), high-capacity corridors reserved for freight transport. We find that the FOCs lead to about a 29% reduction in freight train delays and a 12% improvement in train speeds. Then, we also evaluate non-capacity-investment-based alternatives to FOCs, like (i) threshold-based releases for freight trains dwelling longer than a specified time limit and (ii) freight capacity consolidation by using vertically stacked trains. Interestingly, we find that our non-capacity interventions can provide benefits similar to those of FOCs while being considerably cheaper. Specifically, a 45-minute threshold release policy leads to around 31% reduction in dwell times and 9% increase in speeds. Similarly, vertically consolidating freight capacity by about 25% leads to around a 10% increase in speed, comparable to the improvement achievable with the FOC. Our policy recommendations for improving freight speeds could enhance the overall efficiency of India's transportation infrastructure, benefiting the country's economic and social development.
... In this way, waiting inbound turns to the outbound task at scheduled moments. The analysis of such systems is conducted in Armony & Maglaras [9,10], Hathaway et al. [67], Legros et al. [87], and Legros et al. [89]. ...
We give an overview of the practice and science of call center workforce planning, in which we evaluate the commonly used methods by their quality and the theory by its applicability. As such, this paper is useful for developers and consultants interested in the background and advanced methodology of workforce management and for researchers interested in practically relevant science.
Supplemental Material: The e-companion is available at https://doi.org/10.1287/stsy.2021.0008 .
... Buell (2021) focuses on the position of the wait and finds that waiting in last place exacerbates the probability of abandonment. Another recent and position-related study is by Hathaway et al. (2021), who find that customers experience significant less discomfort when waiting offline (i.e., via callback option) as opposed to waiting in a queue. Contrary to waiting's negative effect on abandonment, Ülkü et al. (2020) find that customers tend to consume more after long waits, provided that they do not abandon the queue. ...
It is well known that the waiting time a customer experiences in a service system is determined by the service processing time of preceding customers, among other factors. We argue that a directionally opposite effect, which diffuses from waiting time to her own service time, also exists in co‐productive service contexts where a significant fraction of the service time is contributed by the customer. Multiple underlying customer behavioral mechanisms lead us to hypothesize that waiting's impact is dependent on the service stage and magnifies as the service process approaches completion. Our empirical analysis uses a unique operational data set that combines server log information with instant‐messaging transcripts collected from a live‐chat contact center. We show that pre‐service waiting accelerates customer engagement—one dimension of customer instigated service time—only at the beginning of the conversation and then exhibits a slowdown effect as the conversation proceeds. In contrast, in‐service waiting consistently slows down customer responses—another dimension of customer instigated service time, the magnitude of which is higher in later episodes of the agent‐customer message exchanges. We discuss the practical implications of our findings on operational policies employed in contact centers.
... Yet, exploring this type of behavioral change is beyond the scope of our study and deserves a multidisciplinary research effort involving psychology and marketing as well. We refer the reader to Hathaway et al. (2019) who attempt to fill this gap via an empirical study. ...
In this paper, we study the system with two finite-size queues where underlying exponential random variables have state-dependent rates. When all servers are busy, upon arrival customers may join the online or the offline/callback queue or simply balk. Customers waiting in the online queue are impatient and if their patience expires, they may choose to join the callback queue instead of abandoning the system for good. Customers in the callback queue are assumed to be patient. Customers are served following a threshold policy: when the number of customers in the callback queue surpasses a threshold level, the next customer to serve is picked from here. Otherwise, only after a predetermined number of agents are reserved for future arrivals, customers remaining in the callback queue can be served. We conduct an exact analysis of this system and obtain its steady-state performance measures. The times spent in both queues are expressed as Phase-type distributions. With numerical examples, we present how the policy responds when shorter callback times are promised or customer characteristics vary.
... In this way, waiting inbound turns to the outbound task at scheduled moments. The analysis of such systems is conducted in [8,9,64,84,82]. ...
We give an overview of the practice and science of call center workforce planning, where we evaluate the commonly used methods by their quality and the theory by its applicability. As such this paper is useful for developers and consultants interested in the background and advanced methodology of workforce management, and for researchers interested in practically relevant science.
... Aksin et al. (2013) model caller abandonment as an optimal stopping problem, and show that modeling endogenous caller behavior can be important when operational changes such as those in service discipline are implemented at call centers. Hathaway et al. (2020) empirically study callers' behavior when callback options are offered in the queue. In the context of abandonment, they find that offering reasonable callback options allows demand management by postponing demand; and decreases callers' average waiting costs, the percentage of calls lost due to abandonment, and the percentage of callers not answering callbacks. ...
In many emerging economies, callers may abandon ambulance requests due to a combination of operational (small fleet size), infrastructural (long travel times) and behavioral factors (low trust in the ambulance system). As a result, ambulance capacity, which is already scarce, is wasted in serving calls that are likely to be abandoned later. In this paper, we investigate the design of an ambulance system in the presence of abandonment behavior, using a two-step approach. First, because the callers’ actual willingness to wait for ambulances is censored, we adopt a Maximum Likelihood Estimator estimation approach suitable for interval censored data. Second, we employ a simulation-based optimization approach to explicitly incorporate customers’ willingness to wait in: (a) tactical short-term decisions such as modification of dispatch policies and ambulance allocations at existing base locations; and (b) strategic long-term network design decisions of increasing fleet size and re-designing base locations. We calibrate our models using data from a major metropolitan city in India where historically 81.3% of calls were successfully served without being abandoned. We find that modifying dispatch policies or reallocating ambulances provide relatively small gains in successfully served calls (around 1%). By contrast, increasing fleet size and network re-design can more significantly increase the fraction of successfully served calls with the latter being particularly more effective. Redesigning bases with the current fleet size is equivalent to increasing the fleet size by 8.6% at current base locations. Similarly, adding 29% more ambulances and redesigning the base locations is equivalent to doubling the fleet size at the current base locations and adding 34% more ambulances and redesigning base locations is equivalent to a three-fold increase. Our results indicate that in the absence of changes in behavioral factors, significant investment is required to modify operational factors by increasing fleet size, and to modify infrastructural factors by redesigning base locations.
... Aksin et al. (2013) model caller abandonment as an optimal stopping problem, and show that modeling endogenous caller behavior can be important when operational changes such as those in service discipline are implemented at call centers. Hathaway et al. (2020) empirically study callers' behavior when callback options are offered in the queue. In the context of abandonment, they find that offering reasonable callback options allows demand management by postponing demand; and decreases callers' average waiting costs, the percentage of calls lost due to abandonment, and the percentage of callers not answering callbacks. ...
We model the decision-making process of callers in call centers as an optimal stopping problem. After each waiting period, a caller decides whether to abandon a call or continue to wait. The utility of a caller is modeled as a function of her waiting cost and reward for service. We use a random-coefficients model to capture the heterogeneity of the callers and estimate the cost and reward parameters of the callers using the data from individual calls made to an Israeli call center. We also conduct a series of counterfactual analyses that explore the effects of changes in service discipline on resulting waiting times and abandonment rates. Our analysis reveals that modeling endogenous caller behavior can be important when major changes (such as a change in service discipline) are implemented and that using a model with an exogenously specified abandonment distribution may be misleading.
This paper was accepted by Assaf Zeevi, stochastic models and simulation.
In many service industries, companies compete with each other on the basis of the waiting time their customers' expe-rience, along with other strategic instruments such as the price they charge for their service. The objective of this paper is to conduct an empirical study of an important industry to measure to what extent waiting time performance measures impact different firms' market shares and price decisions. We report on a large scale empirical industrial organization study in which the demand equations for fast-food drive-thru restaurants in Cook County are estimated based on so-called structural estimation methods. Our results confirm the belief expressed by industry experts, that in the fast-food drive-thru industry customers trade off price and waiting time. More interestingly, our estimates indicate that consumers attribute a very high cost to the time they spend waiting.
Call centers are an increasingly important part of today's business world, employing millions of agents across the globe and serving as a primary customer‐facing channel for firms in many different industries. Call centers have been a fertile area for operations management researchers in several domains, including forecasting, capacity planning, queueing, and personnel scheduling. In addition, as telecommunications and information technology have advanced over the past several years, the operational challenges faced by call center managers have become more complicated. Issues associated with human resources management, sales, and marketing have also become increasingly relevant to call center operations and associated academic research.
In this paper, we provide a survey of the recent literature on call center operations management. Along with traditional research areas, we pay special attention to new management challenges that have been caused by emerging technologies, to behavioral issues associated with both call center agents and customers, and to the interface between call center operations and sales and marketing. We identify a handful of broad themes for future investigation while also pointing out several very specific research opportunities.
Motivated by practices in customer contact centers, we consider a system that offers two modes of service: real-time and postponed with a delay guarantee. Customers are informed of anticipated delays and select their preferred option of service. The resulting system is a multiclass, multiserver queueing system with state-dependent arrival rates. We propose an esti- mation scheme for the anticipated real-time delay that is asymptotically correct, and a routing policy that is asymptotically optimal in the sense that it minimizes real-time delay subject to the deadline of the postponed service mode. We also show that our proposed state-dependent scheme performs better than a system in which customers make decisions based on steady-state waiting-time information. Our results are derived using an asymptotic analysis based on "many-server" limits for systems with state-dependent parameters. Subject classifications: service networks; service level guarantees; multiclass queueing systems; call-back option; call centers; Halfin-Whitt regime; real-time delay notification. Area of review: Manufacturing, Service, and Supply Chain Operations. History: Received June 2002; revision received January 2003; accepted July 2003.
Organizations worldwide use contact centers as an important channel of communication and transaction with their cus- tomers. This paper describes a contact center with two channels, one for real-time telephone service, and another for a postponed call-back service offered with a guarantee on the maximum delay until a reply is received. Customers are sensi- tive to both real-time and call-back delay and their behavior is captured through a probabilistic choice model. The dynamics of the system are modeled as an M/M/N multiclass system. We rigorously justify that as the number of agents increases, the system's load approaches its maximum processing capacity. Based on this observation, we perform an asymptotic analysis in the many-server, heavy traffic regime to find an asymptotically optimal routing rule, characterize the unique equilibrium regime of the system, approximate the system performance, and finally, propose a staffing rule that picks the minimum number of agents that satisfies a set of operational constraints on the performance of the system. Subject classifications: service networks; call centers; heavy traffic; service level guarantees; choice models; Nash equilibrium; Halfin-Whitt regime. Area of review: Manufacturing, Service, and Supply Chain Operations. History: Received September 2001; revision received May 2002; accepted May 2003.
We undertake an empirical study of the impact of delay announcements on callers' abandonment behavior and the performance of a call center with two priority classes. A Cox regression analysis reveals that in this call center, callers' abandonment behavior is affected by the announcement messages heard. To account for this, we formulate a structural estimation model of callers' (endogenous) abandonment decisions. In this model, callers are forward-looking utility maximizers and make their abandonment decisions by solving an optimal stopping problem. Each caller receives a reward from service and incurs a linear cost of waiting. The reward and per-period waiting cost constitute the structural parameters that we estimate from the data of callers' abandonment decisions as well as the announcement messages heard. The call center performance is modeled by a Markovian approximation. The main methodological contribution is the definition of an equilibrium in steady state as one where callers' expectation of their waiting time, which affects their (rational) abandonment behavior, matches their actual waiting time in the call center, as well as the characterization of such an equilibrium as the solution of a set of nonlinear equations. A counterfactual analysis shows that callers react to longer delay announcements by abandoning earlier, that less patient callers as characterized by their reward and cost parameters react more to delay announcements, and that congestion in the call center at the time of the call affects caller reactions to delay announcements.
The authors develop a model of customer channel migration and apply it to a retailer that markets over the Web and through catalogs. The model identifies the key phenomena required to analyze customer migration, shows how these phenomena can be modeled, and develops an approach for estimating the model. The methodology is unique in its ability to accommodate heterogeneous customer responses to a large number of distinct marketing communications in a dynamic context. The results indicate that (1) Web purchasing is associated with lower subsequent purchase volumes than when buying from other outlets; (2) marketing efforts are associated with channel usage and purchase incidence, offsetting negative Web experience effects; and (3) negative interactions occur between like communications (catalog x catalog or e-mail x e-mail) and between different types of communications (catalog x e-mail). The authors find that over the four-year period of their data, a Web-oriented "migration" segment emerged, and this group had higher sales volume. Their post hoc analysis suggests that marketing efforts and exogenous customer-level trends played key roles in forming these segments. The authors rule out alternative explanations, such as that the Web attracted customers who were already heavy users or that the Web developed these customers into heavier users. They conclude with a discussion of implications for both academics and practitioners.
Cutting in line: Social norms in queues
- G Allon
we will be right with you: Managing customer expectations with vague promises and cheap talk
- G Allon
The impact of delaying the delay announcements
- G Allon