The Modern Call Center: A Multi-
Disciplinary Perspective on Operations
Management Research
Zeynep Aksin • Mor Armony • Vijay Mehrotra
College of Administrative Sciences and Economics, Koc University, Rumeli Feneri Yolu,
34450 Sariyer-Istanbul, Turkey
Leonard N. Stern School of Business, New York University, West 4th Street, KMC 8–62, New York,
New York 10012, USA
Department of Decision Sciences, College of Business, San Francisco State University, 1600 Holloway Avenue,
San Francisco, California 94132-1722, USA
zaksin@ku.edu.tr • marmony@stern.nyu.edu • vjm@sfsu.edu
C
all 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.
Key words: call centers; staffing; skill-based routing; personnel scheduling; outsourcing
Submissions and Acceptance: Submissions and Acceptance: Received April 2007; revision received October
2007; accepted October 2007.
1. Introduction
Virtually all businesses are interested in providing
information and assistance to existing and prospective
customers. In recent years, the decreased costs of tele-
communications and information technology have
made it increasingly economical to consolidate such
information delivery functions, which led to the emer-
gence of groups that specialize in handling customer
phone calls. For the vast majority of these groups,
their primary function is to receive telephone calls that
have been initiated by customers. Such operations,
known as “inbound” call centers, are the primary
topic of this paper.
Inbound call centers are very labor-intensive oper-
ations, with the cost of staff members who handle
phone calls (also known as “agents”) typically com-
prising 60– 80% of the overall operating budget. In-
bound call centers may be physically housed across
several different locations, time zones, or countries.
Inbound call centers make up a large and growing
part of the global economy. Although reliable in-
dustry statistics are notoriously hard to come by, the
Incoming Call Management Institute (ICMI), a
highly reputable industry association, regularly
tracks published industry statistics from several
sources (www.incoming.com/statistics/demographics.
aspx). By 2008, various studies cited by ICMI predict
the following:
•
The United States will have over 47,000 call cen-
ters and 2.7 million agents.
POMS
PRODUCTION AND OPERATIONS MANAGEMENT
Vol. 16, No. 6, November-December 2007, pp. 665–688 doi 10.3401/poms.
issn 1059-1478 兩 07 兩 1606 兩 665$1.25 © 2007 Production and Operations Management Society
665
•
Europe, the Middle East, and Africa together will
have 45,000 call centers and 2.1 million agents.
•
Canada and Latin America will have an estimated
305,500 and 730,000 agents, respectively.
Meanwhile, the demand for call center agents in India
has grown so fast that the labor supply has been
unable to keep up with it: by 2009, the demand for
agents in India is projected to be over 1 million, and
more than 20% of those positions will be unfilled
because of a shortage of available skilled labor.
When a customer calls an inbound call center, var-
ious call handling and routing technologies will at-
tempt to route the call to an available agent. However,
there is often no agent available to immediately an-
swer the phone call, in which case the customer is
typically put on hold and placed in a queue. The
customer, in turn, may abandon the queue by hanging
up, either immediately after being placed on hold or
after waiting for some amount of time without receiv-
ing service. Once connected to an agent, a customer
will speak with that agent for some random time, after
which either the call will be completed or the customer
will be “handed off” to another agent or queue for
further assistance. The quality of the service is typi-
cally viewed as a function of both how long the cus-
tomer must wait to receive service and the value that
the customer attributes to the information and service
that is received.
Call center managers are increasingly expected to
deliver both low operating costs and high service
quality. To meet these potentially conflicting objec-
tives, call center managers are challenged with de-
ploying the right number of staff members with the
right skills to the right schedules in order to meet an
uncertain, time-varying demand for service. Tradi-
tionally, meeting this challenge has required call cen-
ter managers to wrestle with classical operations man-
agement decisions about forecasting traffic, acquiring
capacity, deploying resources, and managing service
delivery.
In recent years, the call center landscape has been
altered by a wide variety of managerial and techno-
logical advances. Reduced information technology
and telecommunications costs—the same forces that
contributed significantly to the growth of the call cen-
ter industry— have also led to rapid disaggregation of
information-intensive activities (Apte and Mason
1995). For call centers, this translated into increased
contracting of call center services to third parties (com-
monly referred to as “outsourcing”) and the disper-
sion of service delivery to locations across the globe
(“offshoring”). In addition, advances in telecommuni-
cations technologies enabled richer call center work-
flow, including increasingly intelligent routing of calls
across agents and physical sites, automated interac-
tion with customers while on hold, and call messaging
that results in automatic callbacks to customers once
an agent is available.
Also, as call centers now serve as the “public face”
for many firms, there is increasing executive consid-
eration of their vital role in customer acquisition and
retention. Similarly, the managerial awareness of call
centers’ potential to generate significant incremental
revenue by augmenting service encounters with po-
tential sales opportunities has also been growing rap-
idly: for example, a recent McKinsey study revealed
that credit card companies generate up to 25% of new
revenue from inbound call centers (Eichfeld, Morse,
and Scott 2006). However, for call center managers,
there is significant additional complexity associated
with managing this dual service-and-sales role with-
out compromising response times, service quality, and
customer satisfaction.
Finally, every call center manager is acutely aware
that phone conversations between customers and
agents are interactions between human beings. This
suggests that the psychological issues associated with
the agents’ experience can have a major impact on
both customer satisfaction and overall system perfor-
mance. Although these types of issues have been re-
searched extensively by behavioral scientists, opera-
tions management researchers have only recently
begun to explicitly include such factors in richer ana-
lytic models.
Given the size of the call center industry and the
complexity associated with its operations, call centers
have emerged as a fertile ground for academic re-
search. A relatively recent survey paper (Gans, Koole,
and Mandbelbaum 2003) cites 164 papers associated
with call center-related problems, and an expanded
on-line bibliography (Mandbelbaum 2004) includes
over 450 papers along with dozens of case studies and
books. In addition, there have been several more spe-
cialized surveys associated with call center operations,
including that of Koole and Mandelbaum (2002), who
focused on queueing models for call centers; L’Ecuyer
(2006), who focused on optimization problems for call
centers; and Koole and Pot (2006) and Aksin, Karaes-
men, and Ormeci (2007), who both focused on multi-
skill call centers.
This survey seeks to provide a broad perspective on
both traditional and emerging call center management
challenges and the associated academic research. The
specific objectives and major contributions of this pa-
per are as follows:
1. To provide a survey of the academic literature
associated with traditional call center problem areas
such as forecasting, queueing, capacity planning, and
agent scheduling over the past few years;
2. To identify several key emerging phenomenon
that affect call center managers and to catalog the
Aksin, Armony, and Mehrotra: The Modern Call Center
666 Production and Operations Management 16(6), pp. 665– 688, © 2007 Production and Operations Management Society
academic research that has been done in response to
these developments;
3. To recognize new call center operations manage-
ment paradigms that consider the role of the call cen-
ter in helping firms to attract, retain, and generate
revenue from customers and to propose some impor-
tant implications of these new paradigms on future
research;
4. To chronicle research on psychological aspects of
call center agent experience, survey recent operations
management papers that have incorporated some of
these ideas into their modeling, and suggest ways in
which such work can be incorporated into future op-
erations management research; and
5. To highlight gaps in the current literature on call
center operations management and opportunities ar-
eas for future research.
The remainder of the paper is organized as follows. In
Section 2, we survey recent work on traditional call
center operations management problems. Section 3
reviews research that considers demand modulation
as an alternative to supply side management. In Sec-
tion 4, we look at the research literature that emerged
as a result of technology-driven innovations, includ-
ing multi-site routing and pooling, the design of
multi-skill call centers, the blending of inbound calls
with other types of workflow such as outbound calls
and emails, and increased call center outsourcing. In
Section 5, we examine several key human resources
issues that affect call centers and chronicle recent op-
erations management research that sought to incorpo-
rate some of these factors into their models. In Section
6, we explore research that integrates call center oper-
ations with sales and marketing objectives, focusing
on cross-selling and long-term customer relationship
management. In each of the above sections, we sug-
gest specific opportunities for future research. Con-
cluding comments are provided in Section 7.
2. Managing Call Center Operations:
The Traditional View
Traditional operations management challenges for call
center managers include the determination of how
many agents to hire at what times based on a long-
term forecast of demand for services (“resource acqui-
sition”) and the scheduling of an available pool of
agents for a given time period based on detailed short-
term forecasts for a given time period (“resource de-
ployment”). In addition, once initial resource deploy-
ment decisions have been made, there may be
additional shorter-term decisions to be made, includ-
ing forecast updating, schedule updating, and real-
time call routing.
Resource acquisition decisions must be made sev-
eral weeks and sometimes months ahead of time be-
cause of lead times for hiring and training agents.
Also, because most call centers have fairly high em-
ployee turnover and absenteeism levels, models that
support resource acquisition decisions must explicitly
account for random attrition and absenteeism.
Resource deployment decisions are typically made
1 or more weeks in advance of when the calls actually
arrive. A cost-effective resource deployment plan at-
tempts to closely match the supply of agent resources
with the uncertain demand for services. The (highly
variable) demand for resources is expressed in terms
of call forecasts, which are typically composed of call
arrival distributions and service time distributions,
both of which vary over time. This variability means
that both forecasting and queueing models play an
important role in modeling resource deployment de-
cisions. From a scheduling perspective, agents can
typically be assigned to a range of shift patterns, and
the process of determining an optimal (or near-opti-
mal) schedule has a significant combinatorial com-
plexity.
In addition, as new data about forecasts and agent
availability becomes available for a given day or week,
this information can be used to modify both the near-
term call arrival forecasts and the agent schedules that
are driven by them. Finally, as calls actually arrive,
there may be specific decisions to be made about
queuing policies or call routing.
In this section, we begin our survey by looking at
recent work on these call center operations manage-
ment problems. We focus on call forecasting in Section
2.1, resource acquisition in Section 2.2, and perfor-
mance evaluation, staffing, scheduling, and routing in
Section 2.3. Next, we consider the basic problems of
staffing, scheduling, and routing when arrival rates
are random in Section 2.4. Finally, Section 2.5 provides
a brief overview of developments in performance
evaluation models for call centers, reflecting some of
the newer characteristics of modern call centers.
2.1. Call Forecasting
Call forecasts are defined by (a) the specific queue or
call type associated with the forecast; (b) the time
between the creation of the forecast and the actual
time period for which the forecast was created (often
referred to as the forecasting “lead time”); and (c) the
duration of the time periods for which the forecasts
are created, which can range from monthly (to sup-
port resource acquisition decisions) to short time
frames, such as 15-, 30-, or 60-minute periods (to sup-
port resource deployment decisions). Over the years,
there have been relatively few papers that focused on
forecasting call volumes, prompting Gans et al. (2003)
to assert that call forecasting was “still in its infancy.”
However, in the past few years, there have been a
handful of important developments in the call fore-
Aksin, Armony, and Mehrotra: The Modern Call Center
Production and Operations Management 16(6), pp. 665–688, © 2007 Production and Operations Management Society 667
casting field, driven by increased availability of his-
torical databases of call volumes and by utilization
and adaptation of new techniques that have been ap-
plied to similar forecasting problems in other applica-
tion areas.
Weinberg, Brown, and Stroud (2007) propose a mul-
tiplicative effects model for forecasting Poisson arrival
rates for short intervals, typically 15, 30, or 60 minutes
in length, with a 1-day lead time. In their setting, the
call arrival rate for a given time interval of a particular
day of the week is modeled as the product of the
forecasted volume for that day of the week and the
proportion of calls that arrive in that time interval plus
a random error term. To estimate the model’s param-
eters, the authors adopt a Bayesian framework, pro-
posing a set of prior distributions, and using a Monte
Carlo Markov chain model to estimate the parameters
of the posterior distribution.
Although computationally intensive, the methodol-
ogy proposed by Weinberg, Brown, and Stroud (2007)
is quite valuable from an operational perspective. In
particular, because the model produces forecasts of
Poisson arrival rates on an intra-day interval basis,
these results can be used in conjunction with perfor-
mance models and agent scheduling algorithms. In
addition, the authors propose a modification of this
method to allow for intra-day forecast updating,
which can in turn be used to support intra-day agent
schedule updating. The paper includes a forecasting
case study in which data from a large North American
commercial bank’s call centers are used to test both the
1-day-ahead forecasts and intra-day forecast updates,
with very promising results.
Soyer and Tarimcilar (2007) introduce a new meth-
odology for call forecasting that draws on ideas from
survival analysis and marketing models of customer
heterogeneity. Specifically, this paper models call ar-
rivals as a modulated Poisson process, where the ar-
rival rates are driven by advertisements that are in-
tended to stimulate customers to contact the call
center. The parameters for the call intensity associated
with each particular type of advertisement and future
time interval are modeled by a Bayesian framework,
using a Gibbs sampler (Dellaportes and Smith 1993) to
approximate the posterior distributions. The authors
also test their methodology by conducting numerical
experiments using call volume data from a call center
for which all calls can be traced directly to specific
advertisements, with the forecasts being created for
single- and multi-day time periods.
Shen and Huang (2007) develop a statistical model
for forecasting call volumes for each interval of a given
day and also provide an extension of their core mod-
eling framework to account for intraday forecast up-
dating. Their model is based on the use of singular
value decomposition to achieve a substantial dimen-
sionality reduction, and their approach also decom-
poses predictive factors into inter- and intra-day fea-
tures. For the empirical cases presented, the
methodology produces forecasts that are more accu-
rate than both the (highly unsophisticated) standard
industry practice and the results from Weinberg,
Brown, and Stroud (2007); the methodology is also
significantly less computationally intensive than the
Monte Carlo Markov chain methods of Weinberg,
Brown, and Stroud (2007).
Taylor (2007) presents an empirical study that com-
pares the performance of a wide range of univariate
methods in forecasting call volumes for several UK
bank call centers as well as for the Israeli bank call
center data from Brown et al. (2005), considering lead
times ranging from 1 day to 2 weeks. Taylor’s perfor-
mance comparison includes methods that have ap-
peared previously in the call center literature, such as
seasonal Auto Regressive Moving Average modeling
(Andrews and Cunningham 1995) and dynamic har-
monic regression (Tych et al. 2002), as well as several
other models that have not previously been used for
call center forecasting. The latter group includes an
exponential smoothing model for double seasonality
that was originally developed for forecasting short-
term electric utility demand (Taylor 2003); a periodic
Auto Regressive model; and a model based on robust
exponential smoothing based on exponentially weighted
least absolute deviations (Cipra 1992). The empirical
comparison showed no clear “winner,” because differ-
ent methods proved to be more effective under differ-
ent lead times and different workloads.
2.2. Personnel Planning: Resource Acquisition
The call center resource acquisition problem has been
studied by a handful of researchers. Gans and Zhou
(2002) model a process in which agents are hired and
experience both learning and attrition over time, dem-
onstrating that a threshold policy for hiring agents is
optimal in their setting. Ahn, Righter, and Shanthiku-
mar (2005) look at a general class of service systems
and demonstrate that under the assumption of contin-
uous number of agents who can be hired and fired at
will, the optimal policy is of a “hire-up-to/fire-
down-to” form. Bordoloi (2004) combines control the-
ory and chance-constrained programming techniques
to derive steady-state workforce levels for different
knowledge groups and a hiring strategy to achieve
these targets. Bhandari, Harchol-Balter, and Scheller-
Wolf (2007) consider both the hiring of regular work-
ers and the contracting of part-time workers along
with the operational problem of determining how
many part-time workers to deploy under different
load conditions. Ryder, Ross, and Musacchio (2008)
examine the impact of different routing strategies on
employee learning in a multi-skill environment in an
Aksin, Armony, and Mehrotra: The Modern Call Center
668 Production and Operations Management 16(6), pp. 665– 688, © 2007 Production and Operations Management Society
attempt to understand the connection between rout-
ing, learning, and overall staffing needs.
Given the importance of the resource acquisition
decision, there is significant need for additional re-
search in this area, including models for long-term
forecasting, personnel planning for general multi-skill
call centers, and resource acquisition planning for in-
creasingly complex networks of service providers (as
described by Keblis and Chen 2006, for example).
2.3. Personnel Planning: Staffing, Scheduling,
and Routing
The traditional approach to call center resource de-
ployment decisions is to attempt to build an agent
schedule that minimizes costs while achieving some
customer waiting time distribution objectives. As
such, targeted staffing levels for each period of the
scheduling horizon are typically key inputs to the
scheduling and rostering problems. These targets de-
pend on both how much work is arriving into the call
center at what times (as estimated by the call volume
forecasts and the forecasted mean service times) and
how quickly the call center seeks to serve these cus-
tomers (estimated by some function of the customer
waiting time distribution). Once the forecasts and
waiting time goals have been established, queueing
performance evaluation models are used to determine
the targeted number of service resources to be de-
ployed. The actual performance obtained from the
deployed resources also depends on the operational
problem of allocating incoming calls to these resources
dynamically, known as the call routing problem. Our
review follows the same hierarchical order that would
be followed in the resource deployment problem for
call centers: we first review staffing problems, then
provide an overview of scheduling and rostering
problems, and finally demonstrate how the call rout-
ing problem interacts with them.
2.3.1. Staffing Problems. Simulation models and
analytic queueing models are the two alternatives to
performance evaluation. Mehrotra and Fama (2003)
provides an overview of the inputs required for build-
ing a call center simulation model, while Koole and
Mandelbaum (2002), and Mandelbaum and Zeltyn
(2006) are good sources for a detailed overview of
queueing models of call centers.
The simplest queueing model of a call center is the
M/M/s queue, also known as an Erlang-C system.
This model ignores blocking and customer abandon-
ments. The Erlang-B system incorporates blocking of
customers. The Erlang-C model is further developed
to incorporate customer impatience in the Erlang-A
system (Garnett, Mandelbaum, and Reiman 2002).
Performance measures and approximations for the
Erlang-A system are discussed by Mandelbaum and
Zeltyn (2007b). Sensitivity of this model to changes in
its parameters is analyzed by Whitt (2006c), where it is
demonstrated that performance is relatively insensi-
tive to small changes in abandonment rates.
For most inbound call centers, the management ob-
jective is to achieve relatively short mean waiting
times and relatively high agent utilization rates. Gans
et al. (2003) refer to such an environment as a “Quality
and Efficiency Driven” regime. In this context, let R be
the system-offered load measured in terms of the
mean arrival rate times and the mean service time. The
so-called “square-root safety-staffing rule” stipulates
that if R is large enough then staffing the system with
R ⫹

公R servers (for some parameter

) will achieve
both short customer waiting times and high server
utilization.
This rule was first observed by Erlang (1948) and
was later formalized by Halfin and Whitt (1981) for
the Erlang-C model (i.e., an M/M/s queue). Its prac-
tical accuracy was tested for service systems by Kole-
sar and Green (1998). This rule was further supported
by Borst, Mandelbaum, and Reiman (2004) and Mag-
laras and Zeevi (2003) under various economic con-
siderations. This rule has since been demonstrated to
be robust with respect to model assumptions such as
customer abandonment (Garnett, Mandelbaum, and
Reiman 2002; Zeltyn and Mandelbaum 2005), an in-
bound call center with a call-back option (Armony and
Maglaras 2004a,b), and call centers with multiple
queues and agent skills (Gurvich, Armony and Man-
delbaum 2006, Armony and Mandelbaum 2004),
which will be discussed in more detail below.
Borst, Mandelbaum, and Reiman (2004) have also
identified two other operating regimes: the quality
driven and the efficiency driven (ED) regimes, which
are rational operating regimes under certain costs
structures. In the ED regime server utilization is em-
phasized over service quality; however, with cus-
tomer abandonment, this regime can also result in
reasonable performance as measured by expected
waiting time and fraction of customer abandonment
(Whitt 2004b). Whitt has proposed fluid models for
system approximation under the ED regime (Whitt
2006a,b) and has shown its applicability in staffing
decisions under uncertain arrival rate and agent ab-
senteeism.
Most of the early literature on staffing deals with
these problems in settings with a single pool of ho-
mogenous agents (see references in Gans et al. 2003;
Garnett, Mandelbaum, and Reiman 2002; Borst, Man-
delbaum, and Reiman 2004; Atlason, Epelman, and
Henderson 2004; and Massey and Wallace 2006). Re-
cent literature on staffing models focuses on multi-
skill settings, that is, in call centers where calls of
different types are served using service representa-
tives with different skills (Pot, Bhulai, Koole, 2007;
Bhulai, Koole, and Pot, 2007; Cezik and L’Ecuyer,
Aksin, Armony, and Mehrotra: The Modern Call Center
Production and Operations Management 16(6), pp. 665–688, © 2007 Production and Operations Management Society 669