Decision Support Systems 58 (2014) 3142

Contents lists available at ScienceDirect

Decision Support Systems

journal homepage: www.elsevier.com/locate/dss

Capacity planning and performance contracting for service facilities

Yabing Jiang , Abraham Seidmann
a
Lutgert College of Business, Florida Gulf Coast University, Fort Myers, FL 33928 USA
b
Simon Graduate School of Business, University of Rochester, Rochester, NY 14627 USA

a r t i c l e i n f o a b s t r a c t

Available online 16 January 2013 Market demand uncertainty and time-based competition make capacity investment and managerial incentive
decisions for service facilities such as high-end diagnostic medical imaging centers, modern IT services, or
Keywords: contract manufacturing shops particularly challenging. These facilities compete on service quality, short
Cost allocation queuing times and speed. Therefore, having insufcient capacity can be economically devastating for them.
Capacity planning Given the high up-front costs involved, rms want to make sure that they neither over- nor under-invest in
Franchise contract
service capacity. These problems get exasperated by the fact that typically rms are unfamiliar with the local
Information asymmetry
Mechanism design
market conditions and do not closely observe the demand-generating efforts of the hired managers. Most
Service center prior studies of cost allocation methodologies, contract design, and service resource management tend to
address these aspects of the problem separately. They ignore the interaction effects between the capacity
decisions and the managerial adverse selection and moral hazard issues, which are crucial elements for successfully
running services with xed capacity, random arrivals, and stochastic service times. Our paper instead focuses on
the development of an integrated-approach to the simultaneous design of efcient managerial contracts and of
capacity planning for capital intensive service facilities. We derive optimal linear contracting structures under
information asymmetry between the rms and management, and analyze their impact on capacity decisions,
service levels, service volumes, and the allocations of costs. Surprisingly, we prove that even though a franchise
(charge-back) contract induces the rst-best effort from the manager, it is not always the best choice for the
rms as it may lead to inferior prots for them. In fact, our results explain why a rm's eventual contract choice
should be a function of its prior on the probability distribution of the local market demand. We also explain
when it may be optimal (for both the rm and the manager) to charge the manager up front a xed franchising
fee that is even greater than the total costs of capacity. Our study applies to many capital-intensive and
congestion-prone service systems, where the success of signicant up-front capacity investments also hinges on
the daily operations of those facilities run by hired managerswho typically possess specic knowledgethat
gives them a signicant information advantage.
2013 Elsevier B.V. All rights reserved.

1. Introduction
Effectively managing service centers such as call centers, computer-
ized diagnostic imaging facilities, data centers, SaaS businesses, and
telecommunication networks has always been a challenging task.
Owners of the centers (rms) are responsible for investing in capacity,
which is often capital intensive and involves signicant up-front xed
costs for equipment, software, and installation. While high utilization is
a critical protability driver, facing uncertain market demand, rms
also have to maintain an acceptable service level and relatively short
waiting times in order to compete in the market successfully. Managers
of these centers are often contracted to run them as prot centers.
As such, they are usually responsible for directing daily activities and
generating demand through additional marketing or service quality
Corresponding author.
E-mail addresses: yjiang@fgcu.edu (Y. Jiang), seidmannav@simon.rochester.edu
(A. Seidmann).
efforts. For example, we have seen that at most free-standing radiol-
ogy facilities that are run by the national networks in the US (such
as Insight Imaging and Radnet Inc.), the local managers are responsible
for stafng, scheduling, marketing, and other demand-generating
activities in their territory.
What makes the issue complicated is that managers, as the agents
running the service centers, often possess private information about
local market conditions, referral patterns, demographic preferences,
and their own demand-generating efforts, and they may not willingly
or truthfully share that information with their rms. Working with
a national rm that operates high-end diagnostic medical imaging
centers in various states, we had to make the correct capacity invest-
ments in each market and to obtain the true market demand informa-
tion from the local managers who are typically being incentivized by
the patient volume at their center. If the local managers overestimate
the demand, to allow ample capacity and fast turnaround times for all
exams, the rm will nd itself over-investing in capacity, such as MRI,
CT, or PET-CT equipment. On the other hand, if the local managers

0167-9236/$ see front matter 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.dss.2013.01.010
32 Y. Jiang, A. Seidmann / Decision Support Systems 58 (2014) 3142
underestimate the demand, in order to have a lower service quota in
their contract, the rm will under-invest in capacity, resulting in long
lines and lost prot opportunities.
In this paper, we outline the solution to this problem that has been
motivated by our practical experience and by the latest research
on optimal mechanism design. In our model, the manager (agent)
possesses private information regarding the local market and the
level of his own demand-generating effort; facing uncertain market
demand and the combined agency issues, the rm has to make up-
front capacity investment decisions and design incentive contracts to
motive the manager to truthfully share his specic market knowledge
and to exert the desired level of effort to generate demand. Several
studies [4,10,22,27] have pointed out that optimal mechanisms predi-
cated by economic theory are often sophisticated and rarely used in
the real world, whereas linear contracts are commonly used in practice.
Our focus is on evaluating three effective linear performance-based
incentive contracts under information asymmetry: a uniform contract,
a variable-rate contract, and a charge-back contract. We show that the
uniform contract, which offers the same contract terms regardless of
market conditions, fails to elicit market information from the manager
and leads to a distorted capacity investment decision. The variable-rate
and charge-back (franchise) contracts, on the other hand, can success-
fully elicit true market information from the manager. We show that
the latter even induces the rst-best (full-information) effort level
from the manager and hence leads to the optimal capacity investment,
the same as the full-information level.
We embed a queuing model within the principal-agent frame-
work because most service rms are constrained by a nite installed
capacity, yet they have to deliver reasonably fast turnaround times
when both demand and service rates could be random and outside
their control in the short run. Integrating the capacity decision
with the incentive contract design and the service level constraints,
therefore, enables us to address a complicated and yet practical
problem that has not been fully studied before. We not only derive
analytical solutions for this intricate problem but also nd some
interesting results. For example, we show that if the rm selects the
proper xed and variable charge terms, its franchise contract can
induce the rst-best demand-generating effort and it will successfully
elicit truthful market information from the manager; however
franchise contracting fails to produce the rst-best prot and it may
not always be the best choice for the rm. The xed charge term in
the franchise contract is naturally associated with the classical xed
cost allocation. It has been shown in managerial accounting that
rms can benet from xed cost allocation because the allocation
may inuence agents' consumption of perquisites or serve as a
proxy for the difcult-to-calculate opportunity costs [28]. Overall,
the accounting literature tends to emphasize the role of xed cost
allocation as an effective cost control to hold agents accountable so
that they will not over-consume resources at the expense of the
rm [2,15,27]. We extend that literature and show here in a service
queuing setting rms should also use cost allocation as a way to pre-
vent the managers from under-consuming resources. A high capacity
request is associated with a high xed charge, and the managers have
a natural incentive to request low capacity when they privately
observe a high-demand market. By doing so, they can shirk in
their demand-generating effort and gain greater utility. But, under-
consuming resources can be harmful to the rms as well, because
low capacity restricts the imaging center from serving more patients,
for instance, and also causes long patients' waiting times, which
translate to high delay costs and potential losses in future revenue
due to a degraded service-level reputation. We show that with proper
xed cost allocations rms can avoid both over-investing and under-
investing in capacity.
Our work contributes to the literature in multiple ways. First,
we develop and analyze a new contracting problem in a service
environment in which rms compete on meeting certain response
times and local managers possess valuable private information on
market demand. These managers can inuence market demand by
exerting demand-generating effort, such as marketing promotions
or by providing an outstanding customer service experience. Second,
we optimally integrate rms' operations capacity decision with
their incentive contract decision for a business environment with sto-
chastic arrivals of customers and randomly distributed service times
per customer. We investigate three practical mechanisms and show
that two of them can successfully solicit the manager's market infor-
mation. Third, we demonstrate the value of the local manager's
private information and prove that the rm's optimal contract choice
is a function of the market demand distribution. As expected, we show
that the cost of the manager's private information is associated with
market demand variance, with a higher demand variance correspond-
ing to a higher information cost. However, the manager's market
knowledge is more valuable when there is a low demand variance.
Overall, this research provides guidelines for rms that deal with
congestion-prone systems and sheds light on how to effectively manage
service facilities with combined moral hazard and adverse selection
issues.
The rest of the paper is organized as follows. We rst review the
relevant literature in Section 2. Then we present the model setting
in Section 3 and analyze the benchmark case with full information in
Section 4. In Section 5, we derive three contracts for the information
asymmetry case in which the manager's effort is not observable
and the manager has private information regarding local market
condition. We explain why two of these contracts can effectively
solicit true market information from the manager. We then analyze
the rm's optimal contract choice and evaluate the factors affecting
the installed capacity, the average throughput rates, and the expected
waiting times in Section 6 and conclude the paper with practice
guidelines for cost allocation and incentive design for service systems
in Section 7.
2. Literature review
This paper is related to research that applies agency theory as well
as the service resource management and accounting cost allocation
literature. Prior research on IT resource management has modeled
IT service centers as queuing systems and mostly focused on capacity
allocation within a rm through an internal pricing scheme [6,7,18].
In these works, demand is exogenous and there is no agency issue.
Clearly, this model setting does not apply to many modern IT service
centers that provide services to external clients because external
random demand cannot by controlled by the proposed internal transfer
pricing scheme, and it fails for any service centers that do experience
agency issues.
Marketing literature that applies the principal-agent framework, for
example, Basu et al. [3] and Lal and Srinivasan [14], often focuses on the
compensation contract design of a sales force selling a commodity good
with an unlimited supply and does not work for a service environment.
Harris et al. [8] study effective resource allocation in a manufacturing
setting under incomplete information. This is one of the early works
on mechanism design. In their model, agents have private information
on their productivity and effort levels. The objective of the rm is to
minimize the costs of producing a given level of output by designing
an efcient mechanism. However, since the output level is given there
is no demand uncertainty in their model.
Among recent research that combines queuing models with agency
theory, Jiang and Seidmann [12] model the capacity management and
contract design issue of a service organization with a nite capacity.
In their model, the service demand can be affected by the agent's
marketing effort, and the agent is both risk and effort averse. They
extend prior agency theory literature by introducing delay costs and
the capacity decision into the model and taking an integrated approach
to derive the optimal incentive contract and capacity investment for the
 Y. Jiang, A. Seidmann / Decision Support Systems 58 (2014) 3142
principal. Plambeck and Zenios [20] study dynamic incentive contracts
for make-to-stock production systems. They deal with a moral hazard
problem in which the agent dynamically controls the production rate
and has no inuence on market demand. Kim et al. [13] analyze and
compare two contractsa material contract and a performance-based
contractin the after-sales product support context in which a multi-
tasking agent is responsible for service part inventory and product
reliability decisions. They focus on a moral hazard problem only, and
in their model the agent has no inuence on demand. Baiman et al.
[1] also introduce a queuing model in the principal-agent framework,
but they model a manufacturing environment, and the agent's effort
affects the throughput rate of production, not the demand arrival rate.
Porteus and Whang [21] study a similar multiple agent problem in
a manufacturing setting as in Harris et al. [8], but they introduce
demand uncertainty and stochastic capacity and focus on the
moral hazard issue (agents' efforts) without information asymmetry
(no hidden productivity). They propose two incentive contracts;
one of them is a franchising plan that achieves the rst-best
outcomes. Hasija et al. [9] study call-center outsourcing contracts. In
their model, the call-center vendor has private information on its
staff's efciency (i.e., the service rate). The client offers outsourcing
contracts that specify nancial and service quality terms. In their
work, the vendor makes a stafng decision that affects the call
center's service quality, but it has no inuence on market demand.
While we also embed a queuing model in the principal-agent
framework, our research differs from these papers because we study
a combined moral hazard (hidden action) and adverse selection
(hidden information) problem in a service environment. The agent
possesses private information on market demand, and the agent's
effort can also inuence market demand. In a service queuing setting,
the principal needs to make a capacity investment decision while
facing uncertain market demand and service standard constraints
and design proper incentive contracts to motivate the agent to take
desired actions. We propose a charge-back contract that is similar to
the franchise contract in [21], and we show that the contract can
induce the rst-best effort from the agent, but unlike that in [21], it
cannot produce the rst-best prot and may not even produce the
highest prot for the principal.
Our paper is also related to Chen [5], in which the agent (a sales
agent with a negative exponential utility function) has private
information on market conditions and exerts selling effort that is not
observable by the principal. The principal makes an inventory stocking
decision and designs a compensation scheme (linear contract) to
induce the agent to truthfully reveal market conditions so that it can
better match supply with demand. The focus of [5], however, is to
derive the corresponding Gonik's scheme and compare it with a menu
of linear contracts. We also study a combined moral hazard and adverse
selection problem, but in a service setting, and we focus on evaluating
different incentive contracts and recommending the optimal contracting
choice for the principal.
This research is also connected to accounting cost allocation liter-
ature [2,15,19,25,26,28]. These papers all provide justications for the
allocation of xed costs to agents, but they either do not consider
agency issue or only consider the cases in which the agents possess
private information on the efciency of resources or their own types
or productivity, and they do not model a service environment and
the impact of capacity investment and agents' effort on service
quality. Our paper instead models a practical service environment
with complicated agency issues, and we further demonstrate that
incentive contracts that involve xed cost allocation terms can
effectively motivate the agent to communicate his private market infor-
mation to the principal so that he will not over-demand or under-
request capacity from the principal.
Overall, we extend the literature by considering incentive contract
design together with operations resource management. Our work
provides a complete solution for rms dealing with xed capacity
33
planning decisions combined with adverse selection and moral hazard
issues in an uncertain service environment characterized by time-
based competition, whereas prior studies on cost allocation methodolo-
gies, contract design, and service resource management tend to address
these aspects of the problem separately.
3. The model
Consider a service facility with random demand, such as a data
center or a computerized diagnostic imaging center. We model
the service center as an M/M/1 queuing system, where the arrivals
of service requests are independent and follow a Poisson process
with a rate of requests per unit time, and the service time has an
exponential distribution with a service rate . The rm is responsible
for investing in capacity, and the hired manager is responsible for
managing the center's daily activities and exerting additional effort
to boost the center's service demand.
Let be the expected local base demand, which is the center's
expected service demand affected by pricing, general advertising,
and other decisions made by the rm but without the inuence
of the manager's demand-generating effort. Different from prior
researches [6,7,18], here service demand can be inuenced by the
manager's marketing and service effort , where R+. In
particular, = + (), i.e., when the manager invests in marketing
and service effort , the expected demand increases by (), which
measures managerial effectiveness. Thus the center's service demand
N is a random variable with an expected value , which is a function
of the base demand and the manager's effort .
Because the manager is usually responsible for managing the daily
operations of the service center, prior research that utilizes a queuing
model in the principal-agent framework only considers how the
agent's effort affects service rate [1] or product availability [13] and
often ignores the fact that the manager can inuence service demand
by exerting marketing and service effort. For instance, the manager
can ease patient anxiety and provide patients with a pleasant service
environment in a digital imaging center or provide additional value-
added features in a data center. Clearly, such service efforts as well
as the manager's direct marketing efforts can help the center attract
additional demand. Here, we focus on the manager's responsibility
for generating demand.
Let r be the revenue from processing a service request. We assume
that the marginal cost of processing a request is constant, and it
is normalized to zero. Because of demand uncertainty, stochastic
arrivals of service requests, and random service times, the service
center often experience queuing delays. Let W be the expected time
in the system for a service request. Here, W includes waiting time in
queue and service time. In order to compete in the market, the service
center has to meet a service standard: W W0, which means that the
center promises to its clients that a service request's expected time in
the system is no longer than W0. We can think of W0 as an industry-
wide service standard. Here, the rm's capacity investment decision
directly affects the center's service quality, measured by the user
waiting time W. For the service industry, long waiting times reduce
users' satisfaction and negatively affect a center's reputation and its
future referrals. Thus to ensure the center's long-term protability
and competitiveness in the market, the rm needs to make a corre-
sponding capacity investment so that the center will meet the service
standard. Let c be the marginal capacity cost. We assume a linear
capacity cost function in the form of c0 + c. The linear capacity cost
function may not capture the exact cost structure of service facilities,
which often have a stepwise cost feature, but it is a reasonable
model for approximating IT costs and costs of other service centers;
see [6,18,24,26,28], among others.
We assume that the managerial effectiveness is linear (() =
k), since any nonlinear increasing managerial effectiveness function
can be renormalized to be linear [3]. In addition, the expected local
34 Y. Jiang, A. Seidmann / Decision Support Systems 58 (2014) 3142
base demand is a random variable, which could be h and l (h > l)
with probability q and 1 q, respectively. This setting better captures
the characteristics of many modern service centers, and it creates two
management challenges for the rm.
First, the manager running the center often has better knowledge
about the realized local base demand , whereas the rm only
observes the distribution of . Thus the rm would like to solicit market
information from the manager in order to make the proper capacity
investment to meet the service requirement. However, the manager,
with his own agenda, may not truthfully share the market information
with the rm.
Second, the rm may be able to observe advertisement spending
and some marketing and service activities conducted by the manager,
but it cannot observe or verify the exact level of the manager's
demand-generating effort. While the rm would like the manager
to exert the optimal managerial effort to maximize its net benet,
the manager is effort-averse and incurs a disutility V() = 2/2 for
an effort level . Because given the same compensation the manager
always prefers to exert less effort (i.e., shirking), the rm has to
design proper incentive contracts to induce the manager to reveal the
true market demand information and exert the desired level of effort,
and the compensation contracts should be based on observable
variables such as the number of service requests processed.
Following the general principal-agent adverse selection model
[2,8,9,11,22], we also assume that both the principal and the agent
are risk neutral. Let s be the compensation contract and s(N) be the
compensation to the manager at N service requests. In addition, let
M be the manager's net utility from an outside alternative. Thus, the
manager will choose an optimal effort level to maximize his net
utility given by his expected compensation minus his disutility of
effort EN[s(N)] V(), and he will accept the contract s only when
his expected net utility is no less than M. Here, because the manager
is protected by limited liability, the agent utility assumption is
economically similar to risk aversion [22]. If the rm chooses to
compensate the manager M, then it is only paying the manager for
managing the daily operations of the center without seeking his
demand-generating effort and local market knowledge. Introducing
the marketing aspect of the agent into the model enables us to better
understand the impact of contract design in a service environment in
which the manager has an information advantage and can inuence
market demand. Next, we analyze the model to provide an integrated
solution of operation resource management and compensation contract
design for rms dealing with market uncertainty and with combined
moral hazard and adverse selection issues.
4. Full-information case
Suppose for now the rm can perfectly observe the expected local
base demand and the manager's effort level . Hence the rm will
determine the optimal capacity , the manager's effort , and the
compensation to the manager s to maximize its net benet given by
revenue minus costs of capacity and compensation to the manager,
subject to the service standard (SS) constraint and the manager's
individual rationality (IR) constraint. Suppose the realized expected
local base demand is i (i=l, h); then the rm's problem is characterized
in Eq. (1):
i max i rc i si 1 contracts that reward the manager based on the realized demand, the
i ; i ;si
subject to : W i W 0 i l; h SS
si V i M: IR
We exclude the xed capacity cost term (c0) since only the
variable capacity cost term is relevant for the capacity decision.
Because the rm can verify exactly the manager's effort level, it will
design the contract that compensates the manager only when he
exerts the desired effort level. Given the optimal contract, the (IR)
constraint will be binding, since the rm will not offer the manager
more than what is required to make him do the job. In this case, the
compensation si will be a xed amount (M + V(i)), independent of
the number of requests processed. That means by paying the manager
a xed amount the rm is bearing all of the market uncertainty risk,
since the market demand N and hence revenue are random variables.
We know that for the M/M/1 queue in steady state the expected time
in the system is given by Wi = 1/(i i). Because capacity is costly,
the rm will also only invest in the minimum capacity required to
satisfy the service standard. Thus, the service standard constraint
will be binding as well (i.e., i = i + 1/W0).
Taking the two binding constraints back into Eq. (1), the rm will
determine the optimal effort level to maximize Eq. (2):
i max i k i rci k i 1=W 0 V i M : 2
i
Since Eq. (2) is concave in effort, solving the rst-order condition
(FOC), we have the optimal effort i = (r c)k, which is independent
of the market condition i. Given the optimal effort level, we can solve
for the optimal capacity through the binding service constraint and
the optimal compensation through the binding (IR) constraint. If
the manager exerts the optimal effort level (r c)k, the rm will
compensate the manager a xed amount sl = sh = M + (r c) 2 k 2,
the second term of which is the manager's disutility of effort V(i).
The rm can enforce this contract by penalizing the manager if he
does not exert the desired effort level, and the manager will accept
the offer. The optimal capacity is given by i + (r c)k2 + 1/W0, which
varies for different market conditions. In the full-information case,
with the optimal contract, effort, and capacity, the rm's net prot can
be written as i =ri ci si = (r c)i c/W0 M + (r c) 2 k 2.
The rst three terms (r c)i c/W0 M represent the rm's net prot
when it does not induce the manager's demand-generating effort.
The last term (rc)2 k2 represents the additional prot for the rm
(net of the compensation to the manager) from inducing the manager's
demand-generating effort. We assume that the parameters satisfy
the condition (rc)l c/W0 M>0, which means that the rm
can make a prot even without the manager's demand-generating input.
The result of the full-information case is also called the rst-best
result, which is often used as a benchmark for the case of information
asymmetry. Next, we consider the case in which the manager privately
observes the local market condition and his own effort level .
5. Information asymmetry case
In reality, the rm often cannot directly verify the manager's effort
level, which is known only to the manager. In addition, the manager
often observes the realized local base demand, whereas the rm only
knows its distribution. Both parties can observe the realized demand
N. In this section, we analyze three contracts: a uniform contract, a
variable-rate contract, and a charge-back contract. We focus on linear
contracts in the form of a xed payment (or charge) plus a per-unit
payment term. We choose linear contracts for their simplicity and
popularity in practice and also because a restriction to linear contracts
is standard in the literature [4,22]. We show that by designing incentive
rm can solicit true market information and induce a desired level of
demand-generating effort from the manager.
5.1. Uniform contract
We rst consider a uniform contract; that is, the rm offers a
single contract sU(N)=fU +bUN regardless of the realized market condi-
tion. This contract does not seek market information from the manager,
 Y. Jiang, A. Seidmann / Decision Support Systems 58 (2014) 3142
but it can be used to induce the manager to exert demand-generating
effort. While the full-information case offers the benchmark with the
highest prot that can be achieved, this uniform contract offers another
benchmark, with the prot's lower bound. It can be used to evaluate
the value of incentive contracts that solicit market information from the
manager.
The rm's problem is captured in Eq. (3):
h

 i
U U U U
max q rh b h 1q rl b l c f 3 base demand), the capacity U is close to hF, the full-information capacity
f U ;bU ; U
subject to : W i W 0 i l; h SS
U U
f b i V i M IR
h
i
U U
i arg max f b i V i : IC
i
Here, in addition to the service standard and the manager's (IR)
constraints, the rm also needs to consider the manager's incentive
compatibility (IC) constraint, since the manager's effort level is his
private information. The (IC) constraint indicates that the manager
will exert an effort level i = b Uk, which maximizes his net utility.
Thus, the expected demand satises h = h + k 2b U > l = l + k 2b U.
Since the uniform contract does not seek market information from
the manager, the rm has to consider demand uncertainty when
making its capacity decision. In fact, the rm will invest in a capacity
level such that the service standard constraint is binding in the
high-demand market. Hence the service standard constraint is not
binding in the low-demand market, which means that if the realized
local market demand is low the center has more capacity than what
is required to satisfy the service standard. In addition, the (IR)
constraint is not binding in the high-demand case, which means that
the rm compensates the manager more than what he can get from
an alternative job.
Proposition 1. The optimal uniform contract is given by bU = max{0,
r c q(h l)/k 2} and fU = M bUl (bU) 2k2/2; under this contract
the manager will exert an effort level l = h = bUk, and the rm will
install a capacity level U = h + bUk 2 + 1/W0.
When this uniform contract is offered, the manager exerts the
same level of demand-generating effort in both markets since the
per-unit incentive is the same. However, the expected compensation
differs. Specically, when the expected local base demand is h, the
manager gets paid (h l)bU more. It is also possible that the rm
optimally sets b U =0 and only offers a xed payment M when the
market uncertainty (h l) is high, in which case it is only compensating
the manager for directing the center's daily activities without seeking
his demand-generating effort. Next, we limit to the case of bU >0,
i.e., assuming k2(rc)>q(h l).
Corollary 1. The per-unit rate in the uniform contract is decreasing in
q (the probability of having a high expected local base demand), the
xed term in the contract is increasing in q, and the capacity under
the uniform contract is decreasing in q.
When q is low, it is less likely that there will be a high expected local
base demand. However, the center still needs to invest in enough capacity
to satisfy the service standard should the realized local base demand be
high. Thus, if the realized local base demand is low, with ample capacity
on hand the rm would like the manager to work hard to generate addi-
tional demand, and it induces more effort from the manager by offering a
higher per-unit payment. However, if the realized demand is high, the
rm will overpay the manager by (h l)bU. When the probability of
having a high expected local base demand increases (i.e., q increases),
naturally, the rm should increase its capacity investment. However,
35
since the uniform contract only induces demand-generating effort from
the manager, the rm actually reduces capacity investment when q in-
creases. This is because the expected demand i decreases as the rm re-
duces the per-unit payment bU to reduce the overpayment to the
manager. In fact, when q approaches 1 (the center most likely will have
a high expected local base demand), the capacity U is close to lF, the
full-information capacity level in the low-demand market, and when q
approaches 0 (the center most likely will have a low expected local
level in the high-demand market, where the superscripts represent full
information (F) and the uniform contract (U). That is, the uniform contract
severely distorts the center's capacity investment decision. Comparing with
the full-information case, we nd the following relationships: lF b U b hF,
iU b iF, and iU b iF (i=l, h). In general, the uniform contract induces less
demand-generating effort from the manager, and as a result, the center
processes fewer service requests and realizes less prot.
The value of information is thus clear. First, precise market
information leads to an accurate capacity decision. The rm can
avoid excess capacity costs while still satisfying the service standard
in the low-demand market and process more service requests in the
high-demand market with a higher capacity level. Second, having
the ability to verify the manager's exact effort level enables the
rm to induce a desired level of demand-generating effort from the
manager without overcompensating him and not to be constrained
by the manager's (IC) constraint.
The uniform contract can motivate the manager to invest in
demand-generating effort (subject to his (IC) and (IR) constraints),
but it fails to seek market information from the manager. For the
rm, a common solution for the adverse selection problem is to
offer a menu of contracts to the manager; each contract is designed
for an expected local base demand i. By observing the contract
selected by the manager, the rm is able to infer the true market
information and invest in an appropriate capacity level. Next, we
analyze two incentive contracts that can successfully solicit market
information from the manager; we continue to focus on the simple
linear contract form.
5.2. Variable-rate contract
Instead of offering a uniform contract, the rm can design a menu
of contracts, again in the form of a xed payment plus a commission
term: si(N) = fi + biN. Note that the xed payment term can be nega-
tive, which indicates a charge for the manager. The decision timeline
is as follows: First, the rm announces a set of contracts; each
contract (si(N)) corresponds to an expected local base demand (i).
We call this a variable-rate contract because generally the unit rate
bl bh. Second, the manager observes the realized local base demand
and reports to the rm (or selects a contract). Note that the manager
may or may not report truthfully to the rm. Third, the rm, based
on the manager's report (or the contract selected by the manager),
invests in capacity. Fourth, the manager exerts demand-generating
effort. Lastly, the nal demand is realized, and both parties observe
the realized demand N. The rm compensates the manager according
to the selected contract, and it incurs a penalty if the service standard
is not met. Since the demand N is a random variable, the rm cannot
infer the value of even after learning the expected local base
demand and observing the realized value of N. The challenge for
the rm is to design a menu of contracts so as to solicit true market
information and induce the desired level of effort from the manager
and thereby maximize its net benet.
Following the revelation principle [16], we only need to search for the
optimal menu of contracts from the class of truth-telling contracts. Under
such contracts, it is in the manager's interest to truthfully reveal the
expected local base demand information to the rm. Assume the cost
of not satisfying the service standard is high such that the rm always in-
vests in sufcient capacity to meet the service standard. The rm
36 Y. Jiang, A. Seidmann / Decision Support Systems 58 (2014) 3142
therefore needs to solve Eq. (4) for the optimal contracts and capacities,
subject to a series of constraints:
max 1ql rc l f l bl l qh rc h f h bh h ; i l:h: local market information from the manager so that the rm can make
i ;f i ;bi
First, in order for the manager to accept it, the contract si(N) (i =l, h)
has to satisfy the manager's individual rationality (IR) constraint:
EN si NV i M; IR
where EN[si(N)]= fi + bii. Second, because the rm cannot observe and
verify the manager's effort level, the manager will choose the effort
level that maximizes his own net utility. This gives the manager's incen-
tive compatibility (IC) constraint:


 contract is decreasing and concave in q (the probability of having a
i arg max EN si NV i : IC
0
i
These constraints are for the case in which the manager reports
truthfully about the realized market condition to the rm. In order
for the manager to report truthfully, the optimal contracts also need
to satisfy the following two incentive compatibility constraints:
EN sl N=l V l EN sh N=l V l IC  LH mation asymmetry the rm has to pay a price to obtain this information.
EN sh N=h V h EN sl N=h V h ; IC  HL the high-demand market. This information rent, which comes as the
where i refers to the optimal effort level when the manager observes
i but reports j (i, j = l, h and i j). The left-hand sides of these two
constraints represent the manager's net utility if he truthfully reports
the realized local market condition, and the right-hand sides represent
the manager's net utility if he misreports the market condition. These
two constraints ensure that it is in the manager's interest to report
truthfully to the rm about the realized local market condition. In addi-
tion, the rm also needs to choose the optimal capacity to satisfy the
service standard constraints: Wi W0 (i =l, h).
Proposition 2. Following the decision timeline outlined above, this
combined adverse selection and moral hazard problem has a unique
solution of the optimal capacity and the menu of variable-rate
contracts that solicit truthful market information from the manager:
(i) the optimal menu of contracts are given by sl(N) = fl + blN with
fl = M bll bl2k 2/2 and bl = max{0, r c q(h l)/(k 2(1 q))}
and sh(N) = fh + bhN with fh = M hbh bh2k 2/2 + bl(h l) and
bh = r c; under these contracts the manager will exert the optimal
efforts h = kbh and l = kbl; (ii) the optimal capacity levels are
given by h = h + bhk 2 + 1/W0 and l = l + blk 2 + 1/W0.
Comparing the contract terms for different market conditions,
we see that the rm offers a higher xed payment (fl > fh) in the
low-demand market so that the manager will not misreport. In the
high-demand market, to counter the manager's misreporting tenden-
cy since fl > fh, the rm offers a higher per-unit payment (bh > bl) in
the contract. The proposed contracts in Proposition 2 ensure that
the manager will accept the contracts and truthfully reveal the real-
ized local market condition to the rm. When the expected local
base demand is h, the contract sh(N) can in fact induce the same
level of effort as that in the full-information case, and the rm will
accordingly invest in the same level of capacity. However, because
of information asymmetry the rm has to compensate the manager
more than in the full-information case in order to elicit his market
information. Thus, the rm realizes less prot when the expected local
base demand is h. On the other hand, when the expected base demand
is l, the contract sl(N) provides a lower incentive (bl b rc) and therefore
induces less effort than in the full-information case. Accordingly, the rm
invests less in capacity, provides less compensation to the manager, and
in the end realizes less prot.
Here, the incentive contract serves two purposes: (1) it elicits true
the correct capacity investment decision; (2) it induces demand-
4
generating effort from the manager so that the rm can realize a
higher prot. The rm accomplishes the rst purpose by adjusting
both the xed term (fi) and the per-unit rate (bi) in the contract so
that it is in the manager's interest to reveal market information. The
rm accomplishes the second purpose mostly by offering the manager
a per-unit payment bi, which motivates the manager to exert effort
i = bik to increase the demand for the center's services, and the rm
shares the prots generated by the manager's effort with the manager
through the xed term in the contract (the bi2k 2/2 term in fi).
Corollary 2. i) The per-unit rate bl (when bl > 0) in the variable-rate
high expected local base demand), the xed term fl in the contract
is increasing in q, and the capacity l is decreasing and concave in q.
ii) The per-unit rate bh and capacity h are independent of q, but the
xed term fh in the contract is decreasing in q.
By offering different contract terms for different market conditions,
the rm is able to elicit true market information from the manager
and make corresponding capacity decisions. However, because of infor-
In this case, the manager is able to keep an information rent in
overpayment term bl(h l), is increasing in (h l) and bl. When q
increases, the probability of having a high expected local base demand
increases, and so does the chance of overpaying the manager. Thus, to
counter the overpayment effect, the rm will reduce the per-unit rate
bl and accordingly increase the xed payment term fl. As a result, the
manager incurs less effort when the expected local base demand is
low, the center's expected demand decreases, and the rm invests
less in capacity. When (h l) is sufciently large, the benet from
inducing the manager's demand-generating effort in the low-demand
market cannot offset the overpayment to the manager in the high-
demand market. Thus, it is optimal to set bl = 0; i.e., the rm induces
no demand-generating effort in the low-demand market. As a result,
it does not have to over-compensate the manager and is still able to
elicit true market information and induce the rst-best effort level in
the high-demand market. Thus, when bl =0, the rm can achieve the
same prot level as that in the full-information case in the high-
demand market.
To summarize, with the proposed variable-rate linear contracts in
Proposition 2 the rm is able to elicit true market information from the
manager, but this comes at a cost because of information asymmetry.
Compared with the full-information case, the rm will realize less
expected prot, even though the expected compensation to the manager
is lower (since (1q)(fl +bll)+q(fh +bhh)b M+(rc)2 k2 =full-
information compensation). The manager, on the other hand, realizes a
higher expected net benet, since in the high-demand market his net
utility satises fh +bhh (rc)2 k2 >M.
5.3. Charge-back contract
We now consider an alternative contracting approach that not
only elicits true market information but also induces the rst-best
effort level from the manager. Suppose the rm offers a uniform
per-unit rate to the manager and delegates the capacity investment
decision to the manager by letting him select a capacity level. In re-
turn, it charges the manager a xed fee for using capacity and sharing
prot, and this xed charge term is like a franchise fee. The decision
timeline is as follows: First, the rm announces a menu of contracts;
each contract corresponds to a capacity level i and includes a
 Y. Jiang, A. Seidmann / Decision Support Systems 58 (2014) 3142
uniform per-unit payment to the manager (p), a xed charge (Fi), and
a penalty term (Ti) if the service standard Wi W0 is not met. In the
second stage, the manager observes the realized local market condi-
tion and asks the rm for a specic capacity level; i.e., the manager
selects a contract. Here, the manager is subjected to the service stan-
dard constraint directly, and he is penalized if the service standard is
not met. Third, the rm delivers the capacity level selected by the
manager. Fourth, given the capacity provided by the rm, the manag-
er then invests in demand-generating effort. Fifth, the nal demand is
realized. The manager is compensated at the given unit rate, and he
also pays the rm the pre-specied xed amount for using capacity,
and possibly a penalty. This is equivalent to delegating all the decision
rights to the manager, and the rm then can subtract a fee to extract
all surpluses.
Suppose for now that the rm sets a penalty term such that the
manager will always choose to satisfy the service standard. We can
then omit the penalty term in both parties' objective functions. The
rm's prot function is given by Fi ci + (r p)i. Its problem is to
determine the contract terms, which include the uniform per-unit
payment rate (p) and the capacity charge (Fi). The rm will select
the contract terms {p, Fi} to ensure that the manager will participate,
report truthfully regarding the local market conditions, and exert
the rst-best effort level. The manager's problem is to determine
the optimal effort level and capacity choice to maximize his net
utility, given by pi V(i) Fi, subject to the service standard con-
straint. As the manager maximizes his net utility, clearly, he prefers
the optimal effort level i = pk (derived from the FOC). Thus, the
rm can set p = r c, which equals the per-unit prot, to induce
the rst-best effort level from the manager. The rm can then
extract all surpluses while ensuring that the manager will partic-
ipate (i.e., pi V(i) Fi = M) by setting Fi = pi V(i) M. Given
that the manager exerts the optimal effort level i = pk, and suppose
for now that he also reports truthfully, the optimal capacity choices
should be l = l + k 2p + 1/W0 and h = h + k 2p + 1/W0. Thus, the rm
can offer the capacity and contract set {p, Fi, i} from which the manager
will choose.
We next check whether this proposed menu of contracts {p, Fi, i}
can induce the manager to reveal market information truthfully and
invest in the rst-best effort level. First, suppose the expected local
base demand is l. Even at a higher capacity level, the manager still
invests in the optimal effort level l = pk, since effort is costly. Because
the capacity charge term satises Fh > Fl, the manager will truthfully
reveal his private information about the realized local market demand
l, ask for capacity l, and pay Fl. In this case, the service standard can
be met even with the penalty term Tl = 0.
However, if the realized expected local base demand is h, with a
zero penalty term the manager would prefer the contract {p, Fl, l}
over {p, Fh, h}, since the former charge for capacity is lower. In that
case, the manager would ask for capacity l and still prefer to invest
in the same effort level = pk, and the center cannot meet the service
standard, since the capacity level is too low. One quick x is to include
a high penalty term in the contract. To prevent the manager from
choosing l when he observes h, the rm can set a high enough
penalty term Tl, which leads to a negative net utility for the manager
if he does not comply with the service standard. This, however, still
does not solve the problem, because given l, the manager can either
invest in no managerial effort or invest less in managerial effort (
with h k ^ h l such that the center meets the service constraint,
assuming k 2(r c) > (h l), which indicates h b l.
First, we rule out the case in which misreporting and incurring no
managerial effort are a viable choice. We assume the parameters
satisfy k 2(r c) > 2(h l), which implies that if the manager
observes h but asks for capacity l, without incurring demand-
generating effort his net utility is less than M, since ph Fl =
M + (r c)(h l) k 2(r c) 2/2 b M. The manager still has a sec-
ond option, which is to incur effort ^ h such that h k ^ h l . By
37
doing so the service standard is met when he observes h but asks
for capacity l, and he still can realize a higher net utility
^ h F l > pl V l F l M, since V
pl V ^ h bV l . To prevent
the manager from shirking and under-requesting capacity, the
rm has to adjust the xed charge term to induce him to share
market information truthfully and exert the rst-best effort level.
Proposition 3. When the parameters satisfy k 2(r c) > 2(h l), the
following menu of contracts with a uniform unit rate plus a capacity
charge and a service standard penalty term will solicit true market in-
formation and induce the rst-best effort levels from the manager:
The rm offers a uniform per-unit rate p=rc to the manager and in
return charges the manager Fl =(rc)l +(rc)2k2/2M for capacity
level l =l +k2(rc)+1/W0 and Fh =Fl +(h l)2/(2 k2) for capacity
level h =h +k2(rc)+1/W0. In addition, there is a penalty term at-
tached to the capacity selection i: the manager has to pay a ne Tl =
(rc)(h l)+ and Th =>0 if the service standard Wi W0 is not
met.
Given the proposed charge-back contracts in Proposition 3, the
manager nds it is in his best interest to report local market conditions
truthfully and exert the rst-best effort level. This charge-back contract
is equivalent to a franchise contract with Fi as the franchise fee and r p
as the royalty rate. That is, in order to run the center and retain the
center's revenue, the manager has to pay the rm the xed franchise
fee and a per-unit royalty (c). Similar to the variable-rate contracts
described in Section 5.2, when there is market information asymmetry,
the manager is able to reserve an information rent for the high-demand
market under the proposed charge-back contracts. The two contracts
both have a per-unit payment and a xed term, but the variable-rate
contract offers different per-unit rates for different markets, and the
xed term could be a payment, whereas it is always a charge in the
charge-back contract. Also, the manager is not explicitly subject to
the service standard constraint and the rm makes the capacity deci-
sion under the variable-rate contract. In addition, although both kinds
of contract can induce the manager to reveal the true local market
condition, the charge-back contracts can induce the rst-best effort
levels in both markets, whereas the variable-rate contracts only induce
the rst-best effort in the high-demand market. The two contracts also
differ in that the variable-rate contract terms and capacity are a function
of the probability distribution of the expected local base demand, but
the charge-back contract terms and capacity are independent of the
probability distribution of the expected local base demand.
6. Contract analysis
In this section, we rst illustrate the operational and nancial out-
comes of the center in a simple numerical example. We then analyze
the results under the three contracts with information asymmetry,
recommend the contract choice for the rm, and evaluate the cost and
value of the manager's private information.
6.1. Numerical example
Consider a service center such as a computerized medical imaging
center characterized by the following parameter values: r = $100, c =
^ h) $30, W0 = 0.1, M = $100, k = 0.7, l = 30, h = 45, and q = 0.6. Table 1
summarizes the optimal contracts, the manager's effort, the center's
operational performance, the costs of capacity, and the center's
prots in the full-information case and the information asymmetry
case with uniform, variable-rate, and charge-back contracts. The
example in Table 1 illustrates that the manager's market knowledge
is valuable for the rm, since both contracts that solicit the manager's
market information outperform the uniform contract in the informa-
tion asymmetry case. Even with information asymmetry, the charge-
back contract is able to induce effort levels identical to those in the
38 Y. Jiang, A. Seidmann / Decision Support Systems 58 (2014) 3142
full-information case, and it therefore generates the same volume
levels and the rm is able to invest in the same capacity levels as in
the full-information case. However, the rm realizes less prot
when there is information asymmetry, even with the charge-back con-
tract, mostly because the manager is able to keep an information rent.
Moreover, while the charge-back contract seems to be superior to the
variable-rate contract in terms of the center's operational outcome,
the variable-rate contract in this case outperforms the charge-back
contract because it generates more prots for the rm.
Under the charge-back contract, the per-unit payment to the
manager equals the center's per-unit prot margin (r c). Hence
the manager is equivalently reserving all the revenues of the center
by paying the variable cost of the consumed capacity (ci out of ci)
and the xed charge term (or franchise fee), which is naturally associat-
ed with the xed cost of operations, with a higher charge for a higher
capacity installation. In this numerical example, the franchise fee is
much higher than the cost of capacity buffer (c/W0 = ci ci), and,
combining that with the variable charges, it means the manager is
paying much more than the total variable cost of capacity. One may
wonder why the manager would need the rm. This can be explained
by examining what each party contributes to the business. In this
service setting, the manager and the rm each brings two unique assets
to the operation. The former possesses knowledge of local market
demand and can help the center increase service demand by managing
day-to-day operations and exerting demand-generating efforts, and the
latter provides capital for capacity and brings in base demand through
its brand name, reputation, and promotional efforts. Thus to reserve
the whole revenues of the center the manager not only needs to pay
the rm for capacity but also for the base demand brought by the rm's
reputation and effort. The charge-back contract is designed to motivate
the manager to share his assets and help the rm recoup the values of
assets that it contributes to the center and extract all surpluses. Paying
the manager the full prot margin per unit effectively motivates him
to exert the rst-best demand-generating effort. The rm's prot
is driven by the xed franchise fee only. This xed charge serves two
purposes: it elicits market information from the manager and recoups
reputation value and extracts all surpluses for the rm. Because part
of the center's demand (local base demand) is generated by the rm's
reputation and effort, not by the manager, the latter has to pay a
reputation premium (the (r c)i term) through the franchise fee.
The analysis of the charge-back contract conrms the positive
theory of cost allocation proposed by [28] that cost allocations are
often used to solve certain organizational problems, here, in particu-
lar, incentive and information asymmetry agency issues. It is also
consistent with the franchise contract literature that the franchising
fee structure is driven by factors such as moral hazard, franchisor
services, and risk sharing between the two parties [17,23]. We did
not model the double-sided moral hazard issue that commonly
appears in the franchising network, but we derive the optimal incentive
Table 1
Summary of contract terms and corresponding operational/nancial results of the center.
Full-information contract
Optimal contracts sl = sh = $1301
Manager's effort l = h = 49 l = h = 36.14
Center's expected l = 64.3
volume (patients/day) h = 79.3
Capacity (patients/day) l = 74.3
h = 89.3
Capacity utilization l = 86.5%
h = 88.8%
Costs of capacity cl = $2229
ch = $2679
Center's prots $3530
contract for the rm facing an effort-averse manager who also pos-
sesses private information regarding local market conditions. Unlike
[21], which shows that a franchising plan achieves the rst-best
outcome in terms of agents' effort and the principal's prot, we nd
that a similar franchising (charge-back) contract can induce the
rst-best effort and truthful local market information from the manager
and lead to the rst-best capacity investment, but it fails to produce the
rst-best prot and may not even produce the highest prot for the rm
in a service environment with a combined moral hazard and adverse
selection problem.
6.2. Contract choice
We now compare the variable-rate and charge-back contracts to
determine the optimal choice for the rm. While both can solicit
true market information from the manager, only the charge-back
contract can induce the rst-best effort levels from the manager
under both market conditions. When the expected local base demand
is low, the variable-rate contract offers a lower per-unit payment than
the charge-back contract, so the manager invests less in demand-
generating effort than that under the charge-back contract. The xed
payment fl in the variable-rate contract can be positive or negative
depending on parameter values. Since paying the manager the xed
payment fl is equivalent to charging the manager fl, we can compare
the xed charge fl with Fl. It is straightforward to show that Fl > fl
when bl > 0. The expected prots have the following relationship:
lF = lC > lV, where the superscript represents full information (F),
charge-back contract (C), and variable-rate contract (V).
When the expected local base demand is high, both contracts offer
the same per-unit payment (bh = p = r c) and can induce the
rst-best effort level from the manager. Comparing the xed term
in the contract, we nd that Fh > fh when q b 1/3. Thus, the expected
prots have the following relationship: hF > hC hV when q 1/3 and
hF > hV > hC when q > 1/3. That is, which contract generates higher
expected prots depends on the probability distribution of the local
base demand.
Proposition 4. As only the manager can privately observe the
realized expected local base demand and his own demand-generating
effort, and the rm only knows the distribution of the former, when
the parameters satisfy k 2(r c) >2(h l): (1) it is optimal for the
rm to offer the charge-back contract when the probability of having
a high expected local base demand satises q b ; (2) it is optimal for
the rm to offer the variable-rate contract when q > .
Recall that the parameter condition in Proposition 4 ensures that
misreporting when observing h and incurring no demand-generating
effort are not a viable choice for the manager. Proposition 4 can be de-
rived by comparing the rm's prot under the two contracts. In general,
Information asymmetry
Uniform contract Variable-rate contract Charge-back contract
s(N) = $2102 + $51.63 N sl(N) = $765 + $24.1 N sl(N) = $3201 + $70 N
sh(N) = $3889 + $70 N sh(N) = $3430 + $70 N
l = 16.86; h = 49 l = h = 49
l = 55.3 l = 41.8 l = 64.3
h = 70.3 h = 79.3 h = 79.3
= 80.3 l = 51.8 l = 74.3
h = 89.3 h = 89.3
l = 68.9% l = 80.7% l = 86.5%
h = 87.5% h = 88.8% h = 88.8%
c = $2409 cl = $1554 cl = $2229
ch = $2679 ch = $2679
$2803 $3107 $3038
 Y. Jiang, A. Seidmann / Decision Support Systems 58 (2014) 3142
the rm realizes less expected prot when the manager possesses
private information on market demand and his effort. However, the
rm can improve its prot by selecting which contract to offer based
on the probability distribution of the local base demand. Proposition 4
states that when the probability of having a low local base demand is
higher than that of a high local base demand (i.e., q b ), the rm can
realize a higher expected prot level by offering the charge-back
contract. This is mostly because the charge-back contract achieves the
same prot level as in the full-information case when the expected
local base demand is low. When the probability of having a high local
base demand is greater than , the rm is better off to offer the
variable-rate contract. This is because the variable-rate contract leads
to a higher prot level in the high-demand market, and this prot
improvement can offset the prot shortage in the low-demand market,
especially as the probability of having a low-demand market is relatively
low.
6.3. Cost and value of the manager's private information
The prot difference between the full-information and information
asymmetry cases can be used as a measure for the overall costs of infor-
mation or agency costs for the rm. Here information includes the
manager's information regarding local market and his effort. Even
though the variable-rate and charge-back contracts can elicit market
information from the manager, because of information asymmetry the
manager is able to reserve an information rent, which represents part
of the costs of information. Since the rm can improve her prot by
offering appropriate contracts based on the distribution of local base de-
mand, we can use F C as the measure for the agency costs when
q b and F V when q > .
Corollary 3. Given that the parameters satisfy k 2(r c) > 2(h l):
(1) the prot difference F C is increasing in the prot margin
(r c), the local market uncertainty (h l), and the probability of
having a high local base demand (q); (2) the prot difference F
V is increasing in (r c); it is increasing in (h l) when bl > 0
and independent of (h l) when bl = 0; it is, however, decreasing
in q when bl = 0; for bl > 0 it increases in q rst, reaches its maxi-
mum at q = q^ and then decreases in q where q^ satises
2
q^ 2q^ =1q^ 2k rc=h l .
2
Corollary 3 states that the manager's information becomes more
costly for higher prot margins (r c) or higher local market uncer-
tainty (h l). In addition, it is more costly for moderate q values,
which correspond to higher base demand variance, rather than ex-
treme q values, since as q approaches 0 (or 1) the charge-back (or
variable-rate) contract leads to a near rst-best prot level. Note
that the variance of local base demand is given by q(1 q)(h l) 2,
which is increasing in q for q b and decreasing in q for q > . Thus
it is high for moderate q values.
We compute the rm's expected prots when offering the
uniform contract, the variable-rate contract, and the charge-back
contract and compare them with the prot in the full-information
case using the numerical example from Section 6.1 but with q varying
from 0.05 to 0.95. Fig. 1 illustrates the percentage of full-information
prot that can be achieved by offering the three contracts. Clearly,
both the variable-rate and charge-back contracts outperform the uni-
form contract since the latter does not elicit market information from
the manager. Similarly, we can use the prot differences C U and
V U as a measure for the value of the manager's market informa-
tion, since the uniform contract that does not elicit such information
distorts the capacity decision. In this example, the prot difference
C U (when q b ) is bigger when q is low and the difference
V U (when q > ) is bigger when q is high. Thus, effectively
eliciting the manager's local market information through contract de-
sign is more valuable for the rm for extreme q values or when the
39
Percentage of Full-Information Profit
variable-rate contract
96% charge-back contract
uniform contract
91%
86%
81%
76%
0.15 0.05 0.25 0.35 0.45 0.55 0.65 0.75 0.85 0.95
Probability of Having a High Expected Local Base Demand (q)
Fig. 1. The percentage of the full-information prot under the uniform, variable-rate, and
charge-back contracts with q varying from 0.05 to 0.95.
variance of the local base demand is low. In this example, the rm
can realize 17.72% more prot by obtaining such information with
the charge-back contract than with the uniform contract when q=
0.05 and 24.4% more prot by offering the variable-rate contract rather
than the uniform contract when q =0.95. In comparison, the two
information-eliciting contracts only generate 10.38% more prot when
q=0.5. In addition, the prot gap between the full-information case
and the charge-back or uniform contract is increasing in q. However,
when q is higher than the bound q^ the prot gap is actually decreas-
ing under the variable-rate contract (Corollary 3). That is, the value
of the optimal contract choice is increasing, since the center's
protability is improved with the variable-rate contract when the
probability of having a high-demand market is relatively high.
7. Conclusions
Market demand uncertainty and time-based competition, make
capacity investment and managerial incentives decisions for service
facilities such as high-end diagnostic medical imaging centers, modern
IT services, or contract manufacturing shops particularly challenging.
These facilities compete on service quality, short queuing times and
speed. Therefore, having insufcient capacity can be economically
devastating for them. Given the high up-front costs involved, rms
want to make sure that they neither over- nor under-invest in service
capacity. These problems get exasperated by the fact that typically
rms are unfamiliar with the local market demand and do not closely
observe the demand-generating efforts of the hired managers. Most
of prior studies of cost allocation methodologies, contract design, and
service resource management tend to address these aspects of the prob-
lem separately. They ignore the interaction effects between the capacity
decisions and the managerial adverse selection and moral hazard issues
which are crucial elements for successfully running services with xed
capacity, random arrivals and stochastic service times. We contribute
to literature by developing an integrated approach that simulta-
neously determines the design of managerial incentive contracts
and the optimal capacity planning for capital intensive service
facilities, facing a combined moral hazard and adverse selection
agency issue, and providing a complete solution to the complicated
practical problem.
We derive the closed-form solutions for three incentive contracts
under information asymmetry between the rm and management,
and analyze their impact on capacity decisions, service levels, service
volumes, and the allocations of costs. We nd that the uniform
contract, which offers the same payment terms regardless of market
conditions, is able to induce the desired demand-generating effort
from the manager but fails to elicit the manager's market knowledge
and greatly distorts the rm's capacity decision. To effectively elicit
market information from the manager, the rm needs to design a
40 Y. Jiang, A. Seidmann / Decision Support Systems 58 (2014) 3142
menu of incentive contracts. Two contracting approaches that are
simple to implement in practice can elicit this information. The rst
is to offer a menu of variable-rate linear contracts in which the
per-request payments are modied for different market conditions.
Under this contracting mode, the manager is induced to report the re-
alized local market demand and the rm is responsible for investing
in capacity and meeting the service standard, based on the manager's
report of market information. Under the second approach, the rm
charges the manager a differential franchise fee for using the
installed capacity and pays the manager a uniform per-request
fee regardless of the market conditions. Here the manager is
explicitly responsible for capacity selection and for meeting the
service standard. We prove that the rm can elicit true market in-
formation by properly setting the xed franchise fee, and this ap-
proach (of charge-back contracting) can also induce the rst-best
(full-information) demand-generating effort from the manager.
As we demonstrate in the numerical example, the manager's mar-
ket knowledge is valuable to the rm. By eliciting this market
information from the manager with the proposed variable-rate
or charge-back contracts, the investing rm can make accurate
capacity decisions and realize 10.38% to 24.4% more prots than
offering the uniform contract.
While both the variable-rate and charge-back contracts can help the
rm successfully elicit the manager's local market information, the
rm's eventual optimal contract choice depends on the probability
distribution of the local base demand. Surprisingly, the charge-back
(franchise) contract, which produces the rst-best operational results
across the board, may not always maximize the rm's prot when the
manager also has private information about the local market. In partic-
ular, this franchise approach only dominates the variable-rate contract
when the probability of having a low local base demand is high (qb ).
This contrasts with and extends results in [21] that a franchising plan
achieves the rst-best outcome in terms of both the agents' effort and
the principal's prot. In our setting, the franchising contract not only
fails to produce the rst-best prots for the rm, it can also be inferior
to the variable-rate contract.
In our service setting, the manager and the rm each brings two
unique assets to the operation. The former possesses knowledge of
local market demand and can help the center increase service demand
by managing day-to-day operations and exerting demand-generating
efforts, and the latter provides capital for capacity and brings in base-
demand through its brand name, reputation, and promotional efforts.
The charge-back contract is designed to motivate the manager to
share his assets and help the rm recoup the values of assets that it con-
tributes to the center and extract all surpluses. Thus it may be optimal
(for both the rm and the manager) to charge the manager more than
the costs of capacity as the rm needs to recoup its reputation values
and extract all surpluses through the franchise fee. The accounting liter-
ature tends to focus on the role of xed cost allocation as an effective cost
control to hold agents accountable so that they will not over-consume re-
sources at the expense of the rm. However, in a service queueing setting,
under-consuming resources can be harmful to the rm as well, because
low capacity restricts the center from serving more customers and also
causes long customer waiting times. We show that when properly
designed, the franchise fee in the charge-back contract can prevent the
manager from both over-using and under-using service capacity.
In summary, we solve a practical capital budgeting and managerial
incentive contracting problem in a service setting and extend prior liter-
ature on cost allocation, contract design, and service resource manage-
ment. We show that by integrating service resource investment
and incentive contract design, rms can optimize their capacity in-
vestments to better cope with market uncertainty and signicantly
improve their protability. Our study applies to many capital-
intensive and congestion-prone service systems, where the success
of signicant up-front capacity investments also hinges on the
daily operations of those facilities run by hired managerswho
typically possess specic-knowledgethat give them a signicant
information advantage.
Appendix A. Derivation of the full-information contract and capacity
Following the discussion in Section 4, the constrained problem (1)
can be reduced to solving the optimal effort from Eq. (2). The rst-
order condition (FOC) and second-order condition (SOC) are given by:
i = i krc i 0 FOC
2 2
i = i 1b0: SOC
Solving the FOC, we have the optimal effort, i = k(r c). Hence
the optimal expected demand is i = i + ki = i + k 2(r c). The
optimal capacity can be derived from the binding service standard
constraint: i = i + 1/W0 = i + k 2(r c) + 1/W0. The compensation
to the manager can be derived from the binding (IR) constraint:
si = M + V(i) = M + (r c) 2 k 2.
Proof of Proposition 1. The uniform contract problem is described
in Eq. (3). Solving the FOC of the manager's net utility function ((IC)
constraint) we know that the manager will exert an effort level
i = b Uk, which maximizes his net utility. Thus, the expected demand
satises h = h + k 2b U > l = l + k 2b U. To satisfy the service standard
constraint, the rm will invest in a capacity level such that the service
standard constraint is binding in the high-demand market because
Wl b Wh. Next, we check the (IR) constraint. Since h > l and h = l,
f U + b Uh V(h) > f U + b Ul V(l) M. The (IR) constraint will be
binding for the low-demand case but not for the high-demand
case. From the binding service standard constraint in the high-
demand case and the binding (IR) constraint in the low-demand case
we have =h +1/W0 =h +k2bU +1/W0 and f U =M+V(l)bUl.
Taking i, , and f U into Eq. (3), which is now a concave function of bU,
and solving the FOC, we have the per-unit rate bU =max{0, rc
q(h l)/k2}. We then can derive the optimal effort, capacity, and xed
payment from bU. Here, depending on the parameter values, it maybe
optimal for the rm to set bU =0 and only offer a xed payment M, and
in that case it is only paying the manager for running the center without
seeking his demand-generating effort.
Proof of Corollary 1. When bU >0, bU/q = (h l)/k2 b 0, U/
q = (h l)b 0, and fU/q = (l +k 2bU)bU/q >0.
Proof of Proposition 2. The rm's problem of capacity investment
and variable-rate contract design is describe in Eq. (4) subject to a
series of constraints: (IR), truth-telling (IC), (IC-LH), (IC-HL), and ser-
vice standard constraints. First, we simplify the constraints described
in Section 5.2. Because the manager's net utility function is concave in
effort, we can replace the (IC) constraint with the FOC of the manager,
which gives the optimal effort level i = bik when the manager
reports truthfully. Similarly, i bj k (i, j = l, h and i j). We then
can simplify the (IR), (IC-LH) and (IC-HL) constraints:
2 2
f l bl l bl k =2M IR  L
2 2
f h bh h bh k =2M IR  H
2 2 2 2
f l bl l bl k =2f h bh l bh k =2 IC  LH
2 2 2 2
f h bh h bh k =2f l bl h bl k =2: IC  HL
Further comparison of the (IC-HL) and (IR-L) constraints shows
b2 k2 b2 k2
that f h bh h h2 > f l bl l l2 M. Thus, the (IR-L) constraint
should be binding, and we can omit the (IR-H) constraint. The
 Y. Jiang, A. Seidmann / Decision Support Systems 58 (2014) 3142
(IC-HL) constraint should be binding as well because the rm
does not need to compensate the manager more than what is re-
quired. Next, from the binding (IC-HL) constraint we can show
that the (IC-LH) constraint can be further simplied as bh bl.
Because capacity is costly, the service standard constraints will
be binding so the optimal capacity is given by i = i + 1/W0 =
i + bik 2 + 1/W0.
The rm's problem can now be reduced to solving:
n h
 i h

max
2 2
1q rcbl l bl k f l c=W 0 q rcbh h bh k f h c=W 0
f l ;bl ;f h ;bh
2 2
subject to : f l bl l bl k =2 M
2 2
f h bh h bh k =2 f l bl h bl k =2
2 2
bh bl 0:
From the binding (IR-L) and (IC-HL) constraints, we can represent fl as
Mbll +bl2k2/2 and fh as fl +blh +bl2k2/2bhh bh2k2/2. Taking these
fl and fh into Eq. (4), we can then derive bl and bh by solving the corre-
sponding FOCs. From bl and bh, we then can solve for the optimal effort,
capacity, and xed payment terms.
Proof of Corollary 2. When bl > 0, bl/ q = (h l)/k 2(1 q) 2 b 0,
l/ q = (h l)/(1 q) 2 b 0, and fl/ q = (l + k 2bl) bl/ q > 0.
In the high-demand case, bh and h are independent of q, but fh/
q = (h l) bl/ q b 0.
Proof of Proposition 3. Following the corresponding analysis in
Section 5.3, since the manager will choose the optimal effort level
i = pk, the rm can set p = r c to induce the rst-best effort level
from the manager. Accordingly, the rm will offer capacity choices
i = i + k 2(r c) + 1/W0. In addition, the rm will set a franchise
fee Fl = (r c)l + (r c) 2k 2/2 M such that the manager's (IR) con-
straint is binding in the low-demand case.
As discussed in Section 5.3, when the parameters satisfy
k 2 (r c) > 2( h l ) if the manager observes h but asks for ca-
pacity l, without incurring additional demand-generating effort his net
utility is less than M. But since the manager can incur an effort ^ h such
that h k ^ h l , he can realize a higher net utility if he misreports
when observing h. In order to induce true market information from
the manager, the rm has to offer the manager more incentives through
the franchise fee term Fh. In particular, the rm allows the manager to
keep an information rent (h l)(r c (h l)/2k 2) by charging
him Fh = (r c)h + (r c) 2k 2/2 M (h l)(r c (h l)/2k 2)
for using capacity h.
Following the discussion in Section 5, we assume the parameters
satisfy k2(rc)>2(h l). Given this assumption, the unit rate bl
under the variable-rate contract satises bl >0 for q2/3. For 2/3b qb 1,
the rm may set bl =0 if the parameters satisfy k2(rc)b q(h l)/
(1q). We prove the next proposition and corollary for both bl >0 and
bl =0.
Proof of Proposition 4. When bl > 0 (corresponding to the parame-
ter condition k 2(r c) > max{2(h l), q(h l)/(1 q)}), a prot
comparison between the variable-rate and charge-back contract
shows that C V = q(h l) 2(1 2q)/2k 2(1 q), which is nonneg-
ative if q and negative otherwise. Thus, it is optimal to offer the
charge-back contract when q and the variable-rate contract other-
wise. When bl =0 (corresponding to the parameter condition 2(h
l)bk2(rc)bq(h l)/(1q)), we have C V =(1q)(rc)2k2/
2q(h l)(rc(h l)/2k2), which can be shown to be negative
41
from the parameter condition for bl =0. Hence in general whether
bl >0 or bl =0 we have C b V if q>.
Proof of Corollary 3. The prot difference between the full-
information contract and the charge-back contract in the information
asymmetry case is given by F C = q(h l)(r c (h l)/2k2).
For bl >0, the prot difference between the full-information contract
and the variable-rate contract in the information asymmetry case is
given by F V =q(h l)(rcq(h l)/2k2(1q)), and when
bl =0, we have F V =(1q)(rc)2k2/2. Accordingly, we can show
io
that for the charge-back contract (F C)/(rc)>0, (F C)/
(h l)>0, and (F C)/q>0.
4
For the variable-rate contract, ( F V)/ (r c) > 0 with bl 0;
IR  L ( F V)/ (h l) > 0 for bl > 0, and F V is independent of
(h l) for bl = 0. Furthermore, for bl > 0, 2( F V)/ q 2 b 0 and
IC  HL ( F V)/ q = (h l)(r c q(h l)(2 q)/2k 2(1 q) 2). Thus,
the prot difference is at its maximum at q = q^ , and it increases in q
IC  LH for q b q^ and decreases in q for q > q^ , where ( F V)/ q = 0 at q =
q^ . For bl = 0, we have ( F V)/ q b 0.
References
[1] S. Baiman, S. Netessine, R. Saouma, Informativeness, incentive compensation
and the choice of inventory buffer, The Accounting Review 85 (2010)
18391861.
[2] R. Balakrishnan, V.D. DeJong, The role of cost allocations in the acquisition and use of
common resources, Contemporary Accounting Research 9 (1993) 395414.
[3] A.K. Basu, R. Lal, V. Srinivasan, R. Staelin, Salesforce compensation plans: an
agency theoretic perspective, Marketing Science 4 (1985) 267291.
[4] S. Bhattacharyya, F. Lafontaine, Double-sided moral hazard and the nature of
share contracts, The Rand Journal of Economics 26 (1995) 761781.
[5] F. Chen, Salesforce incentives, market information, and production/inventory
planning, Management Science 51 (2005) 6075.
[6] S. Dewan, H. Mendelson, User delay costs and internal pricing for a service facility,
Management Science 36 (1990) 15021517.
[7] P.S. Giridharan, H. Mendelson, Free-access policy for internal networks, Information
Systems Research 5 (1994) 122.
[8] M. Harris, C.H. Kriebel, A. Raviv, Asymmetric information, incentives and
intrarm resource allocation, Management Science 28 (1982) 604620.
[9] S. Hasija, E.J. Pinker, R.A. Shumsky, Call center outsourcing contracts under information
asymmetry, Management Science 54 (2008) 793807.
[10] B. Holmstrom, P. Milgrom, Aggregation and linearity in the provision of
intertemporal incentives, Econometrica 55 (1987) 303328.
[11] A.V. Iyer, L.B. Schwarz, S.A. Zenios, A principal-agent model for product specica-
tion and production, Management Science 51 (2005) 106119.
[12] Y. Jiang, A. Seidmann, Integrated marketing and capacity contracting for capital-
intensive service systems, Decision Support Systems 51 (2011) 627637.
[13] S. Kim, M.A. Cohen, S. Netessine, Reliability or inventory? Analysis of product sup-
port contracts in the defense industry, INSEAD Working paper No. 2011/58/TOM,
2011,(Available at SSRN: http://ssrn.com/abstract = 1666100).
[14] R. Lal, V. Srinivasan, Compensation plans for single- and multi-product salesforces:
an application of the Holmstrom-Milgrom model, Management Science 39 (1993)
777793.
[15] R.P. Magee, Variable cost allocation in a principal/agent setting, The Accounting
Review 63 (1988) 4254.
[16] A. Mas-Colell, M.D. Whinston, J.R. Green, Microeconomic Theory, Oxford University
Press, New York, 1995.
[17] G.F. Mathewson, R.A. Winter, The economics of franchise contracts, Journal of Law
and Economics 28 (1985) 503526.
[18] H. Mendelson, Pricing computer services: queuing effects, Communications of the
ACM 28 (1985) 312321.
[19] B. Miller, A.G. Buckman, Cost allocation and opportunity costs, Management
Science 33 (1987) 626639.
[20] E.L. Plambeck, S.A. Zenios, Incentive efcient control of a make-to-stock production
system, Operations Research 51 (2003) 371386.
[21] E.L. Porteus, S. Whang, On manufacturing / marketing incentives, Management
Science 37 (1991) 11661181.
[22] M. Raith, Specic knowledge and performance measurement, The Rand Journal of
Economics 39 (2008) 10591079.
[23] K.C. Sen, The use of initial fees and royalties in business-format franchising,
Managerial and Decision Economics 14 (1993) 175190.
[24] Y. Tan, V.S. Mookerjee, Allocating spending between advertising and infor-
mation technology in electronic retailing, Management Science 51 (2005)
12361249.
[25] D. Wei, Inter-departmental cost allocation and investment incentives, Review of
Accounting Studies 9 (2004) 97116.
[26] S. Whang, Cost allocation revisited: an optimality result, Management Science 35
(1989) 12641273.
42 Y. Jiang, A. Seidmann / Decision Support Systems 58 (2014) 3142

[27] F. Zhang, Procurement mechanism design in a two-echelon inventory system Abraham Seidmann is Xerox Professor of Computers and Information Systems and
with price-sensitive demand, Manufacturing & Service Operations Management Operations Management at the William E. Simon Graduate School of Business Administra-
12 (2010) 608626. tion, University of Rochester. He is the author of over 100 research articles, which appear in
[28] J.L. Zimmerman, The costs and benets of cost allocations, The Accounting Review many of the leading scientic journals, and has been the founding department editor on
54 (1979) 504521. interdisciplinary management research and applications in Management Science for
10 years. He is also an associate or area editor for IIE Transactions, the International Journal
of Flexible Manufacturing Systems, Production Planning and Controls, the Journal of Intelligent
Yabing Jiang is Assistant Professor of Information Systems & Operations Management
Manufacturing, the Journal of Management Information Systems, and Production and
at the Lutgert College of Business, Florida Gulf Coast University. She holds a Ph.D.
Operations Management.
in Computers Information Systems from the William E. Simon Graduate School of
Business Administration, University of Rochester. Her research interests focus on
employing economic theories and methodologies to study IT-related topics such as new
business models and pricing strategies in electronic commerce, incentive contracting in
service facilities, outsourcing contract design, and the role of IT in corporate governance.
Her research appears in Electronic Commerce Research and Applications, the Journal
of Management Information Systems, the Journal of Revenue & Pricing Management, the
Information Resources Management Journal, and Information Systems Management.