Int. J.

Production Economics 155 (2014) 406–417

Contents lists available at ScienceDirect

Int. J. Production Economics

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

Optimal ordering and pricing policy for demand functions

that are separable into price and inventory age
Tal Avinadav a,n, Avi Herbon a, Uriel Spiegel a,b
a
Department of Management, Bar-Ilan University, Ramat-Gan, Israel
b
Department of Economics, University of Pennsylvania, Philadelphia, USA

art ic l e i nf o a b s t r a c t

Article history: We formulate and analyze two models for determining the optimal pricing, order quantity and
Received 20 June 2013 replenishment period for items whose demand function is separable into components of price and
Accepted 1 December 2013 inventory age. The first model assumes a multiplicative demand function. We provide conditions, which
Available online 12 December 2013
are satisfied by most common price-dependent demand functions, to reduce the three-variable profit
Keywords: maximization problem into a single-variable problem, which can be solved using an efficient line-search
Lot sizing method. Next, we show that a genuine additive model cannot exist, and instead suggest and analyze a
Price and inventory age dependent demand pseudo-additive model. However, this model is more limited than the multiplicative model in its ability
Separability to incorporate various combinations of price and inventory age effects, and reduction of the maximiza-
tion problem into a single-variable problem is more complicated, except in the case of a linear price
effect, which is further analyzed. For both models, we show that the optimal solution satisfies the first-
order condition for equilibrium under a monopoly, with a modification that includes inventory holding
costs. We solve numerical examples to illustrate the solution procedures.
& 2013 Elsevier B.V. All rights reserved.

1. Introduction a single-variable problem (the replenishment period), which can

Most inventory management policies involve two decision
variables: order quantity and replenishment period. These variables
directly affect the inventory level trajectory. For some time, the
literature has emphasized the unit selling price as a decision variable
that affects the inventory level indirectly, through the demand
function. Only recently have researchers begun considering how the
demand function is influenced by the inventory age (i.e., the elapsed
time measured from the most recent replenishment) in conjunction
with the selling price. This is especially important for inventory
modeling of perishable items, whose demand might be affected by
their freshness (see, for example, You, 2005; Tsao and Sheen, 2008;
Avinadav and Arponen, 2009; Valliathal and Uthayakumar, 2011;
Maihami and Kamalabadi, 2012; Avinadav et al., 2013).
In this paper we optimize the order quantity, replenishment
period and unit selling price to maximize profits, given a demand
function that is affected by both price and inventory age. Most of
the results pertain to cases in which the demand function can be
separated into the latter two factors, taking either a multiplicative
form or an additive form. The contribution of this study to the
literature is threefold: (i) showing that under common conditions
the multiplicative form maximization problem can be reduced into
n
Corresponding author. Tel./fax: þ 972 3 5318143.
E-mail address: tal.avinadav@biu.ac.il (T. Avinadav).
be solved using the efficient line-search methods; (ii) showing that
a genuine additive demand function cannot exist and suggesting
and analyzing a pseudo-additive model instead; and (iii) providing
economic interpretations for the optimality conditions.
We assume that demand rate (demand per unit of time) is non-
increasing in inventory age, and that unit retail price is fixed.
These assumptions suit inventory systems of perishable items –
i.e., items that physically deteriorate or whose quality decreases
over time – in stores, such as supermarkets, that replenish items
before expiration but avoid discounts for reduced freshness.
Examples of perishable items are food products and beverages
(e.g., fresh vegetables and fruit, dairy and meat products), fashion
goods, ink cartridges for printers, and batteries.
In general, the age of inventory is likely to negatively affect the
demand rate for perishable items. Sarker et al. (1997) claim that
this effect occurs because consumers tend to feel less confident
purchasing perishable items as their expiration dates grow nearer.
Disregarding the effect of perishable products’ limited shelf-life on
their demand may lead to significant losses. For example, accord-
ing to van Donselaar and Broekmeulen (2012), the United States
Department of Agriculture (USDA) estimates that average annual
food losses due to leftover inventory in supermarkets in 2005 and
2006 were 11.4% for fresh fruit, 9.7% for fresh vegetables and 4.5%
for fresh meat, poultry and seafood.
In the next section we present the literature dealing with the
influence of price and time on demand rate. In Section 3, we explain

0925-5273/$ - see front matter & 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.ijpe.2013.12.002
 T. Avinadav et al. / Int. J. Production Economics 155 (2014) 406–417
the approach we use to find the optimal inventory policy and to
evaluate its sensitivity to changes in the decision variables. We then
formulate the model and present general properties of the optimal
policy. In Section 4, we thoroughly analyze demand functions that
take a multiplicative form (i.e., functions in which the components of
price and inventory age are multiplied by each other), and in Section
5, we analyze demand functions that take an additive form and
discuss their limitations. Each analysis is followed by numerical
examples that demonstrate the procedure for obtaining the corre-
sponding optimal solution. We conclude with the main findings and
directions for future work.
2. Literature review
The literature on the effects of price on demand in inventory
models is comprehensive and well developed (see a literature
review in Petruzzi and Dada (1999); more recently, Chang (2013)
proposed an economic lot size model for price-dependent demand
under quantity and price discount). Most studies in this stream of
literature assume that demand is a decreasing function of price
(see examples of common convex and concave demand functions
in Kocabıyıkoğlu and Popescu, 2011).
Nasr et al. (2013) and Hsu and Hsu (2013) discuss the effect of
items’ quality – specifically, the presence of defective items – on the
optimal order quantity; however, they do not consider the negative
effect of time on product quality. Much of the literature on the
effects of time on demand focuses on the additional depletion of
inventory (beyond depletion resulting from sales to consumers) due
to deterioration over time, or to declining/growing markets. A
survey regarding different types of deteriorating inventory appears
in Raafat (1991), and a review of subsequent developments in this
field is presented in Goyal and Giri (2001). Some models assume
that inventory deterioration is a stochastic process (e.g., Philip,
1974; Tadikamalla, 1978), whereas others assume that it is deter-
ministic (e.g., Wu et al., 2006; Musa and Sani, 2012; Taleizadeh
et al., 2013). Nahmias et al. (2004) assume that prior to the
inventory's expiration, demand is not affected by the age of the
products, whereas Avinadav and Arponen (2009) assume that
the demand rate increases in the remaining shelf-life, owing to
consumers’ preference for fresh items.
Most models that consider the negative effect of elapsed time,
measured from the most recent replenishment, on demand rate do
so implicitly. For example, in models that assume that demand is
dependent on inventory level (see a review by Urban, 2005 and a
recent study by Pando et al., 2013), solving the differential equation
of the inventory level yields a demand rate that diminishes over time
(within the period). In general, existing modeling approaches that
assume an inventory-level-dependent demand rate are not applic-
able to inventory systems of perishable items, owing to the following
two reasons: First, many models assume that when inventory is
replenished, leftover inventory from the previous cycle is offered to
customers alongside fresh units (see Baker and Urban, 1988). The
analysis and implementation of such models is highly complex in
cases of items whose demand is sensitive to freshness. Second, the
assumption that inventory level is known to the customers is not
necessarily accurate. For example, in many cases customers see only
part of the inventory (e.g., in supermarkets), while additional stock is
placed in a warehouse or other backroom storage area (Urban, 2002).
Wu et al. (2006), You and Hsieh (2007), and Chang et al. (2010)
propose models that assume an inventory-level-dependent demand
rate and take inventory deterioration or perishability into account.
The models proposed by You and Hsieh (2007) and Chang et al.
(2010) also incorporate the effects of price on the demand rate.
Recent studies in inventory management of a single product
with a deterministic demand rate have considered how demand
407
rate is affected by both price and elapsed time, measured from the
most recent replenishment. You (2005) and Tsao and Sheen (2008)
studied a case with a known sales season, a linear price effect and
an exponential time effect on the demand rate, an opportunity to
adjust prices in fixed time-intervals, and no backlogging. While
You (2005) assumed an additive influence of price and time on the
demand rate, Tsao and Sheen (2008) assumed a multiplicative
influence of these factors. Both Valliathal and Uthayakumar (2011)
and Maihami and Kamalabadi (2012) assumed a multiplicative
influence of price and time on the demand rate with partial
backlogging. Maihami and Kamalabadi (2012) studied a general
exponential time effect (even when inventory is exhausted) that
could be either positive or negative (for deteriorating items),
whereas Valliathal and Uthayakumar (2011) required only that
the time effect be a non-negative continuous function. The two
studies formulated a profit maximization problem with three
decision variables: price, replenishment period and the time at
which shortage starts, and suggested two-dimensional searches to
locate the optimum. However, it seems that the papers do not fully
prove that their respective searches converge to optimal solutions.
Avinadav et al. (2013) also assumed a multiplicative effect of price
and time on the demand rate but without backlogging. Their paper
studied a case in which the price effect is negative-linear (similar
to Maihami and Kamalabadi, 2012) and the time effect is negative-
polynomial (similar to Avinadav and Arponen, 2009). They showed
that the three-variable (order quantity, replenishment period and
price) profit maximization problem can be reduced into a single-
variable (replenishment period) problem, whose objective func-
tion is quasi-concave. Consequently, the optimal replenishment
period can be obtained using an efficient line-search, and then the
other decision variables can be calculated analytically.
3. Model formulation and general properties of the optimal
policy
We formulate a model to obtain the optimal order quantity,
replenishment period and selling price of perishable items when
demand rate is affected by price and inventory age. It is assumed
that the age of an item at delivery time is zero, so that the elapsed
time measured from the most recent replenishment indicates the
inventory age. We allow the inventory age effect on demand to be
either negative or zero. In contrast to Valliathal and Uthayakumar
(2011) and Maihami and Kamalabadi (2012), backlogging is not
considered, as it is not common for most perishable items, which
are at the focus of this work. First, when a retailer stocks multiple
brands (e.g. milk of competitive dairy producers), shortage of a
single brand followed by backlogging is less likely to occur (see,
e.g., Krishna, 1992, p. 268), since the likelihood that consumers are
willing to wait for replenishment is low. Second, perishable items
are usually purchased to satisfy immediate needs of the consumer,
so that shortages lead to loss of sales.
We generalize the work of Avinadav et al. (2013) as follows:
First, the separability of the demand function into price and
inventory age is expressed either by a multiplicative or by an
additive model. Second, the effects of price and inventory age
on the demand rate are not limited to specific functions, but to
decreasing and non-increasing functions, respectively. For each
model we show that the optimal solution satisfies the basic law in
economics: marginal revenue equals marginal cost including the
associated inventory holding costs. By solving the optimality
equation, we reduce the profit-maximization problem into a
single-variable problem (in which the variable is the replenish-
ment period) that can be solved using a line-search. We also
provide a sufficient condition for quasi-concavity of the single-
variable profit function, which renders the efficient line-search
408 T. Avinadav et al. / Int. J. Production Economics 155 (2014) 406–417

methods applicable for finding the optimal policy. This condition optimal replenishment occurs exactly when the inventory is
depends only on the price effect on the demand rate, and is also exhausted. This property holds in all EOQ models (Hadley and
satisfied by most common price-dependent demand functions in Whitin, 1963, p. 32) with, possibly, only one exception – the case of
the literature. Then, we present the additive model and show that inventory-level-dependent demand (Baker and Urban, 1988). In
such a model cannot exist, as it violates a basic economical axiom. our model, we assume that demand is affected by the inventory
Instead, we suggest a pseudo-additive model, which is suitable for age, but not by the inventory level.
a limited variety of price-dependent demand functions, and
analyze two specific cases, in which the price effect is linear and Proposition 1. A profitable policy ðQ ; p; TÞ is optimal only if the
the inventory age effect is either linear or negative-exponential. inventory is exhausted exactly at time T.
The model considered here is a generalization of the common
single-item deterministic economic order quantity (EOQ) model, Proof. We prove the proposition by contradiction. If IðQ ; p; TÞ is
originally proposed by Harris (1913), where shortages are not positive, then reducing the order from Q to Q  IðQ ; p; TÞ reduces
allowed, and batch replenishment is performed periodically (e.g., the holding costs; hence, Q is not optimal. If IðQ ; p; t 0 Þ ¼ 0 for
Hadley and Whitin (1963, Section 2-2)). In addition to the common t 0 o T, then reducing T to t 0 increases the profit per unit of time.
decision variables of when and how much to order, we consider
In a profitable policy with given p and T, let I p;T ðtÞ be the
also the selling price, determined at the beginning of the replen-
optimal inventory level at time tð o TÞ, and Q p;T be the optimal
ishment period, as a decision variable. Specifically, we assume that
order quantity. Since ðd=dtÞI p;T ðtÞ ¼  λðp; tÞ and, by Proposition 1,
the demand rate is a function of price and inventory age, and that
I p;T ð0Þ ¼ Q p;T , we state
inventory is depleted only by demand. Our results pertain primarily
to cases in which the demand function can be separated into these
Corollary 1.
two factors, either in a multiplicative form (see multiplicative
Z T
model formulated in Section 4) or in an additive form (Section 5).
I p;T ðtÞ ¼ λðp; τÞdτ; ð3Þ
t
3.1. Notation
and
Z T
Parameters
Q p;T ¼ λðp; τÞdτ: ð4Þ
0
K the fixed ordering cost per replenishment ($)
c the purchasing cost per unit ($/unit)
h the unit inventory holding cost per unit time ($/[unit  day]) By applying Corollary 1 to Eq. (1), the optimal profit per period
for given p and T is
Decision variables
Z T Z T Z T
Π ðp; TÞ ¼ ðp  cÞ λðp; tÞdt  h λðp; τÞdτdt  K
Q the order quantity (units) 0 0 t
p the selling price of the item ($/unit) Z T Z T
T the inter-replenishment timeI (days) ¼ ðp  cÞ λðp; tÞdt  h t λðp; tÞdt  K; ð5Þ
0 0

Variables so that the three-variable maximization problem (Eq. (2)) can be

replaced by the two-variable problem:
t the elapsed time measured from the most recent replen-
maxfπ ðp; TÞ  Π ðp; TÞ=Tg: ð6Þ
ishment (days) p;T

λ(p,t) the demand rate at t for a given p (units/day)

I(Q,p,t) the inventory level at t for given p and Q (units)
The profit per period, Π ðQ ; p; TÞ, includes three components:
the profit obtained from sold units excluding holding costs; the
total inventory holding costs between replenishments; and the
fixed ordering cost per replenishment. Thus,
Z T
Π ðQ ; p; TÞ ¼ ðp  cÞQ  h IðQ ; p; tÞdt  K: ð1Þ
0
The objective function is the profit per unit of time, π ðQ ; p; TÞ 
Π ðQ ; p; TÞ=T, so that the optimization problem addressed in this
paper is
maxfπ ðQ ; p; TÞg:
Q ;p;T
Clearly, T 4 0 and p 4 c; otherwise, there is no profitable solution.
We further incorporate into the model two properties (which
do not necessitate that the demand rate be separable into price
and inventory age) that reduce the number of decision variables
from three to two and limit the boundaries of the search interval
for the remaining decision variables. The first property is that
I
The time elapsed between two consecutive replenishments.
Eq. (6) can be solved by decomposing the maximum operator in
two stages: first with respect to p and then with respect to T:
8 9
<maxfΠ ðp; TÞg=
p
max : ð7Þ
T : T ;
The second property of the optimal policy, as presented in the
following lemma, states that in a profitable optimal policy the last sold
unit guarantees a positive marginal profit. For simplicity, we assume
that when λðp; tÞ 4 0 then λðp; tÞ is differentiable in t for a given p.
Lemma 1. If p and T are optimal and satisfy π ðp; TÞ 40 and
ð2Þ
λðp; TÞ 4 0, then p  c  hT 40.
Proof. By Eq. (6), ð∂π ðp; TÞ=∂TÞ ¼ ð1=TÞð∂Π ðp; TÞ=∂TÞ  ðΠ ðp; TÞ=
T 2 Þ ¼ ð1=TÞð∂Π ðp; TÞ=∂T  π ðp; TÞÞ. By Eq. (5), ð∂Π ðp; TÞ=∂TÞ ¼
ðp  c  hTÞλðp; TÞ, so that a necessary condition for maximizing
π ðp; TÞ is ðp  c  hTÞλðp; TÞ ¼ π ðp; TÞ. Hence, if optimal p and T
satisfy π ðp; TÞ 40 then they also satisfy p  c  hT 40.
By Lemma 1, the optimal solution for Eq. (6) must be restricted by
T o ðp  cÞ=h in order to be profitable. This restriction simplifies the
procedure used to find this optimal solution, as we show in the
following.
 T. Avinadav et al. / Int. J. Production Economics 155 (2014) 406–417 409

Table 1
Properties of common price-dependent demand functions.

Attribute Demand

Linear Iso-elastic Exponential Logit Logarithmic Polynomial

λ1 ðpÞ a  bp ap  b ae  bp a a b ln p a  bpn
a 40; b 4 0 a 4 0; b 41 a 4 0; b 4 0 1 þ ebp a4 0; b 4 0 a 40; b 4 0;
a 4 0; b 4 0
0o na 1
λ1 0 ðpÞ b ab ab abebp b  nbpn  1
   
pb þ 1 ebp ð1þ ebp Þ2 p
λ1 ″ðpÞ 0 abðb þ 1Þ ab
2 2
ab ebp ðebp  1Þ b  nðn  1Þbpn  2
40 40 40 40
pb þ 2 ebp ðebp þ 1Þ3 p2 ðλ1 ″ðpÞ 40 for 0o no 1;
λ1 ″ðpÞ o 0 for n 4 1Þ
pλ1 0 ðpÞ 
p b  bp bp 1 npn
ηðpÞ ¼ a=b  p   
λ1 ðpÞ 1þ e  bp a=b  ln p a=b  pn
pmax a=b 1 1 1 ea=b ða=bÞ1=n
ηðpmax Þ 1 bo 1 1 1 1 1
rðpÞ 2p  a=b ð1 b
1
Þp pb
1
1 þ e  bp pð1 a=b þ ln pÞ p1  n a
p pð1þ n  1 Þ
b nb
r 0 ðpÞ 2 1 b
1
40 1 1 þ e  bp 4 1 2  a=b þ ln p ðn  1Þa
1þ n  1 þ n
np b
ðr 0 ðpÞ 4 1 for n 41Þ
r}ðpÞ 0 0 0 b p1 40 nðn  1Þa
 o0 
ebp npn þ 1 b
r″ðpÞ 40 for 0o no 1
pT ¼ r  1 ðc þ hwðTÞÞ a=b þ c þ hwðTÞ c þ hwðTÞ b
1
þ c þ hwðTÞ No closed-form expression No closed-form expression No closed-form expression
2 1
1b

4. The multiplicative effect of price and inventory age on all demand functions with a finite pmax satisfy it. M3(ii) indicates
demand that when pmax is infinite, demand elasticity is relatively (but not
necessarily infinitely) high when p approaches pmax . This require-
There may be several ways to combine the effects of price and ment is supported by Marshall (1890, Book III, Chapter IV) and is
inventory age on the demand rate. In this section, we assume that satisfied, as presented in Table 1, by three price-dependent
the two variables have a multiplicative effect, i.e., demand functions, which are also common in the literature and
have an infinite pmax (iso-elastic, exponential and logit; see, e.g.,
λðp; tÞ ¼ λ0 λ1 ðpÞλ2 ðtÞ; p Z 0; t Z 0; ð8Þ
Kocabiyikoğlu and Popescu, 2011).
where λ0 is a non-zero scaling parameter, and λ1 ðpÞ and λ2 ðtÞ Next, according to the formulation of the maximization pro-
express the effects (in percentages) of price and inventory age, blem in Eq. (7), the optimal p given T is extracted in Section 4.1,
respectively, on the demand rate. According to the multiplicative and in Section 4.2 the optimal T is obtained.
model and the assumptions below, the demand vanishes if the
price of the item exceeds a certain threshold pmax (4 c, possibly 4.1. Optimal price given the replenishment period
infinite), or if the inventory age passes the items’ maximal shelf
life duration T max (possibly infinite): The objective herein is to solve
 
Assumption M1. λ1 ðpÞ is continuous, positive and strictly decreas- max Π ðp; TÞ ; c o p o pmax ; 0 oT o T max : ð9Þ
p
ing over ðc; pmax Þ, and is zero over ½pmax ; 1Þ.
Substituting Eq. (8) in Eq. (5) yields
Assumption M2. λ2 ðtÞ is continuous, positive and non-increasing
Z T
over ½0; T max Þ, and is zero over ½T max ; 1Þ.
Π ðp; TÞ ¼ λ0 λ1 ðpÞ λ2 ðtÞdtðp c  hwðTÞÞ  K; ð10Þ
For simplicity, we further assume that λ1 ðpÞ is twice differenti-
0

able over ðc; pmax Þ and that λ2 ðtÞ is differentiable over ½0; T max Þ, so where
0 0 Z Z
that λ1 ðpÞ o0 over ðc; pmax Þ and λ2 ðtÞ r0 over ½0; T max Þ. T T
The third assumption concerns the point-price elasticity of wðTÞ  t λ2 ðtÞdt= λ2 ðtÞdt ð11Þ
demand (see p. 28 in Nicholson and Snyder, 2011), ηðpÞ  ðp=λðp; tÞÞ
0 0
0
ð∂λðp; tÞ=∂pÞ, p o pmax , which, by Eq. (8), is p λ1 ðpÞ=λ1 ðpÞ. is the average amount of time, given T, during which a unit is held
in inventory prior to being purchased. The following proposition,
Assumption M3. proved in Appendix A, presents properties of wðTÞ.

(i) limp-pmax ηðpÞ ¼ 1 when pmax is finite; Proposition 2. (i) 0 owðTÞ r 0:5T; (ii) 0 o w 0 ðTÞ r 0:5.
(ii) limp-pmax ηðpÞ o  1 when pmax is infinite. To apply the necessary condition for optimality, we begin with the
following formulation:
Z T
By M3(i), demand is infinitely elastic when price approaches a ∂
Π ðp; TÞ ¼ λ0 λ01 ðpÞ λ2 ðtÞdtðrðpÞ  c  hwðTÞÞ; ð12Þ
finite pmax . Table 1 shows that three price-dependent demand ∂p 0
functions, which are commonly used in the literature and where
have a finite pmax (linear, logarithmic and polynomial; see, e.g.,
0
Kocabiyikoğlu and Popescu, 2011), satisfy M3(i). To our knowledge, rðpÞ  p þ λ1 ðpÞ=λ1 ðpÞ ¼ pð1 þ 1=ηðpÞÞ: ð13Þ
410 T. Avinadav et al. / Int. J. Production Economics 155 (2014) 406–417

By equating Eq. (12) to zero, we obtain the necessary condition for
the optimal price for a given T:
rðpÞ ¼ c þ hwðTÞ:
RT
By Eqs. (3) and (4), Q ¼ λ0 λ1 ðpÞ 0 λ2 ðtÞdt, so that by Eq. (13),
, RT , RT
dðpQ Þ dðpQ Þ dQ dðpλ0 λ1 ðpÞ 0 λ2 ðtÞdtÞ dðλ0 λ1 ðpÞ 0 λ2 ðtÞdtÞ
¼ ¼
dQ dp dp dp dp
, 0
dðpλ1 ðpÞÞ dðλ1 ðpÞÞ pλ1 ðpÞ þ λ1 ðpÞ
¼ ¼ ¼ rðpÞ:
dp dp λ01 ðpÞ
Hence, rðpÞ is the marginal revenue, and Eq. (14) represents a
modification of the first-order condition for equilibrium under a
monopoly seller: marginal revenue is equal to marginal purchase
cost, c, together with the inventory holding cost of an average unit,
hwðTÞ. When there is no holding cost (i.e., h ¼ 0), we obtain the
standard condition for equilibrium that is commonly referred to in
the economics literature, i.e., at equilibrium, marginal revenue is
equal to marginal cost. The optimal price for a given T, denoted
pðTÞ, is extracted on the basis of the necessary condition in
Eq. (14).
The following properties of the marginal revenue and marginal
cost ensure that a solution exists for Eq. (14).
Lemma 2.
(i) rðpÞ o p over ½0; pmax Þ.
(ii) limp-pmax rðpÞ ¼ pmax .
(iii) c o c þ hwðTÞ o 0:5ðc þ pmax Þ opmax for 0 o T o ðpmax  cÞ=h.
Proof.
(i) By (13), since ηðpÞ o0 over ½0; pmax Þ.
(ii) By applying Assumption M3 in Eq. (13).
(iii) By Proposition 2(ii), c o c þ hwðTÞ r c þ 0:5hT, which proves
the claim.
By Lemma 2 and the definition of pmax , rðcÞ o c opmax ¼
limp-pmax rðpÞ, so that rðpÞ crosses c þ hwðTÞ from below at least
once over ðc; pmax Þ. Thus, pðTÞ exists.
In order to solve Eq. (14) and extract a simple formula for pðTÞ,
we present a property of rðpÞ, which, to our knowledge, is satisfied
by most price-dependent demand functions in the literature (see
Table 1):
Assumption M4. rðpÞ is strictly quasi-convex over ðc; pmax Þ.
Lemma 3. Under M4, rðpÞ is strictly increasing over ðp0 ; pmax Þ for
some p0 A ðc; pmax Þ while rðpÞ o c for p A ½c; p0 .
ð14Þ
Proof. By Theorems 3.5.9 and 3.5.12 in Bazaraa et al. (2006), rðpÞ
has a unique local minimum at some p0 A ðc; pmax Þ that is also
global. Thus, rðpÞ strictly increases over ðp0 ; pmax Þ (as illustrated by
the solid curves in Fig. 1(a) and (b)). By Lemma 2(i), rðcÞ o c;
therefore, rðpÞ o c for p A ðc; p0 .
According to Lemma 3, r  1 , the inverse function of r over
ðp0 ; pmax Þ, is well defined, so that pðTÞ ¼ r  1 ðc þ hwðTÞÞ solves
Eq. (14).
Theorem 1. Under M4, for 0 o T oðpmax  cÞ=h:
(i) pðTÞ satisfies c o r  1 ðcÞ o pðTÞ o r  1 ð0:5ðc þ pmax ÞÞ o pmax .
(ii) p0 ðTÞ ¼ hw0 ðTÞ=r 0 ðpðTÞÞ 0
 4 0 (where p ðTÞ ¼ dpðTÞ=dT and
r 0 ðpðTÞÞ ¼ drðpÞ=dpp ¼ pðTÞ ).
(iii) pðTÞ is the optimal price given T (i.e., the solution of Eq. (9)).
Proof.
(i) The claim is proved by Eq. (14), Lemma 2 and Lemma 3.
(ii) Substituting p ¼ pðTÞ in Eq. (14) and taking the derivative of
both sides of the equation with respect to T yields
r 0 ðpðTÞÞp0 ðTÞ ¼ hw0 ðTÞ. By Proposition 2(i), w0 ðTÞ 4 0, and by
Theorem 1(i) and Lemma 3, r 0 ðpðTÞÞ 4 0, implying that
p0 ðTÞ 4 0.
0
(iii) By M1, λ1 ðpðTÞÞ o0, and by Theorem 1(i) and Lemma 3,
0
r ðpðTÞÞ 40. Hence,
Z T
∂2 0
Π ðp; TÞ9 ¼ λ0 λ1 ðpðTÞÞ λ2 ðtÞdt r 0 ðpðTÞÞ o 0;
∂p2 p ¼ pðTÞ 0
which satisfies the sufficient condition for the solution of
Eq. (9). □
By applying Lemma 3 and Proposition 2(i) to Eq. (14), it is clear
that an increase in h or T results in an increase in the optimal price.
The quasi-convexity of rðpÞ is satisfied by the following price-
dependent demand functions, which are the most common in the
literature (see, e.g., Kocabiyikoğlu and Popescu, 2011): linear, iso-
elastic, exponential, logit, logarithmic and polynomial, which
produce rðpÞ that is either increasing or convex, as presented in
Table 1. However, the existence of pðTÞ does not guarantee that it
has a closed-form expression. For example, the first three demand
functions above (linear, iso-elastic and exponential) have closed-
form expressions for pðTÞ, as presented in Table 1, whereas

Fig. 1. Schematic illustration of a quasi-convex rðpÞ with (a) p0 ¼ c; (b) c o p0 o pmax .

 T. Avinadav et al. / Int. J. Production Economics 155 (2014) 406–417 411

the latter three (logit, logarithmic and polynomial) do not. In the Assumption M5(i). r 0 ðpÞ Z0:5 for all p A ðr  1 ðcÞ; pmax Þ.
latter cases, the numerical methods can generally be used to
Lemma 4. Under M5(i)
calculate pðTÞ.

(i) ψ 0 ðTÞ r 0 for all 0 o T o ðpmax cÞ=h.

4.2. The optimal replenishment period
(ii) S is a convex set (actually an interval).
When pðTÞ is given, then, by Eq. (10), the maximal profit per
Proof.
period as a function of T is
Z T
(i) By Proposition 2(i), w0 ðTÞ r 0:5 for T 4 0. r 0 ðpÞ Z 0:5 for
Π ðTÞ ¼ λ0 λ1 ðpðTÞÞ λ2 ðtÞdt ðpðTÞ c  hwðTÞÞ  K: ð15Þ
0 all p A ðr  1 ðcÞ; pmax Þ implies that r 0 ðpðTÞÞ Z 0:5 for all
T A ð0; ðpmax cÞ=hÞ, since, by Theorem 1(i), r  1 ðcÞ o pðTÞ
Let π ðTÞ  Π ðTÞ=T. Then the maximization problem in Eq. (7) can
opmax over this interval of T. Hence, w0 ðTÞ r r 0 ðpðTÞÞ, and by
be formulated as a single-variable maximization problem:
Eq. (22), the claim is proved.
max fπ ðTÞg: ð16Þ (ii) The claim holds true because lim ψ ðTÞ 4 0 and ψ 0 ðTÞ r 0
T-0
n
Solving Eq. (16) yields T , the optimal replenishment n period, over S. □
RT
which is used to calculate pn  pðT n Þ and Q n ¼ λ0 λ1 ðpn Þ 0 λ2 ðtÞdt.
However, it is quite difficult to find this value analytically even
when pðTÞ has a closed-form expression. The conversion of the Let T S  supfSg. By Proposition 3, ð0; T S Þ is the interval in which
two-variable maximization problem (Eq. (6)) into a single-variable the search forT
T n is done.
problem (Eq. (16)) enables the line-search methods (which are
more efficient than the two-dimensional search methods) to be Theorem 2. Under M5(i)
used to identify an (epsilon-approximated) optimal solution.
In the previous section we showed that, under M4, r 0 ðpÞ 4 0 (i) Π ðTÞ is strictly concave (and increasing) over ð0; T S Þ.
over ðr  1 ðcÞ; pmax Þ. In the following we show that if r 0 ðpÞ Z 0:5 for (ii) π ðTÞ is a strictly quasi-concave (also pseudo-concave) function
all p A ðr  1 ðcÞ; pmax Þ then Π ðTÞ is strictly concave and π ðTÞ is strictly over ð0; T S Þ.
quasi-concave over a convex set that includes T n , so that an (iii) Every local maximum of π ðTÞ is also global.
efficient search can be used to obtain the value of T n as closely
as required (see Chapter 8 in Bazaraa et al. (2006)).
To show that Π ðTÞ is strictly concave, we define two functions Proof.
of T:
0
The profit obtained from the last sold unit: (i) We show that Π ðTÞ 4 0 and Π }ðTÞ o 0 over S. By definition of
0
S, ψ ðTÞ 4 0, and by M1 and M2, λðTÞ 4 0; hence, Π ðTÞ 40. To
ψ ðTÞ  pðTÞ  c  hT; 0 o T o ðpmax  cÞ=h; ð17Þ 0
show Π }ðTÞ o 0 it is sufficient to show that λ ðTÞ o 0 and
and the demand rate at the end of the replenishment period: ψ 0 ðTÞ r 0. Indeed, ψ 0 ðTÞ r0 by Lemma 4(i), and λ0 ðTÞ o 0 by
λðTÞ  λ0 λ1 ðpðTÞÞλ2 ðTÞ; 0 o T oðpmax  cÞ=h: ð18Þ M1, M2 and Theorem 1.
(ii) The result follows known results in convex analysis (e.g.,
In Appendix B, we show that chapter 3.5 of Bazaraa et al., 2006) since π ðTÞ ¼ Π ðTÞ=T is the
Π 0 ðTÞ ¼ λðTÞψ ðTÞ: ð19Þ ratio of a differentiable concave function to a linear positive
function.
Consequently,
(iii) According to Theorem 3.5.9 of Bazaraa et al. (2006). □
Π ″ðTÞ ¼ λ0 ðTÞψ ðTÞ þ λðTÞψ 0 ðTÞ; ð20Þ
where Now we show that M5(i) is satisfied for most of the common
0 0 0 price-dependent demand functions in the literature. First, all
λ ðTÞ ¼ λ0 ðλ1 ðpðTÞÞp ðTÞλ2 ðTÞ þ λ1 ðpðTÞÞλ2 ðTÞÞ:
0
ð21Þ
concave price-dependent demand functions (e.g., the linear func-
By substituting p0 ðTÞ ¼ w0 ðTÞ=r 0 ðpðTÞÞ (from differentiating Eq. (14) tion and the polynomial function with exponent greater than
0
with respect to T), (1) satisfy r 0 ðpÞ ¼ 2  λ1 ðpÞλ1 ″ðpÞ=ðλ1 ðpÞÞ2 Z 2 for all p o pmax , since
ψ 0 ðTÞ ¼ hðw0 ðTÞ=r 0 ðpðTÞÞ  1Þ: ð22Þ λ1 ″ðpÞ r 0. The exponential and the logit price-dependent demand
functions, which are convex, satisfy r 0 ðpÞ Z 1 for all p (see Table 1).
For simplicity, we let T max be infinitely large (a finite T max is The logarithmic price-dependent demand function, which is also
discussed in the following). Let convex, satisfies 1 o r 0 ðpÞ o 2 for all p A ðr  1 ðcÞ; pmax Þ (see Appendix
S  fT j ψ ðTÞ 4 0g: ð23Þ C). Finally, the iso-elastic price-dependent demand function with
b 42 satisfies r 0 ðpÞ 4 0:5 (see Table 1).
S is non-empty, since, by modifying Lemma 2(i),
Given Proposition 3, it is interesting to identify conditions that
limT-0 ψ ðTÞ ¼ limT-0 ðr  1 ðc þ hwðTÞÞ c  hTÞ ¼ r  1 ðcÞ  c 4 0: ensure that a system is profitable. We have identified one such
ð24Þ condition, stated in the following assumption, which also verifies
that T S is finite. This condition is identical to M5(i), except that
Proposition 3. If the system is profitable (i.e., Π ðT n Þ 4 0), then here r 0 ðpÞ ¼ 0:5 is excluded.
T n A S. Assumption M5(ii). r 0 ðpÞ Z 0:5 þ ε, ε 40, for all p A ðr  1 ðcÞ; pmax Þ.
Proof. Profitability of the system means that π ðT Þ 40. By Lemma
n
Theorem 3. Under M5(ii)
1, ψ ðT n Þ ¼ pn c  hT n 4 0. □

The following assumption is sufficient to enable an efficient (i) T S r UBðT S Þ  ðr  1 ðcÞ cÞð0:5 þ εÞ=ðhεÞ,
line-search method to be used to find the optimal replenishment (ii) π ðT S Þ ¼  T S π 0 ðT S Þ,
period, as explained in Lemma 4 and Theorem 2. (iii) the inventory system is non-profitable if and only if π ðT S Þ r 0.
412 T. Avinadav et al. / Int. J. Production Economics 155 (2014) 406–417

Proof. 2000
RT 1800
(i) ψ ðTÞ ¼ a þ 0 ψ 0 ðtÞdt where, by Eq. (24), a ¼ lim ψ ðTÞ ¼
T-0 1600
r  1 ðcÞ  c 40. By Eq. (22), Proposition 2(i) and the condition
1400
r ðpÞ Z0:5 þ ε for all p A ðr ðcÞ; pmax Þ, ψ ðTÞ ¼ hðw0 ðTÞ=
0 1 0

r 0 ðpðTÞÞ  1Þ r hð0:5=r 0 ðpðTÞÞ  1Þ r hð0:5=ð0:5 þ εÞ  1Þ ¼  hε= 1200

Profit
ð0:5 þ εÞ for all 0 o T o ðpmax cÞ=h. Thus, by setting 1000
b   hε=ð0:5 þ εÞ o0, the inequality ψ ðTÞ r a þ bT is valid. 800
Hence, ψ ðTÞ ¼ 0 produces T S , which satisfies T S r 600
UBðT S Þ ¼  a=b. 400
0
(ii) By differentiating π ðTÞ ¼ Π ðTÞ=T, we obtain π 0 ðTÞ ¼ ðΠ ðTÞT
0 200
 Π ðTÞÞ=T ¼ ðΠ ðTÞ  π ðTÞÞ=T, so that, by Eqs. (18) and (19),
2
0 T
π ðTÞ ¼ λ0 λ1 ðpðTÞÞλ2 ðTÞψ ðTÞ  T π 0 ðTÞ. By (i), ψ ðT S Þ ¼ 0, so that
π ðT S Þ ¼ T S π 0 ðT S Þ.

0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
(iii) Clearly, if the inventory system is non-profitable, π ðT S Þ r 0. Fig. 2. πðTÞ for demand rate that is linear in both price and inventory age.
For the other direction, if π ðT S Þ r 0 then, by (ii), π 0 ðT S Þ Z 0. By
Theorem 2(ii), π ðTÞ is strictly quasi-concave over ð0; T S Þ, so
pðTÞ ¼ 0:5ðpmax þ c þ ð1=3ÞhTð2  1=ð2  T=T max ÞÞÞ for 0oT r
that π 0 ðTÞ Z0 over this interval. Hence, π ðTÞ r π ðT S Þ r 0 for all
min fT max ; ðpmax  cÞ=hg.
T A ð0; T S Þ. □ For example, let λ0 ¼ 1000, a1 ¼ 1, b1 ¼ 0:1, a2 ¼ 1, b2 ¼ 0:05,
h ¼ 0:1, c ¼ 1 and K ¼ 100. Then pmax ¼ 10, ðpmax  cÞ=h ¼ 90,
Under M5(ii), T S , which satisfies ψ ðT S Þ ¼ 0, can be obtained T max ¼ 20, wðTÞ ¼ Tð3  0:1TÞ=ð6  0:15TÞ for 0 oT r 20, and
(epsilon-approximated) using an efficient line-search method pðTÞ ¼ 0:5ð11 þ 0:1wðTÞÞ for 0 o T r minf20 ; 90g ¼ 20. Since
(such as the bisection method) over the interval ð0; UBðT S ÞÞ, where ψ ð20Þ ¼ 2:833 4 0, then T S ¼ 20. After algebraic manipulation, the
UBðT S Þ is given in Theorem 3(i). Then T n can be obtained (epsilon- maximization problem, Eq. (16), can be formulated as
8 !2 9
approximated) using an efficient line-search over ð0; T S Þ. To further < 270  8:25T þ 0:05T 2 100=
reduce the computational complexity, the search for T S and T n can max π ðTÞ ¼ ð1  0:025TÞ  ; 0 o T o 20:
: 6  0:15T T ;
be performed in one application of a search algorithm.
When T max is finite and satisfies T max r T S then, by Theorem 1, Fig. 2 presents π ðTÞ for T A ð0 ; 20Þ. A golden section search
T n A ð0; T max  in a system that has a profitable optimal solution. within this interval yields T n ¼ 1:185, which implies pn ¼ 5:53,
Consequently, (ii) and (iii) of Theorem 3 are valid after replacing T S Q n ¼ 514:1 and π n ¼ 1855:1.
with T max (see Appendix D). The procedure for obtaining
the optimal solution of the multiplicative model is summarized
4.3.2. Price has a linear effect and inventory age has an exponential
in Appendix E, and Section 4.3 provides examples of optimal
effect on the demand
solutions for different multiplicative demand functions. 
a1  b1 p p o pmax  a1 =b1
After T n is obtained then the optimal price is pnn ¼ pðT n Þ, and Let λ1 ðpÞ ¼ and λ2 ðtÞ ¼ a2 e  b2 t ,
RT
the optimal order quantity is Q n ¼ λ0 λ1 ðpn Þ 0 λ2 ðtÞdt (by 0 otherwise
RT
Eqs. (3) and (4)). where a1 ; b1 ; a2 ; b2 40. Then, by Eq. (11), wðTÞ ¼ 0 ta2 e  b2 t dt=
RT  b2 t 1
0 a2 e dt ¼ b2  T=ðeb2 T  1Þ, and as presented in Table 1,
4.3. Examples 1
pðTÞ ¼ 0:5ðpmax þ c þ hðb2  T=ðeb2 T  1ÞÞÞ.
For example, let λ0 ¼ 1000, a1 ¼ 1, b1 ¼ 0:1, a2 ¼ 1, b2 ¼ 0:05,
In this section we demonstrate the use of an efficient line-
h ¼ 0:1, c ¼ 1 and K ¼ 2500. Then pmax ¼ 10, T max ¼ 1,
search method to find the optimal solution (i.e., T n ; pn ; Q n and
π ðT n Þ) for four demand functions. In each of the first three wðTÞ ¼ 20  T=ðe0:05T  1Þ for T 4 0, and pðTÞ ¼ 0:5ð11 þ 0:1wðTÞÞ for
examples, the demand function satisfies M1–M5(ii), pðTÞ has a 0 o T r ðpmax  cÞ=h ¼ 90. Since ψ ð1Þ ¼ 4:42 40 and ψ ð90Þ ¼  3:55
closed-form expression, and the search interval of T n (i.e., o 0, we use the bisection search within the interval ð1 ; 90Þ, and
0 o T o min fT S ; T max g) is finite (in the third example we show find that ψ ðT S Þ ¼ 0 for T S ¼ 53. Hence, after algebraic manipula-
how to identify an unprofitable item based only on the profit value tion, the maximization problem, Eq. (16), can be formulated as
( )
at T S ). In the fourth example we show how to solve a case 2000ð3:5 þ 0:05T=ðexpð0:05TÞ  1ÞÞ2 ð1  expð  0:05TÞÞ  2500
that does not satisfy M5(ii) and in which T s and T max are max π ðTÞ ¼ ;
T
infinitely large.
0 oT o 53:
4.3.1. Price and inventory age each have a linear effect on the Fig. 3 presents π ðTÞ for T A ð0 ; 53Þ. A golden section search
demand rate within this interval yields T n ¼ 6:807, which implies pn ¼ 5:66,
Let Q n ¼ 2503:7, and π n ¼ 1228:8.

a b p p o pmax  a1 =b1
λ1 ðpÞ ¼ 1 1 4.3.3. Price and inventory age each have an exponential effect
0 otherwise
on the demand rate
and let Let λ1 ðpÞ ¼ a1 e  b1 p and let λ2 ðtÞ ¼ a2 e  b2 t , where a1 ; b1 ; a2 ;
 RT RT
a2  b2 t t o T max  a2 =b2 b2 4 0. Then, by Eq. (11), wðTÞ ¼ 0 ta2 e  b2 t dt= 0 a2 e  b2 t dt ¼
λ2 ðtÞ ¼ ; 1
b2  T=ðeb2 T  1Þ, and as presented in Table 1, pðTÞ ¼ b1 þ
1
0 otherwise 1
c þ hðb2  T=ðeb2 T  1ÞÞ.
For example, let λ0 ¼ 1000, a1 ¼ 1, b1 ¼ 1, a2 ¼ 1, b2 ¼ 0:05,
where a1 ; b1 ; a2 4 0 and b2 Z0. Then, for 0 o T o T max , by Eq. (11), h ¼ 1, c ¼ 1 and K ¼ 500. Then T max ¼ 1, wðTÞ ¼ 20
RT RT
wðTÞ ¼ 0 tða2  b2 tÞdt= 0 ða2  b2 tÞdt ¼ Tð3a2  2b2 TÞ=ð6a2  3b2 TÞ  T=ðe0:05T 1Þ and pðTÞ ¼ 31  T=ðe0:05T 1Þ for T 4 0. By Theorem
¼ ð1=3ÞTð2  1=ð2 T=T max ÞÞ, and, as presented in Table 1, 3(i), λ1 ðpÞ ¼ a1 e  b1 p satisfies the condition for a finite T S (since
 T. Avinadav et al. / Int. J. Production Economics 155 (2014) 406–417 413

1500 150
120
1000 90
60
500
30
0 0
-30
-60

Profit
-500
Profit

-90
-1000 -120
-150
-1500 -180
-210
-2000
-240
-2500 -270
-300 T
-3000 T 0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60 64
0 4 8 12 16 20 24 28 32 36 40 44 48 52
Fig. 5. πðTÞ for demand rate that is iso-elastic in price and constant in
Fig. 3. πðTÞ for demand rate that is linear in price and exponential in inventory age. inventory age.

0 5. The additive effect of price and inventory age

-100
In this section we explore demand functions that are separable
-200
into additive components of price and inventory age. An additive
-300 demand function that is analogous to the multiplicative function
-400 would be formulated as follows:
Profit

-500 λðp; tÞ ¼ λ1 ðpÞ þ λ2 ðtÞ; p Z0; t Z 0;

-600 where λ1 ðpÞ and λ2 ðtÞ indicate the effects of price and inventory
-700 age on the demand rate.
In contrast to the multiplicative model, in which each of the
-800
effects must be non-negative, so that λ0 λ1 ðpÞλ2 ðtÞ is also non-
-900 negative, in the additive model, either component can be negative,
-1000 T as long as λ1 ðpÞ þ λ2 ðtÞ Z 0. However, according to the common
0 10 20 30 40 50 60 70 80 90 100 axiom that limp-1 λðp; tÞ ¼ 0 for every t, the additive model must
Fig. 4. πðTÞ for demand rate that is exponential in both price and inventory age. satisfy λ2 ðtÞ ¼  limp-1 λ1 ðpÞ for every t, yielding a model with no
inventory age effect. Thus, we conclude that a non-degenerate
additive model of price and inventory age does not exist.
For example, while negative-exponential price and inventory age
r 0 ðpÞ ¼ 1), so we calculate ψ ðTÞ ¼ 30  T=ð1  expð  0:05TÞÞ for effects can be combined in the multiplicative model (i.e.,
T ¼ 1; 2; 4; 8; ::: until ψ ðTÞ o 0 for the first time, and find that it λðp; tÞ ¼ ae  b1 p e  b2 t ), such effects cannot be combined in the addi-
occurs for T ¼ 32. A bisection search within the interval ð1 ; 32Þ tive model (i.e., λðp; tÞ ¼ a1 e  b1 p þa2 e  b2 t ), since the outcome is a
yields ψ ðTÞ ¼ 0 for T S ¼ 17:484. By Eq. (15), Π ðTÞ ¼ 200000 positive demand rate for a given t even when price is infinitely large.
expð  31 þ T=ðexpð0:05TÞ  1ÞÞð1  expð  0:05TÞÞ  500. Since π ðT S Þ ¼ This critical problem (which is avoided in the multiplicative model
28:6 o 0, then by Theorem 3(iii), the inventory system is not because limp-1 λðp; tÞ ¼ 0 is attained by letting limp-1 λ1 ðpÞ ¼ 0)
profitable, and the optimal solution is not to order at all. Fig. 4 can be circumvented by letting λ1 ðpÞ ¼  λ2 ðtÞ for all large enough
presents π ðTÞ for T A ð0 ; 100Þ. values of p. This leads to the formulation of a pseudo-additive model.
The model conforms to the following assumptions:
4.3.4. Price has an iso-elastic effect and inventory age has no effect Assumption A1. λ1 ðpÞ is continuous and strictly decreasing over
on the demand ½0; 1Þ.
Let λ1 ðpÞ ¼ a1 p  b1 and let λ2 ðtÞ ¼ 1, where a1 4 0 and b1 4 1.
RT RT
Then, by Eq. (11), wðTÞ ¼ 0 tdt= 0 dt ¼ 0:5T; T 4 0, and as pre- Assumption A2. λ2 ðtÞ is continuous and strictly decreasing over
1
sented in Table 1, pðTÞ ¼ ðc þ 0:5hTÞ=ð1  b1 Þ. ½0; 1Þ.
For example, let λ0 ¼ 1000, a1 ¼ 1, b1 ¼ 2, h ¼ 0:1, c ¼ 1 and
1 1 1
K ¼ 500. Then pðTÞ ¼ 2 þ 0:1T for T 4 0. Since r 0 ðpÞ ¼ 1  b1 ¼ 0:5 By A1 and A2, λ1 and λ2 , the inverse functions of λ1 and λ2 ,
1
and T max ¼ 1, then the sufficient condition in Theorem 3(i) for a respectively, are well defined. If T max  λ2 ð  λ1 ð0ÞÞ exists, then
finite T S is not satisfied, and ψ ðTÞ ¼ pðTÞ  c  hT ¼ 1 4 0 for all either it is positive (expressing the age at which demand vanishes
T 40. Hence, after algebraic manipulation, the maximization even at a price of zero) or T max ¼ 1 (see examples (i) and (ii)
problem, Eq. (16), can be formulated as below).
( ) 1
Let pmax ðtÞ  λ1 ð  λ2 ðtÞÞ be the price for which demand
1 þ 0:05T 500
max π ðTÞ ¼ 1000  : vanishes at time t.
ð2 þ 0:1TÞ2 T
Assumption A3. T max and pmax ð0Þ exist.
Fig. 5 presents π ðTÞ for T A ð0 ; 64Þ. The derivative of π ðTÞ is
π ðTÞ ¼ 500=T 2  50=ð2 þ 0:1TÞ2 and π 0 ð1Þ ¼ 488:7 4 0. We calculate
0
By A1–A3, pmax ðtÞ exists and decreases for 0 r t oT max , where
π 0 ðTÞ for T ¼ 2; 4; 8; 16, stopping at 16 because π 0 ð16Þ ¼  1:9 o 0. pmax ðT max Þ ¼ 0. The pseudo-additive model can capture non-
A bisection search within the interval ð1 ; 16Þ yields T n ¼ 9:250, instantaneous deteriorating items by defining λ2 ðtÞ as a constant
which implies pn ¼ 2:925, Q n ¼ 1081:1 and π n ¼ 116:9. for an initial interval ½0; t 1  and then letting it strictly decrease over
414 T. Avinadav et al. / Int. J. Production Economics 155 (2014) 406–417

1
ðt 1 ; 1Þ. In this case we define λ2 only over an interval that Hence, rðp; TÞ is the marginal revenue, and Eq. (31), like Eq. (14) in
excludes λ2 ð0Þ, and pmax ðtÞ is non-increasing in t for 0 rt o T max . the multiplicative model, represents a modification of the first-
Finally, the pseudo-additive model is order condition for equilibrium under a monopoly seller: marginal
( revenue is equal to marginal purchase cost, c, together with the
λ1 ðpÞ þ λ2 ðtÞ 0 rp opmax ðtÞ
λðp; tÞ ¼ : ð25Þ inventory holding cost of an average unit depleted only by the
0 p Z pmax ðtÞ
price effect, 0:5hT. The optimal price for a given T, denoted pðTÞ, is
As examples, consider λ1 ðpÞ ¼ b1 p, b1 40. extracted on the basis of the necessary condition in (31).

(i) When λ2 ðtÞ ¼ a2  b2 t, a2 4 0; b2 Z 0, then pmax ðtÞ ¼ ða2 b2 tÞ= 5.2. The optimal replenishment period
b1 and T max ¼ a2 =b2 .
(ii) When λ2 ðtÞ ¼ a2 e  b2 t , a2 4 0; b2 Z 0, then pmax ðtÞ ¼ Let pðTÞ be given. Then, by Eq. (27), the maximal profit per
ða2 =b1 Þe  b2 t and T max ¼ 1. period (expressed as a function of T) is Π ðTÞ  Π ðpðTÞ; TÞ, and the
corresponding profit per unit of time is π ðTÞ  Π ðTÞ=T. The max-
Clearly, in the general case, the price effect in the pseudo- imization problem Eq. (26) can therefore be formulated as a
additive model is not separable from the inventory age effect, single-variable maximization problem:
 
owing to the formulation of pmax ðtÞ. As a result, this model is more max π ðTÞ
limited than the multiplicative model in its ability to incorporate
S:t c þ hT o pðTÞ opmax ðTÞ ð32Þ
various combinations of price and inventory age effects on the
demand function. It seems, however, that inseparability of the which can be solved using a line search method. The optimal price
price effect from the inventory age effect is not characteristic of and order quantity are calculated as pn ¼ pðT n Þ and Q n ¼ ðλ1 ðpn Þ þ
common behavior of demand. λ2 ðT n ÞÞ T n , respectively. The procedure for obtaining the optimal
Let π ðp; TÞ  Π ðp; TÞ=T. Then, following Lemma 1, the maximi- solution of the additive model is summarized in Appendix F, and
zation problem – similar to that of the multiplicative model – is examples are provided in 5.3.1.
max fπ ðp; TÞg
5.3. The case of a linear price effect
S:t c þ hT o p o pmax ðTÞ ð26Þ
where the profit per period under the pseudo-additive model is In the additive model, in contrast to the multiplicative model, the
obtained by substituting Eq. (25) in Eq. (5): mix of price and inventory age in the necessary condition for
 Z T   Z T  optimality (Eq. (31)) complicates the ability to obtain an analytical
Π ðp; TÞ ¼ ðp  cÞ λ1 ðpÞT þ λ2 ðtÞdt h 0:5λ1 ðpÞT 2 þ t λ2 ðtÞdt  K: extraction of the optimal p for a given T, or even to find conditions
0 0
under which Eq. (31) produces a unique p for a given T, for most
ð27Þ price and inventory age effects. In this section we analyze the case in
Solving Eq. (26) for general λ1 ðpÞ and λ2 ðtÞ requires a two- which price has a linear effect on the demand rate, i.e., λ1 ðpÞ ¼  b1 p,
dimensional search. To facilitate the use of more efficient search b1 4 0, which produces a unique optimal p for a given T. Hence,
methods, we develop conditions under which Eq. (26) can be pmax ðTÞ ¼ λ2 ðTÞ=b1 ; ð33Þ
reduced into a single-variable problem. We start by separating the
maximization problem into two sequential steps following Eq. (7): and by Eq. (31),
first with respect to p and then with respect to T. pðTÞ ¼ 0:5ðc þ 0:5hT þ λ2 ðTÞ=b1 Þ: ð34Þ
Since ∂ Π ðp; TÞ=∂p ¼  2b1 T o 0, then pðTÞ (which satisfies
2 2
5.1. Optimal price given the replenishment period
∂Π ðp; TÞ=∂p ¼ 0) is the optimal price for a given T. In this case,
p0 ðTÞ ¼ ð0:5b1 h þ ðλ2 ðTÞ  λ2 ðTÞÞ=TÞ=ð2b1 Þ, which may be either posi-
To apply the necessary condition for optimality, we begin with
tive, zero or negative, according to the parameter values and the
the following formulation:
value of T (since A2 implies λ2 ðTÞ r λ2 ðTÞ).
∂
Π ðp; TÞ ¼ ðλ1 ðpÞ þ λ2 ðTÞ þ λ1 0 ðpÞðp  c  0:5hTÞÞT In what follows we analyze two common types of inventory-
∂p age effects: linear and negative exponential. In each example we
0
¼ λ1 ðpÞðrðp; TÞ  c  0:5hTÞT; ð28Þ reduce the maximization problem into a single-variable formula-
tion. As the linear inventory-age effect example is mathematically
where
Z tractable, we are able to prove in that case that the profit is quasi-
T
1 concave over a search interval that includes the optimal T.
λ2 ðTÞ  λ2 ðtÞdt; ð29Þ
T 0

and 5.3.1. The case of a linear inventory-age effect

In order to provide more explicit results, we assume
λ1 ðpÞ þ λ2 ðTÞ that inventory age has a linear effect on the demand rate. Let
rðp; TÞ  p þ : ð30Þ
λ1 0 ðpÞ λ2 ðtÞ ¼ a2  b2 t, a2 40; b2 Z 0. Accordingly, by Eq. (33),
By equating Eq. (28) to zero, we obtain the necessary condition for pmax ðTÞ ¼ ða2  b2 TÞ=b1 , so that T max ¼ a2 =b2 ; by Eq. (29), λ2 ðTÞ ¼
the optimal price for a given T: a2  0:5b2 T; and by Eq. (34), pðTÞ ¼ ða2 þ b1 c þ0:5Tðb1 h b2 ÞÞ=ð2b1 Þ,
so that p0 ðTÞ ¼ ðb1 h  b2 Þ=ð4b1 Þ. Thus, the maximization problem
rðp; TÞ ¼ c þ 0:5hT; ð31Þ (Eq. (32)) becomes
By Eqs. (3) and (4), Q ¼ ðλ1 ðpÞ þ λ2 ðTÞÞ T, so that by Eq. (30),
max fπ ðTÞ ¼ AT 2  BT þ C  K=Tg
, ,
dðpQ Þ dðpQ Þ dQ dðpðλ1 ðpÞ þ λ2 ðTÞÞ TÞ dððλ1 ðpÞ þ λ2 ðTÞÞ TÞ S:t c þ hT o ða2 þ b1 c þ 0:5Tðb1 h b2 ÞÞ=ð2b1 Þ o ða2  b2 TÞ=b1
¼ ¼
dQ dp dp dp dp ð35Þ
, 0
dðpðλ1 ðpÞ þ λ2 ðTÞÞ Þ dðλ1 ðpÞ Þ pλ1 ðpÞ þ λ1 ðpÞ þ λ2 ðTÞ where A ¼ ðb2 þ 3b1 hÞð3b2 þb1 hÞ=ð48b1 Þ 4 0, B ¼ ða2  b1 cÞ ðb2 þ
¼ ¼ ¼ rðp; TÞ:
dp dp λ1 0 ðpÞ b1 hÞ=ð4b1 Þ and C ¼ ða2  b1 cÞ2 =ð4b1 Þ 4 0. If a2 r b1 c then there is
 T. Avinadav et al. / Int. J. Production Economics 155 (2014) 406–417 415

140 by both price and inventory age. We showed that an optimal

120 solution is obtained when the inventory is exhausted exactly at
replenishment time, and that the marginal profit of the last sold unit
100
is positive. In order to provide additional insights into the optimal
80 solution, we assumed that the demand function is separable into
60 components of price and inventory age. Two types of separability
were analyzed: multiplicative and additive. In order to extract the
Profit

40
optimal price for a given replenishment period under each model
20 type, we modified the optimality equation of a monopoly (i.e., at
0 equilibrium, marginal revenue is equal to marginal cost) to include
inventory holding costs. We showed that the properties of this
-20
optimality equation – specifically, the way in which it is influenced
-40 by the effect of inventory age on the demand function – are
-60 T dependent on the model type. In the multiplicative case, the effect
0 1 2 3 4 5 6 7 8 9 10 11
of inventory age influences only the integrated marginal cost, while
Fig. 6. πðTÞ for an additive demand rate that is linear in both price and in the additive model it influences only the marginal revenue.
inventory age. For the multiplicative model, we showed that under common
assumptions, the three-variable maximization problem can be
no profitable solution to the system, since λðc; tÞ ¼ a2  b1 c  b2 t reduced into a single-variable problem, in which the variable is the
o0 for all t 4 0 (i.e., there is no demand even when the product is replenishment period. We showed that a genuine additive model
sold without profit). Otherwise, if a2 4 b1 c, then, as proved in does not exist, and suggested a pseudo-additive model instead. In
Appendix G, π ðTÞ is quasi-concave over the search interval defined this model, which is considerably more limited than the multi-
by the constraint, so that the optimal T can be obtained by using plicative model, reducing the maximization problem into a single-
the efficient line-search methods. variable problem is not a straightforward task, as p and T are not
Numeric example. Let b1 ¼ 10, a2 ¼ 100, b2 ¼ 5, h ¼ 0:1, c ¼ 1 separable in the optimality equation, except in special cases. One
and K ¼ 100. Then pmax ðTÞ ¼ 10  0:5T (implying T max ¼ 20), such exception, which we have analyzed, is the case in which price
pðTÞ ¼ 5:5 0:1T, A ¼ 0:267 B ¼ 13:5 and C ¼ 202:5, so that the has a linear effect on demand. Owing to the difficulties associated
maximization problem (Eq. (35)) is with the additive model, any extension to a non-separable function
n o of demand should be carried out carefully to ensure that the essential
max π ðTÞ ¼ 0:267T 2 13:5T þ 202:5  100=T properties of a legitimate demand function are maintained.
T 40

S:t 0 o T o 11:25: Avinadav et al. (2013) modeled demand that decreases linearly in
price and polynomially in inventory age. The current work extends their
Fig. 6 presents π ðTÞ for T A ½0:1 ; 11. Using the bisection search to model to any price-dependent demand function, in which marginal
solve π 0 ðTÞ ¼ 0:533T  13:5T þ100=T 2 ¼ 0 within the interval revenue is a quasi-convex function of price (with derivative greater than
ð0:1 ; 11:25Þ yields T n ¼ 2:892, which implies pn ¼ 5:21, Q n ¼ 0.5), and to any non-increasing time-dependent demand function.
287:26 and π n ¼ 131:11. Avinadav et al. (2013) obtained a condition for profitability of the
optimal solution, given a specific demand function; herein we show
5.3.2. The case of a negative-exponential inventory-age effect that this condition applies to our extended model as well.
You (2005) provides an example of a model in which price has We conclude with three directions for future work. One direction is
a linear effect on demand and time has a negative-exponential relaxing the assumption that price is fixed along the period, and
effect, over a predefined time interval reflecting the sales season. developing a model that allows price to be a function of inventory age
We modified You’s (2005) demand function such that λðp; tÞ ¼ or inventory level. A simple case of time-dependent price is giving a
a2 e  b2 t  b1 p for 0 rp o pmax ðtÞ ¼ ða2 =b1 Þe  b2 t (implying T max ¼ discount at a certain time-point within the period to elevate sales of
1), and λðp; tÞ ¼ 0 otherwise. By Eq. (29) (after substituting items that are close to their expiration date. A second direction is
RT
0 λ2 ðtÞdt ¼ ða2 =b2 Þð1  Þ), λ2 ðTÞ ¼ ða2 =b2 Þð1  e  b2 T Þ=T, so that
 b2 T including additional marketing factors that affect the demand func-
e
by Eq. (34), pðTÞ ¼ 0:5 c þ 0:5hT þ a2 ð1  e  b2 T Þ=ðb1 b2 TÞ . tion, such as observed inventory-level or investment in product
RT
Thus, after substituting 0 t λ2 ðtÞdt ¼ ða2 =b2 Þð1  ð1 þ b2 TÞe  b2 T Þ,
2 promotion. Another direction, which seems to be the most important,
the maximization problem (Eq. (32)) becomes is relaxing the assumption that demand is deterministic and devel-
8 0  19 oping an appropriate model for demand that is a stochastic process.
> ðpðTÞ  cÞ ða2 =b2 Þð1  e  b2 T Þ  b1 TðpðTÞÞ  K >
>
> @ h i A>
> Such a model has to consider ordering-level and shortages, due to
>
> 2 >
>
>
<  h ða2 =b2 Þð1  ð1 þ b2 TÞe  b2 T Þ  0:5b1 ðpðTÞÞT 2 >
= uncertainty of demand, and how shortage affects consumers’ behavior.
max π ðTÞ ¼
>
> T >
>
>
> >
>
>
> >
> Acknowledgments
: ;

S:t c þ hT o 0:5½c þ 0:5hT þ a2 ð1  e  b2 T Þ=ðb1 b2 TÞ o ða2 =b1 Þe  b2 T : The authors thank Mordecai Henig for his constructive com-
ments and suggestions. Thanks are also due to the IJPE main Guest
ð36Þ
Editor (Leopoldo Eduardo Cárdenas-Barrón), the Senior Editor, and
Simplifying the constraints in Eq. (36) requires the use of the two anonymous referees for their help in improving this paper.
numerical methods. The optimal T can then be obtained by using a
line-search method over the domain delimited by the constraints.
Appendix A. By taking the derivative of wðTÞ we obtain
6. Concluding remarks RT RT RT !
T λ2 ðtÞdt  t λ2 ðtÞdt T λ2 ðTÞ t λ2 ðtÞdt
w0 ðTÞ ¼ λ2 ðTÞ 0

R
0
2 ¼ RT 1  0R T :
0 λ2 ðtÞdt T 0 λ2 ðtÞdt
In this paper we investigated optimal ordering and pricing
0 λ2 ðtÞdt
T
policies when the demand function is deterministic and is affected
416 T. Avinadav et al. / Int. J. Production Economics 155 (2014) 406–417

Since λ2 ðtÞ is positive and non-increasing for t A ½0; T max Þ then Step 7: Solve numerically
RT RT
w0 ðTÞ 4 0. Since λ2 ðtÞ Z λ2 ðTÞ; t rT, then 0 t λ2 ðtÞdt Z 0 t λ2 ðTÞ ( )
RT
dt ¼ 0:5T λ2 ðTÞ and
2
λ0 λ1 ðpðTÞÞ λ2 ðtÞdt ðpðTÞ  c  hwðTÞÞ  K
! ! max π ðTÞ ¼ 0
T
T λ2 ðTÞ 0:5T 2 λ2 ðTÞ T λ2 ðTÞ T λ2 ðTÞ
w0 ðTÞ r R T 1 R T ¼ RT 1  0:5R T :
0 λ2 ðtÞdt T 0 λ2 ðtÞdt 0 λ2 ðtÞdt 0 λ2 ðtÞdt
RT and obtain the optimal replenishment period T n over ð0; T search Þ.
Let y ¼ T λ2 ðTÞ= 0 λ2 ðtÞdt, so that w0 ðTÞ r yð1  0:5yÞ. Since Step 8: Obtain the optimal price pn ¼ pðT n Þ.
R Tn
yð1  0:5yÞ r0:5 for all y then 0 o w0 ðTÞ r0:5, which implies Step 9: Obtain the optimal order quantity Q n ¼ λ0 λ1 ðpn Þ 0 λ2 ðtÞdt.
RT RT
0 o wðTÞ ¼ 0 w0 ðtÞdt r 0 0:5dt ¼ 0:5T. Step 10: End.

Appendix B
Appendix F. An algorithm to obtain the optimal solution of
the additive model

RT 
Π 0 ðTÞ ¼ λ0 λ1 0 ðpT ÞpT 0 0 λ2 ðtÞdt þ λ1 ðpT Þλ2 ðTÞ ðpT  c  hwðTÞÞ Step 1: Set problem parameters c; λ0 ; K; h; λ1 ðpÞ and λ2 ðtÞ.
1
Z  Step 2. Find pmax ðtÞ ¼ λ1 ðλ2 ðtÞÞ explicitly, if possible, or
T  numerically (for a given t) otherwise.
þ λ1 ðpT Þ λ2 ðtÞdt pT 0  hw0 ðTÞ

0 Step 3: Extract pðTÞ explicitly from rðp; TÞ ¼ c þ 0:5hT, if possible,
RT 
λ1 0 ðpT ÞpT 0 λ2 ðtÞdt þ λ1 ðpT Þλ2 ðTÞ ðpT  c  hwðTÞÞ or otherwise, find numerically the value of p for a given T.
¼ λ0 0
Step 4: Solve analytically, if possible, or numerically, otherwise,
Z T  h i h i
8 RT RT 9
þ λ1 ðpT Þ λ2 ðtÞdtpT 0  λ1 ðpT Þhλ2 ðTÞðT  wðTÞÞ < ðpðTÞ  cÞ λ1 ðpðTÞÞT þ 0 λ2 ðtÞdt  h 0:5λ1 ðpðTÞÞT 2 þ 0 t λ2 ðtÞdt  K =
0 max π ðTÞ ¼
: T ;
¼ λ0 λ1 ðpT Þλ2 ðTÞðpT  c  hTÞ
Z T
0 S:t c þ hT o pðTÞ o pmax ðTÞ
þ λ0 λ1 ðpT ÞpT 0 λ2 ðtÞdtðpT  c  hwðTÞ þ λ1 ðpT Þ=λ1 0 ðpT ÞÞ
0 and obtain the optimal replenishment period T n .
¼ λ0 λ1 ðpT Þλ2 ðTÞðpT  c  hTÞ Step 5: Obtain the optimal price pn ¼ pðT n Þ.
Z T RT
Step 6: Calculate λ2 ðTÞ ¼ 1=T 0 λ2 ðtÞdt.
þ λ0 pT 0 λ2 ðtÞdt λ1 0 ðpT ÞðrðpT Þ  c  hwðTÞÞ
0 Step 7: Obtain the optimal replenishment quantity Q n ¼


¼ λ0 λ1 ðpT Þλ2 ðTÞðpT  c  hTÞ ¼ λðTÞψ ðTÞ
λ1 ðpn Þ þ λ2 ðT n Þ T n .
Step 8: End.
Appendix C

By Table 1, r 0 ðpÞ ¼ 2  a=b þ ln p in the case of the logarithmic Appendix G

price-dependent demand function, so that 2  a=b þ lnðr  1 ðcÞÞ
0
or 0 ðpÞ o 2  a=b þ ln pmax for p A ðr  1 ðcÞ; pmax Þ. By Table 1, in this By Eq. (35), Π ðTÞ ¼ 3AT 2  2BT þC and Π ″ðTÞ ¼ 6AT  2B. Thus,
case, pmax ¼ ea=b , and r  1 ðcÞ satisfies rðr  1 ðcÞÞ ¼ r  1 ðcÞð1  a=b þ Π ðTÞ is concave, implying that π ðTÞ ¼ Π ðTÞ=T is quasi-concave,
lnðr  1 ðcÞÞÞ ¼ c, from which we get lnðr  1 ðcÞÞ ¼ c=r  1 ðcÞ  1 þ a=b. for all
Combining the two results above, 1 o 1 þ c=r  1 ðcÞ or 0 ðpÞ o 2.
T o B=ð3AÞ ¼ 4ða2  b1 cÞðb2 þb1 hÞ=ððb2 þ 3b1 hÞð3b2 þ b1 hÞÞ: ð7:1Þ

Appendix D The constraints c þ hT oða2 þ b1 c þ 0:5Tðb1 h  b2 ÞÞ=ð2b1 Þ

o ða2  b2 TÞ=b1 are equivalent to
When T max is finite and satisfies T max o T S (implying that  
T o 2ða2  b1 cÞ= max ð3b2 þ b1 hÞ; ðb2 þ 3b1 hÞ : ð7:2Þ
ψ ðT max Þ 4 0) then, by substituting λ2 ðT max Þ ¼ 0, we get π ðT max Þ ¼
λ0 λ1 ðpðT max ÞÞλ2 ðT max Þψ ðT max Þ  T π 0 ðT max Þ ¼  T π 0 ðT max Þ. If b2 r b1 h then

(i) 2ðb2 þb1 hÞ=ð3b2 þ b1 hÞ ¼ 1 þ ðb1 h b2 Þ=ð3b2 þ b1 hÞ Z 1, so that

Appendix E. An algorithm to obtain the optimal solution
of the multiplicative model 4ða2  b1 cÞðb2 þ b1 hÞ=ððb2 þ3b1 hÞð3b2 þ b1 hÞÞ
4 2ða2  b1 cÞ=ðb2 þ3b1 hÞ:
Step 1: Set problem parameters c; λ0 ; K; h; λ1 ðpÞ and λ2 ðtÞ.
RT RT    
Step 2: Obtain wðTÞ ¼ 0 t λ2 ðtÞdt= 0 λ2 ðtÞdt and rðpÞ ¼ p (ii) max ð3b2 þ b1 hÞ; ðb2 þ 3b1 hÞ ¼ b2 þ b1 h þ 2 max b2 ; b1 h ¼
0
þ λ1 ðpÞ=λ1 ðpÞ. b2 þ 3b1 h, so that Eq. (7.2) is
Step 3: Extract pðTÞ ¼ r  1 ðc þ hwðTÞÞ explicitly, if possible, or
otherwise, find numerically the values of p for a given T by T o 2ða2  b1 cÞ=ðb2 þ3b1 hÞ:
solving rðpÞ ¼ c þ hwðTÞ.
Step 4: Extract T S from ψ ðT S Þ ¼ 0, where ψ ðTÞ  pðTÞ  c hT. If If b2 4 b1 h then
ψ ðT S Þ ¼ 0 has no solution set T S ¼ 1.
Step 5: Extract T max from λ2 ðT max Þ ¼ 0. If λ2 ðT max Þ ¼ 0 has no (i) 2ðb2 þb1 hÞ=ðb2 þ 3b1 hÞ ¼ 1 þ ðb2  b1 hÞ=ðb2 þ 3b1 hÞ 4 1, so that
solution set T max ¼ 1.
4ða2  b1 cÞðb2 þ b1 hÞ=ððb2 þ 3b1 hÞð3b2 þ b1 hÞÞ 4 2ða2  b1 cÞ=
Step 6: Let T search  min fT S ; T max g. If π ðT search Þ r0 then Q n ¼ 0;
Goto step 10. ð3b2 þ b1 hÞ:
 T. Avinadav et al. / Int. J. Production Economics 155 (2014) 406–417
   
(ii) max ð3b2 þ b1 hÞ; ðb2 þ 3b1 hÞ ¼ b2 þ b1 h þ2 max b2 ; b1 h ¼
3b2 þ b1 h, so that Eq. (7.2) is
T o 2ða2  b1 cÞ=ð3b2 þ b1 hÞ:
Hence, any T that satisfies Eq. (7.2) also satisfies Eq. (7.1), so that
π ðTÞ is quasi-concave over the interval defined by Eq. (7.2).
References
Avinadav, T., Arponen, T., 2009. An EOQ model for items with a fixed shelf-life and
declining demand rate based on time-to-expiry technical note. Asia-Pac. J. Oper.
Res. 26 (6), 759–767.
Avinadav, T., Herbon, A., Spiegel, U., 2013. Optimal inventory policy for a perishable
item with demand function sensitive to price and time. Int. J. Prod. Eco. 144 (2),
497–506.
Baker, R.C., Urban, T.L., 1988. A deterministic inventory system with an inventory-
level-dependent demand rate. J. Oper. Res. Soc. 39 (9), 823–831.
Bazaraa, M.S., Sherali, H.D., Shetty, C.M., 2006. Nonlinear Programming, Theory and
Algorithms. John Wiley & Sons, Hoboken, NJ.
Chang, C.T., Teng, J.T., Goyal, S.K., 2010. Optimal replenishment policies for non-
instantaneous deteriorating items with stock-dependent demand. Int. J. Prod.
Eco. 123 (1), 62–68.
Chang, H.C., 2013. A note on an economic lot size model for price-dependent
demand under quantity and freight discounts. Int. J. Prod. Eco. 144 (1), 175–179.
Goyal, S.K., Giri, B.C., 2001. Recent trends in modeling of deteriorating inventory.
Eur. J. Oper. Res. 134 (1), 1–16.
Hadley, G., Whitin, T.M., 1963. Analysis of Inventory Systems. Prentice-Hall, Engle-
wood Cliffs, NJ.
Harris, F.W., 1913. How many parts to make at once. Factory, The Magazine of
Management, vol. 10 (2), pp. 135–136, 152.
Hsu, J.T., Hsu, L.F., 2013. An EOQ model with imperfect quality items, inspection
errors, shortage backordering, and sales returns. Int. J. Prod. Eco. 143 (1),
162–170.
Kocabıyıkoğlu, A., Popescu, I., 2011. An elasticity approach to the newsvendor with
price-sensitive demand. Oper. Res. 59 (2), 301–312.
Krishna, A., 1992. The normative impact of consumer price expectations for
multiple brands on consumer purchase behavior. Mark. Sci. 11 (3), 266–286.
Maihami, R., Kamalabadi, I.N., 2012. Joint pricing and inventory control for non-
instantaneous deteriorating items with partial backlogging and time and price
dependent demand. Int. J. Prod. Eco. 136 (1), 116–122.
Marshall, A., 1890. Principles of Economics, 8th edition. Macmillan and Co., London
p. 1920.
417
Musa, A., Sani, B., 2012. Inventory ordering policies of delayed deteriorating items
under permissible delay in payments. Int. J. Prod. Eco. 136 (1), 75–83.
Nahmias, S., Perry, D., Stadje, W., 2004. Perishable inventory systems with variable
input and demand rates. Math. Methods Oper. Res. 60 (1), 155–162.
Nasr, W.W., Maddah, B., Salameh, M.K., 2013. EOQ with a correlated binomial
supply. Int. J. Prod. Eco. 144 (1), 248–355.
Nicholson, W., Snyder, C., 2011. Microeconomics Theory: Basic Principles and
Extensions, 11th ed.. South-Western, Stamford, CT.
Pando, V., San-José, L.A., García-Laguna, J., Sicilia, J., 2013. An economic lot-size
model with non-linear holding cost hinging on time and quantity. Int. J. Prod.
Eco. 145 (1), 294–303.
Petruzzi, N.C., Dada, M., 1999. Pricing and the newsvendor problem: a review with
extensions. Oper. Res. 47 (2), 183–194.
Philip, G.C., 1974. A generalized EOQ model for items with WEibull distribution
deterioration. IIE Trans. 6 (2), 159–162.
Sarker, B.R., Mukherjee, S., Balan, C.V., 1997. An order-level lot size inventory model
with inventory-level dependent demand and deterioration. Int. J. Prod. Eco. 48
(3), 227–236.
Raafat, F., 1991. Survey of literature on continuously deteriorating inventory
models. J. Oper. Res. Soc. 42 (1), 27–37.
Tadikamalla, P.R., 1978. An EOQ inventory model for items with gamma distributed
deterioration. IIE Trans. 10 (1), 100–103.
Taleizadeh, A.A., Mohammadi, B., Cárdenas-Barrón, L.E., Samimi, H., 2013. An EOQ
model for perishable product with special sale and shortage. Int. J. Prod. Eco.
145 (1), 318–338.
Tsao, Y.C., Sheen, G.J., 2008. Dynamic pricing, promotion and replenishment
policies for a deteriorating item under permissible delay in payments. Comput.
Oper. Res. 35 (11), 3562–3580.
Urban, T.L., 2002. The interdependence of inventory management and retail shelf
management. Int. J. Phys. Distrib. Logistics Manage. 32 (1), 41–58.
Urban, T.L., 2005. Inventory models with inventory-level-dependent demand: a
comprehensive review and unifying theory. Eur. J. Oper. Res. 162 (3), 792–804.
Valliathal, M., Uthayakumar, R., 2011. Optimal pricing and replenishment policies of
an EOQ model for non-instantaneous deteriorating items with shortages. Int. J.
Adv. Manuf. Technol. 54 (1–4), 361–371.
van Donselaar, K.H., Broekmeulen, R.A.C.M., 2012. Approximations for the relative
outdating of perishable products by combining stochastic modeling, simulation
and regression modeling. Int. J. Prod. Eco. 140 (2), 660–669.
Wu, K.S., Ouyang, L.Y., Yang, C.T., 2006. An optimal replenishment policy for non-
instantaneous deteriorating items with stock-dependent demand and partial
backlogging. Int. J. Prod. Eco. 101 (2), 369–384.
You, P.S., 2005. Inventory policy for products with price and time-dependent
demands. J. Oper. Res. Soc. 56 (7), 870–873.
You, P.S., Hsieh, Y.C., 2007. An EOQ model with stock and price sensitive demand.
Math. Comput. Modeling 45 (7), 933–942.