Professional Documents
Culture Documents
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.
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) 406–417 407
the approach we use to find the optimal inventory policy and to rate is affected by both price and elapsed time, measured from the
evaluate its sensitivity to changes in the decision variables. We then most recent replenishment. You (2005) and Tsao and Sheen (2008)
formulate the model and present general properties of the optimal studied a case with a known sales season, a linear price effect and
policy. In Section 4, we thoroughly analyze demand functions that an exponential time effect on the demand rate, an opportunity to
take a multiplicative form (i.e., functions in which the components of adjust prices in fixed time-intervals, and no backlogging. While
price and inventory age are multiplied by each other), and in Section You (2005) assumed an additive influence of price and time on the
5, we analyze demand functions that take an additive form and demand rate, Tsao and Sheen (2008) assumed a multiplicative
discuss their limitations. Each analysis is followed by numerical influence of these factors. Both Valliathal and Uthayakumar (2011)
examples that demonstrate the procedure for obtaining the corre- and Maihami and Kamalabadi (2012) assumed a multiplicative
sponding optimal solution. We conclude with the main findings and influence of price and time on the demand rate with partial
directions for future work. 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),
2. Literature review whereas Valliathal and Uthayakumar (2011) required only that
the time effect be a non-negative continuous function. The two
The literature on the effects of price on demand in inventory studies formulated a profit maximization problem with three
models is comprehensive and well developed (see a literature decision variables: price, replenishment period and the time at
review in Petruzzi and Dada (1999); more recently, Chang (2013) which shortage starts, and suggested two-dimensional searches to
proposed an economic lot size model for price-dependent demand locate the optimum. However, it seems that the papers do not fully
under quantity and price discount). Most studies in this stream of prove that their respective searches converge to optimal solutions.
literature assume that demand is a decreasing function of price Avinadav et al. (2013) also assumed a multiplicative effect of price
(see examples of common convex and concave demand functions and time on the demand rate but without backlogging. Their paper
in Kocabıyıkoğlu and Popescu, 2011). studied a case in which the price effect is negative-linear (similar
Nasr et al. (2013) and Hsu and Hsu (2013) discuss the effect of to Maihami and Kamalabadi, 2012) and the time effect is negative-
items’ quality – specifically, the presence of defective items – on the polynomial (similar to Avinadav and Arponen, 2009). They showed
optimal order quantity; however, they do not consider the negative that the three-variable (order quantity, replenishment period and
effect of time on product quality. Much of the literature on the price) profit maximization problem can be reduced into a single-
effects of time on demand focuses on the additional depletion of variable (replenishment period) problem, whose objective func-
inventory (beyond depletion resulting from sales to consumers) due tion is quasi-concave. Consequently, the optimal replenishment
to deterioration over time, or to declining/growing markets. A period can be obtained using an efficient line-search, and then the
survey regarding different types of deteriorating inventory appears other decision variables can be calculated analytically.
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, 3. Model formulation and general properties of the optimal
1974; Tadikamalla, 1978), whereas others assume that it is deter- policy
ministic (e.g., Wu et al., 2006; Musa and Sani, 2012; Taleizadeh
et al., 2013). Nahmias et al. (2004) assume that prior to the We formulate a model to obtain the optimal order quantity,
inventory's expiration, demand is not affected by the age of the replenishment period and selling price of perishable items when
products, whereas Avinadav and Arponen (2009) assume that demand rate is affected by price and inventory age. It is assumed
the demand rate increases in the remaining shelf-life, owing to that the age of an item at delivery time is zero, so that the elapsed
consumers’ preference for fresh items. time measured from the most recent replenishment indicates the
Most models that consider the negative effect of elapsed time, inventory age. We allow the inventory age effect on demand to be
measured from the most recent replenishment, on demand rate do either negative or zero. In contrast to Valliathal and Uthayakumar
so implicitly. For example, in models that assume that demand is (2011) and Maihami and Kamalabadi (2012), backlogging is not
dependent on inventory level (see a review by Urban, 2005 and a considered, as it is not common for most perishable items, which
recent study by Pando et al., 2013), solving the differential equation are at the focus of this work. First, when a retailer stocks multiple
of the inventory level yields a demand rate that diminishes over time brands (e.g. milk of competitive dairy producers), shortage of a
(within the period). In general, existing modeling approaches that single brand followed by backlogging is less likely to occur (see,
assume an inventory-level-dependent demand rate are not applic- e.g., Krishna, 1992, p. 268), since the likelihood that consumers are
able to inventory systems of perishable items, owing to the following willing to wait for replenishment is low. Second, perishable items
two reasons: First, many models assume that when inventory is are usually purchased to satisfy immediate needs of the consumer,
replenished, leftover inventory from the previous cycle is offered to so that shortages lead to loss of sales.
customers alongside fresh units (see Baker and Urban, 1988). The We generalize the work of Avinadav et al. (2013) as follows:
analysis and implementation of such models is highly complex in First, the separability of the demand function into price and
cases of items whose demand is sensitive to freshness. Second, the inventory age is expressed either by a multiplicative or by an
assumption that inventory level is known to the customers is not additive model. Second, the effects of price and inventory age
necessarily accurate. For example, in many cases customers see only on the demand rate are not limited to specific functions, but to
part of the inventory (e.g., in supermarkets), while additional stock is decreasing and non-increasing functions, respectively. For each
placed in a warehouse or other backroom storage area (Urban, 2002). model we show that the optimal solution satisfies the basic law in
Wu et al. (2006), You and Hsieh (2007), and Chang et al. (2010) economics: marginal revenue equals marginal cost including the
propose models that assume an inventory-level-dependent demand associated inventory holding costs. By solving the optimality
rate and take inventory deterioration or perishability into account. equation, we reduce the profit-maximization problem into a
The models proposed by You and Hsieh (2007) and Chang et al. single-variable problem (in which the variable is the replenish-
(2010) also incorporate the effects of price on the demand rate. ment period) that can be solved using a line-search. We also
Recent studies in inventory management of a single product provide a sufficient condition for quasi-concavity of the single-
with a deterministic demand rate have considered how demand 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
Table 1
Properties of common price-dependent demand functions.
Attribute Demand
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 Lemma 3. Under M4, rðpÞ is strictly increasing over ðp0 ; pmax Þ for
the optimal price for a given T: some p0 A ðc; pmax Þ while rðpÞ o c for p A ½c; p0 .
rðpÞ ¼ c þ hwðTÞ: ð14Þ
Proof. By Theorems 3.5.9 and 3.5.12 in Bazaraa et al. (2006), rðpÞ
RT
By Eqs. (3) and (4), Q ¼ λ0 λ1 ðpÞ 0 λ2 ðtÞdt, so that by Eq. (13), has a unique local minimum at some p0 A ðc; pmax Þ that is also
, RT , RT global. Thus, rðpÞ strictly increases over ðp0 ; pmax Þ (as illustrated by
dðpQ Þ dðpQ Þ dQ dðpλ0 λ1 ðpÞ 0 λ2 ðtÞdtÞ dðλ0 λ1 ðpÞ 0 λ2 ðtÞdtÞ the solid curves in Fig. 1(a) and (b)). By Lemma 2(i), rðcÞ o c;
¼ ¼
dQ dp dp dp dp therefore, rðpÞ o c for p A ðc; p0 .
, 0
dðpλ1 ðpÞÞ dðλ1 ðpÞÞ pλ1 ðpÞ þ λ1 ðpÞ According to Lemma 3, r 1 , the inverse function of r over
¼ ¼ ¼ rðpÞ: ðp0 ; pmax Þ, is well defined, so that pðTÞ ¼ r 1 ðc þ hwðTÞÞ solves
dp dp λ01 ðpÞ
Eq. (14).
Hence, rðpÞ is the marginal revenue, and Eq. (14) represents a
modification of the first-order condition for equilibrium under a Theorem 1. Under M4, for 0 o T oðpmax cÞ=h:
monopoly seller: marginal revenue is equal to marginal purchase
cost, c, together with the inventory holding cost of an average unit, (i) pðTÞ satisfies c o r 1 ðcÞ o pðTÞ o r 1 ð0:5ðc þ pmax ÞÞ o pmax .
hwðTÞ. When there is no holding cost (i.e., h ¼ 0), we obtain the (ii) p0 ðTÞ ¼ hw0 ðTÞ=r 0 ðpðTÞÞ 0
4 0 (where p ðTÞ ¼ dpðTÞ=dT and
standard condition for equilibrium that is commonly referred to in r 0 ðpðTÞÞ ¼ drðpÞ=dpp ¼ pðTÞ ).
the economics literature, i.e., at equilibrium, marginal revenue is (iii) pðTÞ is the optimal price given T (i.e., the solution of Eq. (9)).
equal to marginal cost. The optimal price for a given T, denoted
pðTÞ, is extracted on the basis of the necessary condition in Proof.
Eq. (14).
The following properties of the marginal revenue and marginal (i) The claim is proved by Eq. (14), Lemma 2 and Lemma 3.
cost ensure that a solution exists for Eq. (14). (ii) Substituting p ¼ pðTÞ in Eq. (14) and taking the derivative of
both sides of the equation with respect to T yields
Lemma 2. 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
(i) rðpÞ o p over ½0; pmax Þ. p0 ðTÞ 4 0.
0
(ii) limp-pmax rðpÞ ¼ pmax . (iii) By M1, λ1 ðpðTÞÞ o0, and by Theorem 1(i) and Lemma 3,
(iii) c o c þ hwðTÞ o 0:5ðc þ pmax Þ opmax for 0 o T o ðpmax cÞ=h. 0
r ðpðTÞÞ 40. Hence,
Z T
Proof. ∂2 0
Π ðp; TÞ9 ¼ λ0 λ1 ðpðTÞÞ λ2 ðtÞdt r 0 ðpðTÞÞ o 0;
∂p2 p ¼ pðTÞ 0
(i) By (13), since ηðpÞ o0 over ½0; pmax Þ. which satisfies the sufficient condition for the solution of
(ii) By applying Assumption M3 in Eq. (13). Eq. (9). □
(iii) By Proposition 2(ii), c o c þ hwðTÞ r c þ 0:5hT, which proves
the claim. 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.
By Lemma 2 and the definition of pmax , rðcÞ o c opmax ¼ The quasi-convexity of rðpÞ is satisfied by the following price-
limp-pmax rðpÞ, so that rðpÞ crosses c þ hwðTÞ from below at least dependent demand functions, which are the most common in the
once over ðc; pmax Þ. Thus, pðTÞ exists. literature (see, e.g., Kocabiyikoğlu and Popescu, 2011): linear, iso-
In order to solve Eq. (14) and extract a simple formula for pðTÞ, elastic, exponential, logit, logarithmic and polynomial, which
we present a property of rðpÞ, which, to our knowledge, is satisfied produce rðpÞ that is either increasing or convex, as presented in
by most price-dependent demand functions in the literature (see Table 1. However, the existence of pðTÞ does not guarantee that it
Table 1): has a closed-form expression. For example, the first three demand
functions above (linear, iso-elastic and exponential) have closed-
Assumption M4. rðpÞ is strictly quasi-convex over ðc; pmax Þ. form expressions for pðTÞ, as presented in Table 1, whereas
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.
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
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.
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
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
(ii) max ð3b2 þ b1 hÞ; ðb2 þ 3b1 hÞ ¼ b2 þ b1 h þ2 max b2 ; b1 h ¼ Musa, A., Sani, B., 2012. Inventory ordering policies of delayed deteriorating items
3b2 þ b1 h, so that Eq. (7.2) is 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
T o 2ða2 b1 cÞ=ð3b2 þ b1 hÞ: 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
Hence, any T that satisfies Eq. (7.2) also satisfies Eq. (7.1), so that Extensions, 11th ed.. South-Western, Stamford, CT.
Pando, V., San-José, L.A., García-Laguna, J., Sicilia, J., 2013. An economic lot-size
π ðTÞ is quasi-concave over the interval defined by Eq. (7.2). model with non-linear holding cost hinging on time and quantity. Int. J. Prod.
Eco. 145 (1), 294–303.
References 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
Avinadav, T., Arponen, T., 2009. An EOQ model for items with a fixed shelf-life and deterioration. IIE Trans. 6 (2), 159–162.
declining demand rate based on time-to-expiry technical note. Asia-Pac. J. Oper. Sarker, B.R., Mukherjee, S., Balan, C.V., 1997. An order-level lot size inventory model
Res. 26 (6), 759–767. with inventory-level dependent demand and deterioration. Int. J. Prod. Eco. 48
Avinadav, T., Herbon, A., Spiegel, U., 2013. Optimal inventory policy for a perishable (3), 227–236.
item with demand function sensitive to price and time. Int. J. Prod. Eco. 144 (2), Raafat, F., 1991. Survey of literature on continuously deteriorating inventory
497–506. models. J. Oper. Res. Soc. 42 (1), 27–37.
Baker, R.C., Urban, T.L., 1988. A deterministic inventory system with an inventory- Tadikamalla, P.R., 1978. An EOQ inventory model for items with gamma distributed
level-dependent demand rate. J. Oper. Res. Soc. 39 (9), 823–831.
deterioration. IIE Trans. 10 (1), 100–103.
Bazaraa, M.S., Sherali, H.D., Shetty, C.M., 2006. Nonlinear Programming, Theory and
Taleizadeh, A.A., Mohammadi, B., Cárdenas-Barrón, L.E., Samimi, H., 2013. An EOQ
Algorithms. John Wiley & Sons, Hoboken, NJ.
model for perishable product with special sale and shortage. Int. J. Prod. Eco.
Chang, C.T., Teng, J.T., Goyal, S.K., 2010. Optimal replenishment policies for non-
145 (1), 318–338.
instantaneous deteriorating items with stock-dependent demand. Int. J. Prod.
Tsao, Y.C., Sheen, G.J., 2008. Dynamic pricing, promotion and replenishment
Eco. 123 (1), 62–68.
Chang, H.C., 2013. A note on an economic lot size model for price-dependent policies for a deteriorating item under permissible delay in payments. Comput.
demand under quantity and freight discounts. Int. J. Prod. Eco. 144 (1), 175–179. Oper. Res. 35 (11), 3562–3580.
Goyal, S.K., Giri, B.C., 2001. Recent trends in modeling of deteriorating inventory. Urban, T.L., 2002. The interdependence of inventory management and retail shelf
Eur. J. Oper. Res. 134 (1), 1–16. management. Int. J. Phys. Distrib. Logistics Manage. 32 (1), 41–58.
Hadley, G., Whitin, T.M., 1963. Analysis of Inventory Systems. Prentice-Hall, Engle- Urban, T.L., 2005. Inventory models with inventory-level-dependent demand: a
wood Cliffs, NJ. comprehensive review and unifying theory. Eur. J. Oper. Res. 162 (3), 792–804.
Harris, F.W., 1913. How many parts to make at once. Factory, The Magazine of Valliathal, M., Uthayakumar, R., 2011. Optimal pricing and replenishment policies of
Management, vol. 10 (2), pp. 135–136, 152. an EOQ model for non-instantaneous deteriorating items with shortages. Int. J.
Hsu, J.T., Hsu, L.F., 2013. An EOQ model with imperfect quality items, inspection Adv. Manuf. Technol. 54 (1–4), 361–371.
errors, shortage backordering, and sales returns. Int. J. Prod. Eco. 143 (1), van Donselaar, K.H., Broekmeulen, R.A.C.M., 2012. Approximations for the relative
162–170. outdating of perishable products by combining stochastic modeling, simulation
Kocabıyıkoğlu, A., Popescu, I., 2011. An elasticity approach to the newsvendor with and regression modeling. Int. J. Prod. Eco. 140 (2), 660–669.
price-sensitive demand. Oper. Res. 59 (2), 301–312. Wu, K.S., Ouyang, L.Y., Yang, C.T., 2006. An optimal replenishment policy for non-
Krishna, A., 1992. The normative impact of consumer price expectations for instantaneous deteriorating items with stock-dependent demand and partial
multiple brands on consumer purchase behavior. Mark. Sci. 11 (3), 266–286. backlogging. Int. J. Prod. Eco. 101 (2), 369–384.
Maihami, R., Kamalabadi, I.N., 2012. Joint pricing and inventory control for non- You, P.S., 2005. Inventory policy for products with price and time-dependent
instantaneous deteriorating items with partial backlogging and time and price demands. J. Oper. Res. Soc. 56 (7), 870–873.
dependent demand. Int. J. Prod. Eco. 136 (1), 116–122. You, P.S., Hsieh, Y.C., 2007. An EOQ model with stock and price sensitive demand.
Marshall, A., 1890. Principles of Economics, 8th edition. Macmillan and Co., London Math. Comput. Modeling 45 (7), 933–942.
p. 1920.