Professional Documents
Culture Documents
Parameters
Ethe shelf-life duration of an item.
Kthe xed ordering cost per replenishment.
cthe purchasing cost per unit.
hthe unit inventory holding cost per unit time.
Decision variables
Qthe order quantity.
pthe selling price of a unit.
Tthe replenishment period.
Variables
tthe time elapsed from replenishment (a measure of
freshness).
p; tthe demand rate at t, given p.
IQ; p; tthe inventory level at t, given p and Q.
Q; p; Tthe prot per replenishment period, given p, Q and
T.
Q; p; Tthe prot per unit time, given p, Q and T.
In order to extend our conclusions we formulate a polynomial
decreasing demand function, whose exible shape is determined
according to a given shape parameter. Unlike the negative-
exponential function, which is commonly used in the literature
(see You (2005), You and Hsieh (2007), Maihami and Kamalabadi
(2012)), the polynomial function supports concave decrease. The
polynomial function is a better match to describe demand over
time for perishable food products, many of which undergo a mild
(or even negligible) decrease in demand when their expiration
date is distant that accelerates as the products approach their
expiration date. This concave decrease can be explained by the
awareness of consumers that products with short remaining shelf-
life should be used very close to their purchase date, due to their
inferior quality. This element has recently been discussed by Wu
et al. (2006) and Musa and Sani (2012), who claim that almost all
the inventory models for deteriorating items assume that the
quality deterioration occurs as soon as the retailer receives the
commodities. Yet the authors note that, in reality, there are many
products, such as vegetables, fruit, sh, meat, certain breads or
dairy products, whose deterioration in quality is negligible during
a certain initial period on the shelves (or in refrigerators). After the
initial period passes, the inuence of deterioration on the pro-
duct's quality (e.g., taste, shape, utility, etc.) is noticeable, and
consequently the demand for the product diminishes at larger
rates. This phenomenon, which Wu and colleagues refer to as
non-instantaneous deterioration, lies in contrast to the common
assumption of instantaneous deterioration.
For the effect of price on demand, we use the common negative
linear relation (e.g., You (2005), You and Hsieh (2007), Maihami
and Kamalabadi (2012)). We assume that the price is set in
advance of the sale, and is not negotiable with potential
consumers.
According to assumptions 5 and 6 and the above discussion,
t; p
0
1p1t=E
n
t E ; p
1
0 otherwise
_
1
where
0
is the demand scale parameter, n is the shape parameter
of the demand decrease in time n40; and
1
is the reservation
price, at which demand vanishes. While linear decrease of demand
in price is common in the literature, polynomial decrease over
time is rare (e.g., see Avinadav and Arponen (2009)). However, the
function is general enough to support a variety of decreasing
shapes, as illustrated in Fig. 1: convex decrease for 0ono1; linear
decrease for n 1; concave decrease for n41; and no decrease for
n-.
The advantage of the polynomial demand function over t is the
ability to support concave decrease of demand over time. This kind
of decrease ts the demand trajectory of most perishable grocery
products.
4. Model formulation
The prot per replenishment period, Q; p; T, includes three
economic elements: the prot margin of sold units excluding
holding cost, pcQ; the inventory holding costs, h
_
T
0
IQ; p; tdt,
and the xed ordering cost, K. The objective function, prot per
unit time, which should be maximized with respect to the three
decision variables (Q, p and T) is:
Q; p; T
Q; p; T
T
pcQh
_
T
0
IQ; p; tdtK
T
2
It is clear that when the system is protable then the inventory
level should always be positive, since unsatised demand is lost. It
is also known (see, for instance, Silver and Peterson, 1985) that in
the case of a constant demand rate, it is optimal to order exactly
when the inventory vanishes. The only exception where optimal
replenishment may occur at a positive inventory level was
demonstrated by Baker and Urban (1988) in the case of
inventory-level-dependent demand rate. In our model, in which
the demand rate depends only on the selling price and time after
replenishment, the order quantity can be lowered by the positive
amount of the inventory at the time of replenishment to increase
prot. Therefore, we can conclude that
Proposition 1. Optimal replenishment occurs exactly when
inventory is exhausted.
Consequently,
Corollary 1. In a protable system, the optimal replenishment
period is smaller than or equal to the shelf-life duration of
the item.
Proof. Using a replenishment period T, T 4E, leads to subopti-
mal prots per unit of time (since the remaining units after t E
are expired, and by assumption 4, there is no income over the
interval E; T).
Thus, assuming a protable system, T can be restricted to be not
greater than E.
1
E
t
0
0<n<1
n
n>1
n0
n=1
(p,t)
Fig. 1. (p,t) for a given p and various values of n.
T. Avinadav et al. / Int. J. Production Economics 144 (2013) 497506 499
Corollary 2. Given p and T, the optimal inventory level at time t,
I
p;T
t, is equal to the cumulative demand from time t up to the end
of the replenishment period, i.e.
I
p;T
t
_
T
t
p; d
By Corollary 2, given p and T, the optimal order quantity, Q
p;T
, is
equal to the total demand in the replenishment period, so that
Q
p;T
I
p;T
0
_
T
0
p; tdt 3
Substituting Q
p;T
for Q and I
p;T
t for IQ; p; t in (2), the
objective function can be reduced from a three-variable formula-
tion into a two-variable (p and T) formulation:
p; T
pc
_
T
0
p; tdth
_
T
0
_
T
t
p; ddtK
T
Substituting
_
T
0
_
T
t
p; ddt
_
T
0
_
0
p; dtd
_
T
0
tp; tdt,
the objective function can be further simplied to
p; T
pc
_
T
0
p; tdth
_
T
0
tp; tdtK
T
4
In order to formulate (4) explicitly for the specic demand rate
in (1) we use both:
_
T
0
1t=E
n
dt T 1
T=E
n
n 1
_ _
5
and
_
T
0
t1t=E
n
dt
T
2
2
1
2T=E
n
n 2
_ _
:
Now, the prot maximization problem, considering our sug-
gested price and time-dependent demand function, can be refor-
mulated as
max p; T
0
1p pc 1
T=E
n
n 1
_ _
hT
2
1
2T=E
n
n 2
_ _ _ _
K
T
_ _
S:t: T E
copo
1
6
The two added constraints, T E and copo
1
, exclude values
of T and p that are non-optimal or non-protable, respectively.
5. Optimal selling price given a replenishment period
In this section we nd p
T
, the optimal price given a replenish-
ment period, and provide bounds for it that are one sixth of the
trivial bounds on p in (6). Suppose that a replenishment period T
T E is given. Then the rst and second derivatives of the prot
function are,
p
p; T
0
1
c2p 1
T=E
n
n 1
_ _
hT
2
1
2T=E
n
n 2
_ _ _ _
7
and
2
p
2
p; T 2
0
1
T=E
n
n 1
_ _
8
By (6) and the negativity of (8), the prot function, p; T, is
quadratic and strictly concave in p for a given T T E. Hence,
equating (7) to zero, the optimal p for a given T is expressed by
p
T
1
c
2
hT
2
n 1
n 2
1
n
2n 1T=E
n
_ _
9
To prove that p
T
is feasible, i.e. cop
T
o
1
, we rst show that a
positive prot margin imposes a constraint on the replenishment
period.
Lemma 1. The optimal replenishment period of a protable item
cannot exceed the duration that, due to holding costs, zeroes the
prot margin of an inventory unit (i.e., the optimal T satises
T p
T
c=h).
Proof. See Appendix A.
By Corollary 1 and Lemma 1, the optimal replenishment period
satises
T min E; p
T
c=h
_ _
10
We now provide bounds on p
T
that reduce its original and
trivial interval of uncertainty, copo
1
, to one sixth of it, which
starts at the middle of the interval.
Proposition 2.
1
c
2
op
T
o
2
1
c
3
Proof. See Appendix B.
Corollary 3. The percentage change in the optimal price due to
changes in the model parameters, except for the purchasing cost
and reservation price, is less than 33.4%.
Proof. The maximal absolute change in p
T
is 2
1
c=3
1
c=2
1
c=6, and the minimal value of p
T
is
1
c=2. Thus, the percentage available change in p
T
satises
the condition
1
c
6
=
1
c
2
1
c
3
1
c
o33:4%
It is obvious that when the gap between the reservation price
and the purchasing cost is small, we expect changes in any of the
model parameters to produce only very small percentage changes
in the optimal price. By Corollary 3 we conclude that even when
the gap between the reservation price and the purchasing cost is
very large, the percentage change in the optimal price cannot
exceed 33.4%, and often it is much lower, as shown in Tables 1 and
2 in Section 7.
6. The optimal replenishment period
According to the previous section, when p
T
is substituted for p
in the objective and the constraint functions of problem (6), the
prot maximization problem is reduced to a single-variable
problem. In this section we provide means of nding numerically
the optimal value of the replenishment period i.e., the value that
maximizes the prot as closely as required.
For the sake of simplicity, we replace T with xE, where x
denotes the normalized replenishment period relative to the
shelf-life of the item, E. Consequently, according to (9), the optimal
prot given x is
p
x
1
c
2
hxE
2
n 1
n 2
1
n
2n 1x
n
_ _
11
and the constraint on x, based on (10), is
xmin 1; p
x
c=hE
_ _
12
The constraint in (12) includes x in both sides of the inequality.
In order to simplify the constraint to the form xx
max
, where x
max
is a constant depending on the problem parameters, let us dene
1
c=hE. is actually the normalized time that zeroes the
maximal potential prot margin of an inventory unit. Proposition
T. Avinadav et al. / Int. J. Production Economics 144 (2013) 497506 500
3 below determines, based on the values of and n, which of the
two components of the upper bound of x in (12) is redundant the
shelf-life duration or positive prot margin and provides an
explicit upper bound for x.
Proposition 3. The constraint x min 1; p
x
c=hE
_ _
is equiva-
lent to xx
max
, where x
max
is determined as follows:
(I) x
max
1 if 3n 7=2n 4.
(II) x
max
if o3n 7=2n 4, where uniquely satises
n 3=n 2 nn 1=2n 2n 1
n
_
over (0,1].
Proof. Substituting the expression for p
x
from (11) in
xp
x
c=hE and using algebraic manipulations, we obtain that
gx , where
gx x
n 3
n 2
nn 1
2n 2n 1x
n
_ _
The function gx is continuous, monotone increasing over (0,1],
so within this interval the following holds:
0 g0ogx g1 3n 7=2n 4:
Hence,
(I) If 3n 7=2n 4 then gx is satised over (0,1], which
means that x p
x
c=hE is redundant (see Fig. 2(a)).
(II) If o3n 7=2n 4 then x 1 is redundant (see Fig. 2(b)).
In this case, gx can be replaced by x , where satises
g . Since gx is monotone increasing over (0,1] then is
unique within this interval and can easily be found using
efcient numerical search techniques, such as the bisection
method (see Burden et al., 1985).
Note that 1:5o3n 7=2n 4 1:75 for n40, so that
x
max
1 if 1:75, and x
max
if 1:5, independently of n.
Now the two-variable maximization problem, (6), can be
written as a single-variable problem:
x
0
1p
x
p
x
c
_ _
1
x
n
n 1
_ _
hxE
2
1
2x
n
n 2
_ _ _ _
K
xE
S:t: xx
max
13
By inserting (1) and (5) in (3), the optimal order quantity given
x is
Q
x
0
1p
x
xE 1
x
n
n 1
_ _
In order to nd the optimal normalized replenishment period,
x, we analyze the prot function. We start the analysis by showing
in Appendix C that
0
x
0
x
2
f x 14
Table 1
Deviation in % from original value in Example 1.
Parameter Change in % in parameter value Inuenced
variable
50 25 10 10 25 50
0
30.0 11.5 4.1 3.5 8.1 14.3 T
n
39.1 18.3 7.1 6.8 16.5 31.8 Q
n
0.5 0.2 0.1 0.1 0.2 0.3 p
n
61.5 31.3 12.6 12.7 31.9 64.3
n
n 16.0 8.1 3.2 3.0 7.3 13.8 T
n
27.0 11.9 4.5 4.1 9.7 17.8 Q
n
0.4 0.2 0.1 0.1 0.2 0.3 p
n
21.1 7.3 2.4 2.0 4.3 7.1
n
20.8 9.4 3.6 3.4 8.4 16.1 T
n
15.0 6.1 2.2 1.9 4.5 7.8 Q
n
92.8 30.9 10.3 8.4 18.5 30.8 p
n
149.7 49.3 16.3 13.3 29.1 48.1
n
c 1.8 0.9 0.4 0.4 0.9 1.9 T
n
1.2 0.6 0.2 0.2 0.6 1.2 Q
n
2.4 1.2 0.5 0.5 1.2 2.4 p
n
6.9 3.4 1.4 1.4 3.4 6.7
n
h 3.3 1.6 0.6 0.6 1.6 3.1 T
n
3.9 1.9 0.8 0.8 1.9 3.8 Q
n
1.0 0.5 0.2 0.2 0.5 1.0 p
n
3.2 1.6 0.6 0.6 1.6 3.1
n
K 23.2 10.3 3.9 3.7 8.8 16.6 T
n
20.1 8.7 3.2 3.0 7.0 12.7 Q
n
0.5 0.2 0.1 0.1 0.2 0.3 p
n
15.0 7.0 2.7 2.6 6.4 12.3
n
E 33.9 15.3 5.8 5.4 19.7 23.5 T
n
37.4 16.9 6.4 5.9 20.0 26.0 Q
n
0.8 0.4 0.1 0.1 0.4 0.5 p
n
19.4 6.7 2.3 1.9 4.0 6.8
n
Table 2
Deviation in % from original value in Example 2.
Parameter Change in % in parameter value Inuenced
variable
50 25 10 10 25 50
0
71.6 23.7 8.0 6.6 14.6 24.7 T
n
31.7 14.2 5.4 5.1 12.2 23.0 Q
n
1.8 0.6 0.2 0.2 0.4 0.7 p
n
77.9 41.6 17.1 17.6 44.9 92.3
n
n 27.5 7.1 2.0 1.3 2.5 2.9 T
n
25.1 11.0 4.1 3.8 9.1 16.8 Q
n
0.5 0.1 0.0 0.0 0.0 0.0 p
n
57.5 25.0 9.2 8.2 18.8 33.3
n
56.5 33.2 15.1 19.5 68.4 Ends in
loss
T
n
19.7 5.3 1.0 0.8 6.3 Q
n
64.0 20.9 6.9 5.4 11.4 p
n
744.7 216.2 66.0 47.5 93.2
n
c 36.7 22.4 10.3 13.2 42.5 Ends in
loss
T
n
7.4 4.7 2.2 2.8 8.4 Q
n
16.8 8.5 3.4 3.5 9.0 p
n
236.9 103.2 37.6 32.7 72.3
n
h 10.9 5.0 1.9 1.8 4.3 8.1 T
n
12.6 5.8 2.2 2.1 5.1 9.7 Q
n
1.3 0.6 0.2 0.2 0.6 1.1 p
n
17.4 8.5 3.3 3.3 8.1 15.7
n
K 38.1 18.4 7.3 7.2 17.8 35.4 T
n
29.1 13.0 4.9 4.6 11.0 20.6 Q
n
1.1 0.5 0.2 0.2 0.5 0.9 p
n
46.2 20.4 7.7 7.1 17.0 31.6
n
E 7.7 2.9 1.0 0.9 2.0 3.4 T
n
23.3 9.2 3.3 2.9 6.6 11.6 Q
n
0.4 0.2 0.1 0.0 0.1 0.2 p
n
33.5 12.6 4.4 3.8 8.6 15.0
n
1 0
2n+4
3n+7
2n+4
3n+7
g(x)
x
1 0
g(x)
x
Fig. 2. (a) The case of 3n 7=2n 4; (b) the case of o3n 7=2n 4.
T. Avinadav et al. / Int. J. Production Economics 144 (2013) 497506 501
where
f x
K
0
E
1p
x
p
x
cnx
n1
n 1
hx
2
E
2
1
2n 1x
n
n 2
_ _ _ _
15
By (14) and the properties of f x, we next show that the prot,
x, is strictly pseudo-concave (and as demonstrated in Example 2,
it is not necessarily concave).
Lemma 2. f x is strictly decreasing over 0; x
max
.
Proof. As proven in Appendix D, the derivative of f x with
respect to x is
f
0
x 1p
x
hE2p
x
0
x1x
n
p
x
chxEnx
n
xp
x
0
1x
n
1
chxE
f
0
x o0 over 0; x
max
since the following three elements are
positive on this interval: hE2p
x
0
(see proof in Appendix E),
p
x
chxE and
1
chxE (which are satised, according to (12)
and Proposition 2, by c hxE p
x
o
1
).
Lemma 3.
(I) The prot function, x, is strictly pseudo-concave over
0; x
max
.
(II) x
max
x
max
0
x
max
.
Proof.
(I) By (14) and Lemma 2,
0
x changes its sign from positive to
negative at most once, so therefore by our claim in Appendix F,
x is strictly pseudo-concave over 0; x
max
.
(II) According to (14) and (15),
K
xE
x
0
x
0
1p
x
p
x
cnx
n
n 1
hxE
2
1
2n 1x
n
n 2
_ _ _ _
Substituting this relation into (13) followed by algebraic
manipulations yields
x
0
1p
x
1x
n
p
x
chxEx
0
x:
By Proposition 3, either x
max
1 or x
max
p
xmax
c=hE,
so that
x
max
x
max
0
x
max
0
=x
2
f x
max
0
x
max
x
max
x
max
0, for all x0; x
max
, so
that x is an increasing function, which is maximized at x
max
.
Therefore, x x
max
0 for all x0; x
max
.
By the Theorem above, a non-protable product can be a priori
detected according to the problem parameters. In such a case and
for the sake of preventing loss, the optimal policy is not to order.
On the other hand, if the product is protable then a simple line-
search algorithm can be used to nd numerically the optimal
replenishment period as closely as necessary. In Appendix G, we
provide a detailed algorithm that is guaranteed to obtain the
optimal pricing and inventory policies efciently.
7. Numerical examples
In this section we present three numerical examples to illus-
trate different behaviors of the prot function, x, and its
derivative,
0
x, driven by different parameter values.
Example 1. 3n 7=2n 4 and x
max
40
Let
0
200; n 2 (for concave decrease of the demand in
time), 0:04; c 12; h 1, K 4000 and E 10. Thus,
9:5; x
max
1 and x
max
2375:5. The optimal solution is
x
n
0:48, which leads to an optimal inventory policy of
T
n
4:8; p
n
53:65 and Q
n
410 that yields
n
3133:3. Fig. 3
describes the prot and its derivative as a function of x.
Example 2. o3n 7=2n 4 and x
max
40
Let
0
200; n 0:5 (for convex decrease of the demand in
time), 0:04; c 12; h 1, K 200 and E 10. Thus,
1:3; x
max
0:79 and x
max
59:45. The optimal solution is
x
n
0:25, which leads to an optimal inventory policy of
T
n
2:5; p
n
19:06 and Q
n
79:33 that yields
n
108:0. Fig. 4
describes the prot and its derivative as a function of x. In this
case, we see that the prot is not a concave function, as its
derivative does not decrease monotonically over (0, 0.79]; how-
ever, the prot is still a strictly pseudo-concave function over the
same interval.
Example 3. o3n 7=2n 4 and x
max
0
Let
0
200; n 1 (for linear decrease of the demand in time),
0:04; c 12; h 1; K 4000 and E 10. Thus, 1:3;
x
max
0:81 and x
max
377:9. The prot is negative for all x,
and the optimal solution is not to order at all. Fig. 5 describes the
prot and its derivative as a function of x.
We now analyze the sensitivity of the optimal replenishment
period, T
n
; the optimal order quantity, Q
n
; the optimal selling
price, p
n
; and the maximal prot,
n
, to changes in the values of
model parameters:
0
, n, , c, h, K and E. Such changes characterize
-10000
-8000
-6000
-4000
-2000
0
2000
4000
x -3000
0
3000
6000
9000
12000
15000
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x
(x) '(x)
Fig. 3. (x) and '(x) in Example 1.
T. Avinadav et al. / Int. J. Production Economics 144 (2013) 497506 502
a dynamic environment in which numerous elements are subject
to change, including the cost of raw materials, procurement and
delivering the goods; interest rates the rm is charged (which
inuence the unit holding cost); consumer preference (which
inuences consumers' willingness to buy less fresh or more
expensive items); and the technology of packing and storing
(which may inuence the items' shelf-life). We carried out
sensitivity analysis on Examples 1 and 2, the two examples that
resulted in protable optimal pricing and inventory policies. The
deviation in percentage from the original value of each parameter
was measured for six deviations: 710%, 725% and 750%. In
each of the experiments we veried that the constraint xx
max
was satised. The results are summarized in Tables 1 and 2,
respectively.
The sensitivity of the objective function and the decision
variables to changes in each parameter value was dependent on
the values of the other parameters. For one set of parameter values
the objective function and the decision variables were relatively
sensitive to changes in the value of a single parameter, and for the
other set they were relatively not very sensitive. For instance, a
deviation of 50% in the purchasing cost per unit, c, caused only a
6.9% increase in
n
in Example 1, but the same deviation caused a
236.9% increase in Example 2. Increasing the shape parameter n,
which means a more moderate decrease of the demand rate for a
longer initial period, caused Q
n
and
n
to increase, which is
obvious. The direction of the inuence of n on both p
n
and T
n
was not consistent across parameter sets; in any case, however,
the inuence on p
n
was negligible in our examples. Increasing the
full shelf-life of an item, E, resulted in an increase of Q
n
,
n
, p
n
and
T
n
, owing to the negative inuence of perishability on the
demand rate.
In addition, we analyzed the sensitivity of Q
n
, p
n
and
n
to
deviations from the optimal normalized replenishment period, x
n
.
Such a deviation may occur as a result of constraints that require
the supplier to deliver fresh items in a replenishment period that
is shorter or longer than the optimal one. The results for Examples
1 and 2 are summarized in Table 3.
The results in Table 3 reveal that p
n
and
n
are not very sensitive
to deviations up to 725%; however, the deviation in Q
n
is more
sensitive, especially for negative deviations from x
n
.
8. Discussion and conclusions
The novelty of our model is in the formulation of a demand rate
that decreases polynomially over time after replenishment (up to
expiration, where it vanishes) and linearly in the selling price,
where the two factors have a multiplicative effect. We show that a
search for the optimal price, quantity to order and replenishment
period can be reduced to a search for the replenishment period
alone. In addition, we show that the protability of an inventory
system can be tested by computing the prot at one particular
point. A search interval for the optimal replenishment period is
provided, and we prove that the prot per unit time is strictly
pseudo-concave. This property enables us to use a relatively fast-
converging line-search algorithm for nding numerically the
optimal solution as closely as required. Sensitivity analysis of two
numerical examples shows that the sensitivity of the decision
variables to changes in the value of a single parameter is dependent
on the values of the other parameters. The only exception is the
optimal selling price, which is found to be insensitive to changes
-6000
-5000
-4000
-3000
-2000
-1000
0
x 0
1000
2000
3000
4000
5000
6000
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x
(x) '(x)
Fig. 5. (x) and '(x) in Example 3.
-400
-300
-200
-100
0
100
200
x -200
0
200
400
600
800
1000
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x
'(x) (x)
Fig. 4. (x) and '(x) in Example 2.
Table 3
Deviation in % from original value in Examples 1 and 2.
Example Deviation from x
n
in % Inuenced variable
50 25 10 10 25 50
1
46.3 21.8 8.4 7.8 18.5 33.1 Q
n
1.0 0.5 0.2 0.2 0.5 0.9 p
n
15.6 2.8 0.4 0.3 1.9 6.5
n
2
40.1 18.2 6.9 6.4 15.1 27.6 Q
n
1.4 0.7 0.3 0.3 0.7 1.3 p
n
29.8 4.6 0.6 0.5 2.4 7.5
n
T. Avinadav et al. / Int. J. Production Economics 144 (2013) 497506 503
of up to 50% in the value of a single parameter. This result can be
explained by Corollary 3, which restricts the maximal change in the
optimal selling price in our model.
In our model, in contrast to previous models of perishable
inventory, the expiration date is not only a constraint on the replen-
ishment period but a factor that inuences the demand over time, and
thus the prots. Current technology may increase shelf-life duration of
products by several means. The model enables us to evaluate whether
the monetary investment in those means is economically justied.
However, in order to calculate the optimal, prot-maximizing replen-
ishment period for a given product, our model requires knowledge of
the shape of the product's demand over time. In practice, to obtain the
value of this parameter, a manager must collect and analyze data on
the consumption behavior associated with the product at several
levels of freshness. The emergence of automated identication and
tracking technologies and network information systems may provide a
solution to this problem. Managers in food and pharmaceutical supply
chains might be able to use technologies such as radio frequency
identication (RFID) and other complementary techniques to track
products freshness as well as real-time consumption.
We encourage extending our model to t a retailer that sells a
variety of perishable products. Though some products have similar
prices and holding costs, each product is probably characterized by a
unique shape of demand over time. Applying the suggested metho-
dology to each one separately results in diverse replenishment
schedules. Such a policy induces high ordering costs, additional
logistical costs, and supplementary use of resources. A better and
probably more economical policy would be to jointly order several
products that are characterized by similar parameters and demand
shapes over time. Future work should investigate the effects of
relaxing some of our assumptions, e.g., allowing a nite replenishment
rate or backlogging, or considering a stochastic demand function.
However, such relaxations require additional assumptions, such as
how demand rate is determined for a product whose units on-shelf
have diverse expiration dates (according to their arrival times during
the non-instantaneous replenishment interval); or how backlogged
demand is inuenced by the time left until the next replenishment.
Since parameters such as demand, lead time and quality deterioration
can uctuate, our approach can be applied as an approximation only
when the uctuations are small. Future research is also required in
order to extend the current methodology to include other price- and
time-dependent demand functions, as well as stochastic expiration
time and demand.
Acknowledgments
The authors thank Professor Mordecai Henig for his useful
comments and helpful suggestions during the work on this paper,
and especially in the eld of pseudo-concavity analysis. The
assistance of the Asian-Pacic editor and two anonymous referees
is graciously acknowledged for suggestions that signicantly
improved this paper.
Appendix A. Bounding the replenishment period by positive
prot margin
By (4),
p; T
T
1
T
2
p; T
T
Tp; T
_ _
where
p; T
T
pchTp; T
By contradiction, if the optimal T satises T 4pc=h and
p; T40 then it also satises p; T=T o0. Such a T cannot
be optimal, proving the lemma.
Appendix B. Bounds on p
T
Let
wT
n 1
n 2
1
n
2n 1T=E
n
_ _
Since wT is a decreasing function of T over 0; E, then
1
4
o
1
2
n 1
n 2
wE wTow0
1
2
Combining this result with (9) yields that
1
c
2
hT
8
op
T
o
1
c
2
hT
4
Since T 40 and, by (10)
T
p
T
c
h
we obtain
1
c
2
op
T
o
1
c
2
p
T
c
4
which is equivalent to
1
c
2
op
T
o
2
1
c
3
Appendix C. The derivative of the prot with respect to x
The derivative of x with respect to x is
0
x
K
x
2
E
0
1p
x
p
x
cnx
n1
n1
hE
2
1
2n1x
n
n2
_ _ _ _
0
p
0
x
2p
x
c1
n 1x
n
n 1
hxE
2
n 22x
n
n 2
_ _
:
By (11)
2p
x
c1
hxE
2
n 1
n 2
n 22x
n
n 1x
n
so that
0
x
0
x
2
K
0
E
1p
x
p
x
cnx
n1
n 1
hx
2
E
2
1
2n 1x
n
n 2
_ _ _ _ _ _
Appendix D. The derivative of f x
The derivative of f x with respect to x is
f
0
x 1p
x
p
x
cnx
n
nx
n1
n1
p
0
x
hxE1n 1x
n
_ _
p
0
x
p
x
cnx
n1
n 1
hx
2
E
2
1
2n 1x
n
n 2
_ _ _ _
Using algebraic manipulations and
1
c2p
x
n 1x
n
n 1
hxE
n 22x
n
2n 2
0
(by (11)), we obtain
f
0
x 1p
x
hE2p
0
x
x1x
n
p
x
chxEnx
n
xp
0
x
1x
n
1
chxE:
T. Avinadav et al. / Int. J. Production Economics 144 (2013) 497506 504
Appendix E. The derivative of p
x
The derivative of p
x
with respect to x is
p
0
x
hE
2
n 1
n 2
1n
n 1 n1x
n
2n 1x
n
2
_ _
Since
n 1 n1x
n
n 1x
n
2
1
n 1x
n
nx
n
n 1x
n
2
is a positive, monotone increasing function of x over (0,1], then p
x
0
is a monotone decreasing function of x over (0,1], and satises
0 p
0
x
x 1 p
0
x
op
0
x
x 0 0:25hE
Thus, hE2p
0
x
40:5hE40 over (0,1].
Appendix F. Condition for pseudo-concavity
Claim. A differentiable real function on an open interval, whose
derivative changes from a positive sign to negative at most once, is
strictly pseudo-concave.
Proof. Let f x be dened on the interval a; b. According to
Denition 3.5.10 of Bazaraa et al. (2006), we have to show that for
each distinct x
1
; x
2
a; b with f
0
x
1
x
2
x
1
0 we get f x
2
of x
1
.
If f
0
x
1
40 then f
0
x
1
x
2
x
1
0 is satised only for x
2
ox
1
, so that
f x
2
of x
1
. If f
0
x
1
o0 then f
0
x
1
x
2
x
1
0 is satised only for
x
2
4x
1
, so that f x
2
of x
1
. If f
0
x
1
0 then x
1
is the unique
maximum point, and f x
2
of x
1
for every x
2
.
Appendix G. An algorithm for obtaining the optimal solution
Step 1: Set problem parameters c;
0
; K; h; E; n; and calculate
1
c=hE.
Step 2: If 3n 7=2n 4 set x
max
1; otherwise, extract
numerically from
n3
n2
nn1
2n2n1
n
_ _
0 (e.g. using the
bisection method with an initial search interval of [0,1]) and
set x
max
.
Step 3: Calculate p
xmax
1
c
2
hxmaxE
2
n1
n2
1
n
2n1xmax
n
_ _
and
x
max
0
1p
xmax
p
xmax
c
_ _
1
xmax
n
n1
_ _
hxmaxE
2
1
2xmax
n
n2
_ _ _ _
K
xmaxE
.
Step 4: If x
max
0 then the inventory system is not protable.
Go to Step 7.
Step 5: Solve f x 0 numerically (e.g. using the bisection
method with an initial search interval of [0,1]) for obtaining
x
n
, where f x
K
0
E
1p
x
_ _
_
p
x
cnx
n1
n1
hx
2
E
2
1
2n1x
n
n2
_ __
and
p
x
1
c
2
hxE
2
n1
n2
1
n
2n1x
n
_ _
.
Step 6: Obtain the optimal policy by setting T
n
x
n
E,
p
n
1
c
2
hx
n
E
2
n1
n2
1
n
2n1x
n
_ _
, and Q
n
0
1p
n
x
n
E 1
x
n
n
n1
_ _
.
Step 7: End.
References
Avinadav, T., Arponen, T., 2009. An EOQ model for items with a xed shelf-life and
declining demand rate based on time-to-expiry technical note. Asia-Pacic
Journal of Operational Research 26 (6), 759767.
Bahari-Kashani, H., 1989. Replenishment schedule for deteriorating items with time-
proportional demand. Journal of the Operational Research Society 40, 7581.
Baker, R.C., Urban, T.L., 1988. A deterministic inventory system with an inventory-
level-dependent demand rate. Journal of the Operational Research Society 39,
823831.
Balkhi, Z.T., Benkherouf, L., 2004. On an inventory model for deteriorating items
with stock dependent and time-varying demand rates. Computers and Opera-
tions 31, 223240.
Bazaraa, M.S., Sherali, H.D., Shetty, C.M., 2006. Nonlinear Programming, Theory and
Algorithms. John Wiley & Sons.
Benkherouf, L., 1995. On an inventory model with deteriorating items and
decreasing time-varying demand and shortages. European Journal of Opera-
tional Research 86, 293299.
Bose, S., Goswami, A., Chaudhuri, K.S., 1995. An EOQ model for deteriorating
items with linear time-dependent demand rate and shortages under ination
and time discounting. Journal of the Operational Research Society 46,
771782.
Burden, R.L., Douglas, J.F., Reynolds, A.C., 1985. Numerical Analysis, 3rd ed.. Weber &
Schmidt, Prindle.
Chang, C.T., Teng, J.T., Goyal, S.K., 2010. Optimal replenishment policies for non-
instantaneous deteriorating items with stock-dependent demand. International
Journal of Production Economics 123, 6268.
Chun, Y.H., 2003. Optimal pricing and ordering policies for perishable commodities.
European Journal of Operational Research 144, 6882.
Devangan, L., Amit, R.K., Mehta, P., Swami, S., Shanker, K., 2013. Individually rational
buyback contracts with inventory level dependent demand. International
Journal of Production Economics 142, 381387.
Donselaar, K.H., van, Broekmeulen, R.A.C.M., 2012. Approximations for the relative
outdating of perishable products by combining stochastic modeling, simulation
and regression modeling. International Journal of Production Economics 140,
660669.
Gallego, G., van Ryzin, G., 1994. Optimal dynamic pricing of inventory
with stochastic demand over nite horizons. Management Science 40,
9991020.
Guowei, H., Wang, S., Cheng, T.C.E., 2012. Optimal pricing and order quantity for the
newsvendor problem with free shipping. International Journal of Production
Economics 135, 162169.
Herbon, A., Spiegel, U., Templeman, J., 2012. Simulation study of the price
differentiation effect in a stochastic deteriorating inventory with heteroge-
neous consumers - Freshness sensitivity. Applied Economics 44 (24),
31013119.
Karlin, S., Carr, C.R., 1962. Prices and optimal inventory policy. In: Arrow, K.J., Karlin,
S., Scarf, H. (Eds.), Studies in Applied Probability and Management Science.
Stanford University Press, Stanford, CA, pp. 159172.
Kocabiyikoglu, A., Popescu, I., 2011. An elasticity approach to the newsvendor with
price sensitive demand. Operations Research 59 (2), 301312.
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. International Journal of Production Economics 136,
116122.
Mills, E.S., 1959. Uncertainty and price theory. Quarterly Journal of Economics 73,
116130.
Nahmias, S., 1982. Perishable inventory theory: a review. Operations Research 30
(4), 680708.
Musa, A., Sani, B., 2012. Inventory ordering policies of delayed deteriorating items
under permissible delay in payments. International Journal of Production
Economics 136, 7583.
Partt, J., Barthel, M., Macnaughton, S., 2010. Food waste within food supply chains:
quantication and potential for change to 2050. Philosophical Transactions of
the Royal Society B 365, 30653081.
Petruzzi, N.C., Dada, M., 1999. Pricing and the newsvendor problem: a review with
extensions. Operations Research 47 (2), 183194.
Polatoglu, L.H., 1991. Optimal order quantity and pricing decisions in single-
period inventory systems. International Journal of Production Economics 23,
175185.
Sarker, B.R., Mukherjee, S., Balan, C.V., 1997. An order-level lot size inventory model
with inventory-level dependent demand and deterioration. International
Journal of Production Economics 48, 227236.
Silver, E.A., 1981. A simple inventory replenishment decision rule for a linear trend
in demand. Journal of the Operational Research Society 30, 7175.
Silver, E.A., Peterson, R., 1985. Decision Systems for Inventory Management and
Production Planning, 2nd ed.. Wiley, New York.
Teng, J.T., 1996. A deterministic replenishment model with linear trend in demand.
Operations Research Letters 19, 3341.
Urban, T.L., 2005. Inventory models with inventory-level-dependent demand: a
comprehensive review and unifying theory. European Journal of Operational
Research 162, 792804.
Valliathal, M., Uthayakumar, R., 2011. Optimal pricing and replenishment policies of
an EOQ model for non-instantaneous deteriorating items with shortages.
International Journal of Advanced Manufacturing Technology 54, 361371.
Weatherford., L.R., Bodily, S.E., 1992. A taxonomy and research overview of
perishable-asset revenue management: yield management, overbooking, and
pricing. Operations Research 40 (5), 831844.
Whitin, T.M., 1955. Inventory control and price theory. Management Science 2, 6168.
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. International Journal of Production Economics 101,
369384.
T. Avinadav et al. / Int. J. Production Economics 144 (2013) 497506 505
You, P.S., 2005. Inventory policy for products with price and time-dependent
demands. Journal of the Operational Research Society 56, 870873.
You, P.S., Hsieh, Y.C., 2007. An EOQ model with stock and price sensitive demand.
Mathematical and Computer Modeling 45, 933942.
Young, L., 1978. Price, inventory and the structure of uncertain demand. New
Zealand Operations Research 6, 157177.
Zabel, E., 1970. Monopoly and uncertainty. Review of Economic Studies 37,
205219.
Zabel, E., 1972. Multiperiod monopoly under uncertainty. Journal of Economics
Theory 5, 524536.
T. Avinadav et al. / Int. J. Production Economics 144 (2013) 497506 506