See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/301756768

An EOQ model for perishable products with fixed shelf life under stochastic
demand conditions

Article in European Journal of Operational Research · December 2016

DOI: 10.1016/j.ejor.2016.04.036

CITATIONS READS

24 1,651

1 author:

Cinzia Muriana
University of Pittsburgh Medical Center (UPMC ITALY)
21 PUBLICATIONS 201 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Current research area View project

All content following this page was uploaded by Cinzia Muriana on 01 October 2017.

The user has requested enhancement of the downloaded file.

 European Journal of Operational Research 255 (2016) 388–396

Contents lists available at ScienceDirect

European Journal of Operational Research

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

Production, Manufacturing and Logistics

An EOQ model for perishable products with fixed shelf life under
stochastic demand conditions
Cinzia Muriana∗
Independent researcher

artic l e i nf o abstrac t

Article history: The paper presents a mathematical stochastic model for perishable open-dating foods including shortage
Received 5 October 2014 and outdating costs. The demand fluctuations have been taken into account modeling them through a
Accepted 20 April 2016
normal distribution, and their impact on the storage time has been studied considering the stochastic
Available online 27 April 2016
nature of such a parameter in turn. The quantification of perished products has been also addressed, de -
Keywords: termining the probability for a product of remaining in stock beyond the end of its Shelf Life. On the
Stochastic inventory model basis of such premises, the optimal set of parameters that minimize the total unit cost has been deter-
Deterioration process mined. A numerical application and a sensitivity analysis show the practical applicability of the proposed
EOQ model in the context of warehouse management, underlining managerial insights.
SL © 2016 Elsevier B.V. All rights reserved.
Open dating foods

1. Introduction and literature review
Inventory models for deteriorating items have attracted consid-
erable interest in recent decades. The problem of modeling the
deterioration process was firstly tackled by Ghare and Schrader
(1963) that developed an exponentially decaying inventory model.
They observed that certain commodities deteriorate with time by
a proportion which can be approximated by a negative exponen-
tial function of time. Successively, Covert and Philip (1973), con-
sidered a two parameters Weibull deterioration function. Since the
work of Ghare and Schrader (1963) and Covert and Philip (1973),
considerable works have be done on deteriorating inventory sys-
tems, which are summarized in Raafat (1991), that dealt with 70s
and 80s in relation to continuously deteriorating items. Goyal and
Giri (2001), extended the review at 90s. Nahmias (2011), presented
a review referred to fixed and random lifetime models. Finally,
Karaesmen, Scheller-Wolf, Deniz, and Civelek (2008), Li, Lan, and
Mawhinney (2010), and Bakker et al. (2012), considered the in-
ventory theory with regards to the latest results in such field. In
particular Bakker et al. (2012), proposed a review that extended
the literature to the work done in the field of perishable inven-
tory, since the 2001 until the 2012. Such literature considered the
two fundamental decisions that affect the inventory related costs,
in managing perishables: replenishment and issuance.
The question generally dealt in perishable inventory research
field is to determine the optimal batch to stock under either de-
terministic or stochastic conditions of the demand and possibly
∗
Tel.: +393290128446; fax: +39091424094.
E-mail address: cmuriana@live.it
considering constant or time-dependent deterioration (e.g., expo-
nential, Weibull or Gamma deterioration distribution) and shortage
costs (see for example Begum, Sahoo, & Sahu, 2010; Chang et al.,
2010; Ghosh & Chaudhuri, 2005; Khanra et al., 2010; Mahata and
Goswami, 2009; Manna & Chaudhuri, 2006; Mishra & Shah, 2008;
Mishra & Singh, 2011; Ou & Min, 2010; Roy & Chaudhuri, 2011;
Roy et al., 2011; Roychowdhury, 2009; Sana 2011, for determinis-
tic models and Aggoun, Benkherouf, & Tadj, 2001, Brander & Fors-
berg, 2006; Broekmeulen van Donselaar, 2007; Hala & EI-Saadani,
2006; Halim, Giri, & S., 2008; Hayya, Harrison, & Chatfield, 2009;
Søren & Hill, 2000, for stochastic ones). The literature mentioned
is particularly suited for fruits and vegetables. In this case, deterio-
rated units can be identified through manual inspections. However,
today emerging technologies such as the Radio Frequency Identifi-
cation (RFID) with embedded sensors allow us to perform auto-
matic inspections through the application of mathematical models
relying on time-temperature history of products. On the contrary,
such an approach is not well suited for “open dating” or canned
foods, like those managed at the retailing stage of the food chain
(typically iper and super markets), governed by the “use by date”
or “sell by date”. Such a date indicates that the products remain
safe and suitable for human consumption until the reaching of the
end of their Shelf Life (SL). Once this time is achieved, they are
to be considered completely perished and consequently discarded.
For that a reason, such products are considered having a deter-
ministic and fixed life. In this paper, the SL is addressed as the
remaining SL of products as they enter the retailing stage, namely
a portion of the total lifetime starting form the end of the pro-
cess/manufacturing cycle, as better explained in Kilcast and Subra-
maniam (2000).

http://dx.doi.org/10.1016/j.ejor.2016.04.036
0377-2217/© 2016 Elsevier B.V. All rights reserved.
 C. Muriana / European Journal of Operational Research 255 (2016) 388–396
Works found in literature do not consider the link between the
deterioration rate of products and their SL and mark as outdated
those products still in stock at the end of the cycle time (defined as
the time between consecutive batches, see Diwekar, 2014), without
verifying that the cycle time is greater or less with respect to the
SL of products. In practice, such an essential relationship is usu-
ally guaranteed avoiding of managing products with very short SL.
Thus, the outdating units are calculated as the amount of partially
deteriorated products remaining in stock at the end of the cycle
time. Such products are usually discarded at the end of the cycle
time, or someway recovered and sent to markets with less strin-
gent quality standards, at discounted prices. However, being the
cycle time less than the SL, such products should not be consid-
ered outdated, but still suited for the target market. In such con-
text, the approaches proposed in literature until now are not com-
pliant with supply chain goals, that advocate cost efficiency as a
requirement to satisfy. Moreover, while such approaches have a
low impact on the fresh products such as fruits and vegetables,
whose SL is uncertain and, as a result, it can be determined only
by manual or automatic inspections, it has instead a strong impact
on packaged products, with the result of loss of profit for the firm
carrying out such a policy.
Conversely, when the assumption of the overcoming of the SL
during the storage time is allowed (i.e., the cycle time is greater
than the SL), products in stock are to be considered deteriorated
and necessarily discarded.
The literature analysis allowed us to find only few papers in
which the SL of products is taken into consideration in the op-
timization function. This is the case of Avinadav and Arponen
(2009), that discussed an extension of the classical Economic Or-
der Quantity (EOQ) model for items with a fixed SL and a deter-
ministic declining demand rate due to a reduction in the quality
of the item in the course of its SL, and Yan (2012), that presented
an EOQ model for perishable items with freshness-dependent de-
mand. However, they neglect of considering the outdating units in
stock as the consequence of the case in which the cycle time is
greater than the SL. Therefore, the main shortcoming found in the
review of literature relates to the possibility that the SL of products
is overcome during the cycle time and, in such context, the main
contribution of the present paper is to propose an EOQ model for
perishable products, considering the SL as a distinctive parameter
to determine the amount of perished products still in stock. The
goal is to provide a mathematic model for supply chains managing
“open-dating” foods, which allows us to remove outdating costs re-
lated to the discard of still suitable products. The perishability of
products is addressed in terms of the probability that a product
remains in stock beyond the achievement of its SL.
Moreover, the literature analysis led to find EOQ models tak-
ing into consideration freshness-dependent demand (see for ex-
ample Avinadav & Arponen, 2009; Avinadav, Herbona, & Spiegel,
2013; Bai & Kendall, 2008; Piramuthu & Zhou, 2013; Yan, 2012). In
the present paper the demand variability resulting from consumer
preferences, based on weather conditions, freshness of products,
micro–macro economic factors etc., is taken into consideration by
enforcing the assumption of stochastic demand behaviour. The im-
pact of the stochastic demand behavior is determined not only in
terms of stockout costs and safety stock, but also considering the
storage time as a stochastic variable consequently. The stock on
hand satisfying a desire Cycle Service Level (CSL) (i.e. the probabil-
ity from available stock during the replenishment cycle (Li, 2007))
is fixed and the safety stock is determined, according to the de-
mand variability. Given the accepted risk (1-CSL) (i.e. the proba-
bility of having a stockout during the replenishment cycle) that
the market demand overcomes the stock on hand during the cycle
time, the stockout costs are calculated. Information about the SL
of products is employed to calculate the amount of shortages. Ac-
389
cordingly, the proposed model determines the optimal cycle time
and the quantity to be stocked on hand including the shortage and
deterioration costs.
2. Model formulation
The model here presented deals with products having a deter-
ministic and constant SL. In the case of deterministic demand be-
haviour such an assumption would lead to the purchasing of the
quantity sold within the time of the SL, i.e., EOQ = demand rate∗SL
and t1 = Tc = SL. Conversely, the stochastic behavior of the demand
here enforced leads two main results. First of all, it determines
the possibility of stockout occasions due to the demand variability.
Such problem can be partially overcome by stocking an additional
quantity of products, namely the safety stock, for a desired CSL.
Moreover, the demand variability leads that the time the products
are held in stock can be considered in all respects a random vari-
able as well, whose expected value is the SL of the product. The
second problem relates with the possibility of quantifying the per-
ished units in stock. Unlike the models usually present in litera-
ture, relying on the well known differential equation including the
deterioration rate θ (t) (see for example Mishra & Shah, 2008), we
propose of determining the amount of perished products as the
number of products remaining on hand beyond their SL, i.e., cal-
culating the probability that the products remain in stock beyond
the end of their SL. The inventory model relates to the determi-
nation of optimal parameters consisting in the optimal batch size,
the time the products are held in stock and cycle time, allowing at
minimizing the total unit management cost of products stocked.
2.1. Assumptions and notation
The model has been formulated by considering the following
assumptions and notation:
• The inventory system deals with one item.
• The item has a shelf life equal to SL, that is deterministic and
constant, meaning that one item can be sold until the ending
of the SL, when it must be considered completely perished and
discarded.
• Shortages are allowed and the demand completely lost.
• t1 is the time at which the inventory level drops to zero, a ran-
dom variable that must be optimized.
• The lead time (LT), i.e., the amount of time between the place-
ment of the order and its receipt is deterministic and constant.
Stated this assumption, the LT cannot impact the optimality
conditions of SL of products handled at the retailing stage, as
in fact if the LT is such that the remaining SL of products is less
that that established, the products are rejected and sent back
to the wholesaler. This is consistent with the contractual poli-
cies between retailers and wholesalers, in which the latter ac-
tor have to guarantee a specified remaining SL to the first one
at the time at which the products enter the retailing stage (see
Garrone et al., 2012). Nevertheless, the LT will be considered in
the model presented.
• The expected demand rate is assumed to be a random variable
normally distributed with mean μt and variance σ t 2 , referred
to the time unit t. During the LT, the demand is equal to μtLT.
• The safety stock is considered in order to ensure a fixed CSL
and it is equal to F−1(CSL)σ t . In the case of normally dis-
tributed demand, such a quantity can be approximated by kσ t ,
where k is the safety factor (see Alstrøm, 2001; Annadurai &
Uthayakumar, 2010; Brander & Forsberg, 2006; Chung, Ting, &
Hou, 2009). During the LT, the safety stock is equal to kσ tLT.
• A is the setup cost per order.
• h is the inventory holding cost per unit held in stock per unit
time.
390 C. Muriana / European Journal of Operational Research 255 (2016) 388–396

Fig. 2. Case 1.

Fig. 1. The inventory model.

• Co is the outdating cost per unit.

• Cs is the shortage cost per unit per unit time.
• Tc is the cycle time, a variable which must be optimized.
• At the beginning of every period the initial stock level is raised
to the order level, The assumption of deterministic and con-
stant LT leads that a new order is to be placed at the as the
inventory level drops to the reorder point (ROP), equal to: ROP
= (μt ∗ LT)+(k∗ σ t ∗ LT).
Fig. 1 reports a qualitative image representing the inventory
model discussed.
Fig. 3. Case 2.
2.2. Mathematical model

For the purpose of the model it is assumed that the inventory and Tc (see Fig. 3) plus the quantity perished in stock between SL
level is equal to the EOQ at t = 0. The inventory level depletes over and t1 .
time until it drops to zero at certain time t1 . Once the time t1 is a random variable, it is necessary to de-
The differential equation governing the inventory depletion is: termine the probability for t1 to be less or greater than SL. Such
probability is equal to:
dQ (t )
+ SS = −μt 0 ≤ t ≤ t1 (1) P(t1 ≤ SL) (7)
dt
where SS indicates the safety stock.
For normally distributed demand equation (1) can be written P(t1 > SL) = 1 − P(t1 ≤ SL) (8)
as: The expected inventory on hand is equal to:
∫ ∫
Q (t ) = − ∫ μt + kσt dt (2) t1 t1 t12
EIH = Q (t )dt= rt (t1 − t )dt = rt (9)
The solution of the differential equation (2) is: 0 0 2
Q (t ) = −(μt + kσt )t + c = −rt t + c (3) The expected outdating can be determined as:
. ∫
Σ
SL
By considering the boundary condition:
EO = Q (0 ) − rt dt P(t1 > SL) = rt (t1 − SL)P(t1 > SL) (10)
0
Q (t1 ) = 0 (4)
corresponding to the quantity of products that deteriorate between
we obtain:
t = t1 and t = SL if t1 > SL.
Q (t ) = rt (t1 − t ) (5) Between t1 and Tc the differential equation (1) becomes:
dQ (t )
= − μt t1 ≤ t ≤ Tc (11)
Q (0) = rt t1 (6) dt
Because of the demand variability, the time t1 is in all respects meaning that between t1 and Tc the inventory depletion is due
a random variable following the same distribution of the expected only to the consumer demand.
demand, as can be derived from Eq. (6). Q (t ) = −μt (t1 − t ) t1 ≤ t ≤ Tc (12)
Two distinct cases are to be considered:
The expected shortage can be consequently determined as:
Case 1: t1 ≤ SL. In such a case the inventory drops to zero be-
∫ t1 ∫ TC ∫ t1
fore the ending of the SL of the products. The products do not de-
teriorate in stock. The shortage is equal to the quantity of demand ES = − Q (t )dt− Q (t )dt= − rt (t1 − t )dtP(t1 > SL)
SL t1 SL
lost between t1 and Tc (see Fig. 2). ∫ Tc rt (t − SL)2
P(t1 > SL) + μt (Tc2− t1 )
2
Case 2: t1 > SL. In such a case the products remaining in stock − μt (t1 − t ) = dt 1

between SL and t1 are considered perished and no further saleable. t1 2

The shortage is equal to the quantity of demand lost between t1 (13)
 C. Muriana / European Journal of Operational Research 255 (2016) 388–396 391

Eq. (13) underlines that the shortage is due to the demand lost be- By posing:
tween t1 and Tc plus the demand lost between SL and t1 , in the dETCu
case in which t1 > SL. In fact, in such period the products stocked
dTc =0 (21)
are considered perished and no further saleable. Moreover, such
equation highlights that the ES is zero only if t1 = SL = Tc (if the The cost function is always convex and the optimality condition is:
demand variability is equal to zero). In such a condition the pro-
, . . ΣΣ
posed model returns the traditional EOQ model firstly proposed by Tc = ±1/2/Cs /μt 2 Cs μt 4A − 2Cort SL + Cs rt SL2 (22)
Wilson (1934).
The negative root of the (22) cannot be accepted as it violates the
The Expected Total Cost is:
position (17).
For t1 = Tc Eq. (18) becomes:
ETC = A + EIHC + EOC + ESC = A + h ∗ EIH + Co ∗ EO + Cs ∗ ES
(14)
TCu =A + EIHC +TcEOC + ESC
The total cost per unit time is:
. Σ
A + hrt 2Tc + Co rt (T2c −SL) + C rt (Tc −SL)2
2
A + EIHC + EOC + ESC s 4
ETCu =
Tc . = Tc (23)
2
> SL) + C rt ( t1 −SL)2
P(t Σ
A + hr t1 + C r (t − SL)P (t > SL) + μt (T2c −t1 )
2

t 2 ot 1 1 s 2 1
= Tc dETC u d . Σ
= A + EIHC + EOC + ESC
(15) dTc dTc Tc
with hrt Tc + 1 Cort + 1 Csrt (Tc − SL)
2 2
=
Tc
0 ≤ t1 ≤ Tc (16)
A + 21 hrt cT 2 + 21 Co tr (Tc − SL) +41 C c − SL)2
s tr (T
and − 2 (24)
Tc
Tc ≥ 0 (17) By posing:
The optimal condition minimizing the ETCu can be found by de- dETCu
riving equation (15) with respect to t1 and Tc. The possibility of dTc =0 (25)
finding an analytic closed form solution of (15) depends on the The cost function is always convex and the optimality condition is:
probability distribution of t1 .
It can be simply seen that for symmetrical distributions like
, . Σ
Tc = ±1/(C srt + 2hr t ) r t (Cs + 2h ) C sr t SL2 + 4A − 2Cor tSL (26)
the normal one, the quantity P(t1 ≤ SL) is simply determined being
equal to 0.5. In this case, an analytic closed form of the (15) can be The negative root of the (26) cannot be accepted as it violates the
found as will be illustrated in the following Section 3. As the t1 dis-
position (17).
tribution depends on the demand distribution, the proposed model
Eq. (26) underlines that for t1 = Tc the shortage costs are only
can be considered of general kind, as in fact both discrete and con-
related to the possibility of overcoming the SL during the frame
tinue demand distributions can ultimately be approximated with a
time 0- t1.
normal one as the horizon of observation increases.
For 0 < t1 < Tc Eq. (18) leads to:

3. Cost function analysis dETCu Tc − t1

= Cs tμ
T
dTc c
Under the hypothesis of symmetric distribution Eq. (15) be- . Σ
−A + + − SL ) + Cs 14rt (t1 − SL )2 + 1 μ
1
2 hrt t 21 12
Cort (t1 2 t (Tc − t1 )2
comes:
Tc2
ETCu = A + EIHC +TEOC + ESC (27)
c
. Σ .1 Σ
t 2 r (t −SL) rt (t1 −SL)2 μt (Tc −t1 )2
dETCu hrt t1 + 12Cort + Cs rt (t1 − SL ) − μt (Tc − t1 )
A+ hrt 12 + Co t 1 2 + Cs + 2
dt1 =
4 2 (28)
= (18) Tc
Tc
The optimality conditions are:
Considering position (16), it can be shown that for t1 = 0 or t1 =
dETC u
Tc, Eq. (15) is simply to solve via analytic form.
=0 (29)
For t1 = 0 in (18) we have: dTc
dETCu
ETCu = A + EO + ESC =0 (30)
Tc dt1
. Σ
Cort SL rt SL2 μt Tc 2
A− 2 + Cs 4 + 2
Eqs. (29) and (30) highlight that an optimal solution depends on
= (19) the simultaneous optimization of the two decisions variables t1
Tc and Tc. Since it is not possible to solve such equations explicitly
. Σ for t1 and Tc, an iterative procedure will be used, based on Chiu
dETC u d A + EO + ESC (1995b):
=
dTc dTc Tc
. Σ 1. Pose SL = t1 = Tc and solve Eq. (18) determining the Q(0). In
Cort SL 1
rt SL2 + 1 μt Tc2 this case Eq. (18) turns into the traditional EOQ.
4
= Csμt − A − + Cs 2 (20) 2. Set the number of iterations i equal to N.
2 Tc 2 3. For i = 1.
392 C. Muriana / European Journal of Operational Research 255 (2016) 388–396

Table 1
input variables of the experimental plan.

Configuration Input variables Range variation μt k σt h Cs Co A SL LT

1 μt 30–44.323 – 1.65 5 0.001 0.02 0.01 250 30

2 k 1.2–1.772 30 – 5 0.001 0.02 0.01 250 30
3 σt 4.5–6.648 30 1.65 – 0.001 0.02 0.01 250 30
4 h 0.0008–0.001,182 30 1.65 5 – 0.02 0.01 250 30
5 Cs 0.015–0.022,162 30 1.65 5 0.001 – 0.01 250 30
6 Co 0.008–0.01,182 30 1.65 5 0.001 0.02 – 250 30
7 A 230–339.814 30 1.65 5 0.001 0.02 0.01 – 30
8 SL 25–36.936 30 1.65 5 0.001 0.02 0.01 250 –
9 LT 1–1.551,328 30 1.65 5 0.001 0.02 0.01 250 30 –

4. Fixe t1 = Q(0)/rt. Table 2

output variables of the
5. for the value of t1 determine the optimal Tc, solving Eq. (29).
experimental plan.
6. For the value of Tc determined the optimal t1 , solving
Eq. (30). Output variables

7. Pose i = i+1. t1 (percent)

8. While i ≤ N do steps 4-5. Tc (percent)
9. When at the iteration i+1 the values of t1 and Tc are the EOQ(percent)
ETCu (percent)
same of the iteration i, terminate the procedure.
EIHC(percent)
10. If the number of iterations N is reached before the algorithm EOC(percent)
converges, increase the maximum number of iterations. ESC(percent)

Section 4 reports a numerical example in which the solution

has been determined through a Visual Basic macro that iterates the
algorithm of Chiu (1995b). (b) The increase of the EOQ with consequent increase of the
EIHC and the EOC.
4. Numerical example (c) The increase of the ETCu.

In this section a numerical application is illustrated to confirm
the practical applicability of the proposed model that can be taken
as drive of decision-making process in the warehouse management
field.
The input data are: A = 250€, h = 0.001€/unit∗unit time, Cs =
0.02€/unit∗unit time, Co = 0.01€/unit. The safety factor k is equal
to 1.65 (corresponding to a CSL of 0.95). The SL of the products is
equal to 30 unit time. The demand is normally distributed with μt
= 30 unit/unit time and σ t = 5 unit/unit time. The LT is equal to 1
unit time. In such conditions the solution is: t1 = 42.5052 unit time,
Tc = 53.5058 unit time, ETCu = 6.60,034€ and EOQ = 1625.83 units.
The result was achieved in 15 iterations.
5. Sensitivity analysis
In this section a sensitivity analysis is proposed to show the
impact of the variation of input factors, namely the SL, the LT, the
costs and the demand values, on the optimal ETCu, t1 , Tc and the
EOQ. For the purpose of the model, the input factors are indepen-
dent variables, while t1 , Tc, EOQ and ETCu are dependent variables
of the model. Of independent variables, μt, σ t and SL are out of
the control of supply chain’s actors, as well as the costs, while
the safety factor k is dependent on supply chain’s actor choices,
in relation to the accepted risk of stockout. For each configura-
tion of the sensitivity analysis each input factor has been varied
of the 5 percent with respect to the initial value, while other fac-
tors have been kept constant and equal to the base case reported
in Section 4. The input variables of the experimental plan designed
are reported in Table 1, while Table 2 reports the output variables
of the experimental plan. Finally, the results are graphically sum-
marized in Table 3.
The following considerations can be done on the basis of the
results of Table 3:
The increase of the μt and k values causes (Figs. 4 and 5):
(a) The decrease of the optimal t1 and Tc with consequent de-
crease of the ESC.
This means that the increase of the demand and safety fac-
tors has a negative effect on deteriorated units, that increase,
and a positive effect on shortages, that decrease. Because of the
contemporary increase of the EOQ, the total unit cost increases
as well. The quantity mainly influenced by the increasing of the
input parameter under observation is the EOQ. The considerations
done above are to be reversed in the case of demand decrease.
Being the consumer behaviour quite independent from the firm’s
strategy, the managerial insight that can be arisen is that the
retailer has to pay attention on demand variations that should be
accurately counteracted by providing suitable forecasting methods
in order to prevent deteriorated units in stock.
The increase of the σ t value causes (Fig. 6):
(a) The decrease of the optimal t1 and Tc with consequent de-
crease of the EOC and ESC.
(b) The increase of the EOQ with consequent increase of the
EIHC.
(c) The increase of the ETCu.
This means that the increase of the demand variability has an
overall positive effect on the stock management, as it contributes
to decrease the entity of deteriorated units and shortages. The si-
multaneous increase of the quantity stocked leads to the increase
of the total unit cost. The quantity mainly influenced by the in-
crease of the input parameter under observation is the quantity
stocked. The considerations done above are to be reversed in the
case of demand variability decrease. The managerial insight that
can be arisen is that retailers are to pay attention to the demand
variability, as it can directly affect outdated units and shortages.
The increase of the h value causes (Fig. 7):
(a) The decrease of the optimal t1 and Tc with consequent de-
crease of the EOC, ESC and the EOQ.
(b) The increase of the EIHC and the ETCu.
This means that the increase of the holding cost has an over-
all positive effect on the stock management as it contributes to
 C. Muriana / European Journal of Operational Research 255 (2016) 388–396 393

Table 3
Results of the sensitivity analysis.
394 C. Muriana / European Journal of Operational Research 255 (2016) 388–396
decrease the entity of deteriorated units and shortages. The over-
all stocked quantity decreases. In fact, even if the variation of the
holding cost is positive in mean (4.2497 percent), such an increase
is less than the increase of the input factor under consideration
(which is of the 5 percent for each step of the sensitivity anal-
ysis), thus implying an overall decrease of the stocked quantity.
The increase of the inventory holding cost leads that the total unit
cost has an overall increase. The considerations done above are
to be reversed in the case of holding cost decrease. The manage-
rial insight that can be arisen is that retailers are to pay partic-
ular attention to holding cost variation as they involve the incre-
ment/decrement of the outdating and shortage costs. In particu-
lar, in presence of holding costs increase they should decrease the
batch size (and consequently increase the frequency of orders), in
order to limit storage and shortage costs.
The increase of the Cs value causes (Fig. 8):
(a) The decrease of the optimal t1 and Tc with consequent de-
crease of the EOQ and the EOC.
(b) The decrease of the EIHC due to the decrease of the EOQ.
(c) The decrease of the ESC.
(d) The increase of the ETCu.
This means that the increase of the shortage cost has an overall
positive effect on the stock management as it contributes to de-
crease the entity of the deteriorated units and stockouts, even if
the decrease of stockout (−2.3797 percent) is less than the increase
of the input factor under consideration (which is of the 5 percent
for each step of the sensitivity analysis). However, the simultane-
ous decrease of the optimal cycle time causes that the total unit
cost tends to increase. The quantity mainly influenced by the in-
crease of the input parameter under observation is the outdating
cost followed by the stockouts. The considerations done above are
to be reversed in the case of shortage cost decrease. Of course, by
decreasing the entity of EOQ , it is less probably that higher deteri-
oration costs are encountered. The managerial insight that can be
arisen is that retailers are to pay particular attention to shortage
cost variation as they involve the variation of the outdating, con-
sequently. In particular, in presence of shortage costs increase they
should decrease the batch size (and consequently increase the fre-
quency of orders), in order to limit shortage costs.
The increase of the Co value causes (Fig. 9):
(a) The decrease of the optimal t1 and Tc with consequent de-
crease of the EOQ.
(b) The decrease of the EIHC due to the decrease of the EOQ.
(c) The decrease of the ESC due to the decrease of Tc.
(d) The increase of the EOC and ETCu.
This means that the increase of the outdating cost has an over- all
positive effect on the stock management, as it contributes to de-
crease the entity of deteriorated units and stockouts. In fact, even
if the variation of the outdating is positive in mean (4.6269 per-
cent), indicating that the outdating cost increases, such an increase is
less than the increase of the input factor under consideration (which
is of the 5 percent for each step of the sensitivity analysis), thus
implying an overall decrease of the outdating quantities. The
considerations done above are to be reversed in the case of outdat-
ing cost decrease. The model shows that in presence of increase of
outdating costs, the stock management policy is to be prudent, in-
volving a less quantity is to be stocked on hand. However, since the Tc
decreases, such behaviour implies an overall stockouts decrease. The
increase of the A does not have any influence on the opti-
mal values.
The increase of the SL value causes (Fig. 10):
(a) The increase of the optimal t1 and Tc with consequent in-
crease of the EOQ.
(b) The increase of the EIHC due to the increasing of the EOQ.
(c) The decrease of the ESC.
(d) The decrease of the EOC and ETCu.
This means that the increase of the SL factor has an overall pos-
itive effect on the stock management as it contributes to decrease
the entity of deteriorated units and shortages. The simultaneous
increase of the cycle time causes that the total unit cost has an
overall decrease. At the same time, the increase of t1 leads an over-
all increase of the EOQ. The quantities mainly influenced by the
increase of the input parameter under observation are shortages,
outdating units and stocked units. The considerations done above
are to be reversed in the case of SL decrease. The SL variation re-
flects the application of the model to different products having dif-
ferent SL. The managerial insight that can be arisen is that retail-
ers are to pay particular attention to products having shorter SL,
as they easily deteriorate in stock, leading to the increase of the
outdating costs, consequently.
The increase of the LT value causes (Fig. 11):
(a) The decrease of the optimal t1 and Tc with consequent increase
of the EOQ.
(b) The increase of the EIHC due to the increase of the EOQ.
(c) The decrease of the EOC and ESC.
(d) The increase of the ETCu.
This means that the increase of the LT factor has an overall
positive effect on the stock management as it contributes to de-
crease the entity of deteriorated units and shortages. The simul-
taneous decrease of the cycle time causes that the total unit cost
has an overall increase. The quantities mainly influenced by the in-
crease of the input parameter under observation are EOQ, the to-
tal cost and the storage cost. The considerations done above are
to be reversed in the case of LT decrease. The LT variation reflects
different contractual conditions of supplying, and its optimization
is of fundamental importance in order to minimize the entity of
stocked products. The managerial insight that can be arisen is that
the LT can impact the optimal batch size and thus, managers have
to pay attention to contractual conditions, especially for products
with shorter SL.
The sensitivity analysis shows that the optimal values are not
very sensitive; this implies the robustness of the solution provided
by the model.
The applicability of the proposed model relates with several
practical contexts such as, e.g., the replenishment cycle of food
chain at the retailing stage (especially iper-super markets) where,
because of the products seasonality and the variable consumer
preferences, the demand cannot be considered always determinis-
tic and constant. Moreover, in such contexts the products are con-
sidered as having a fixed and deterministic SL corresponding to the
“best before” or “use by date” time, and the regulations relating to
the food safety require that such products be considered suitable
for human consumption only within their SL and that they be dis-
carded when such time is reached. In such cases, the model pro-
posed allows the optimization of stock management parameters by
determining the optimal quantity to be stocked, the optimal cycle
time and storage time allowing at minimizing the total unit cost.
6. Conclusions
The present paper addressed the topic of EOQ for perishable
products. The study starts from the consideration that the perisha-
bility of products is to be put in relation with their SL. Moreover,
in determining the optimal quantity to be stocked on hand, the
current literature neglects of verifying that the SL has not been
overcome during the cycle time. This leads that products remain-
ing on hand at the end of the cycle time are considered outdated
and consequently disposed off or rather sent to alternative markets
 C. Muriana / European Journal of Operational Research 255 (2016) 388–396
with less stringent quality standards. The paper presented aims
at overcoming such gaps, proposing a stochastic inventory model
for deteriorating items including shortage and outdating costs. The
main assumption of the model relates to the demand rate that is
assumed to be a random variable following a normal distribution.
The perishability of products is taken into account and the prob-
ability for a product of staying still in stock beyond the end of
the SL is determined. Shortages are allowed and the related units
completely lost. In particular, if the SL is greater than the time at
which the inventory drops to zero, shortages are determined only
based on the demand rate. On the contrary, if the SL is less than
the time at which the inventory drops to zero, the model provides
the quantity of outdated products and shortages, that also include
the time frame in which the demand cannot be satisfied cause the
presence of deteriorated units in stock. Such conditions represent
a common hypothesis in the practical contexts of perishable
management in which the product deterioration and the demand
fluctuations have to be taken into consideration in optimizing the
replenishment cycle. The lead time has been considered determin-
istic and constant, meaning that it cannot impact the optimality
conditions, as it can only reduce the time frame available for the
wholesaler for delivering the products to the retailer. The results
show that a solution can be found representing the quantity to
be stocked on hand and the corresponding cycle time minimizing
the total cost incurred. The paper aims at showing that the per-
ishability of products can be put in relation with the SL and that
SL itself has to be monitored during the cycle time, in order to
ensure that only products beyond their SL are considered perished
and consequently discarded at the end of the cycle time. Unlike
models present in literature, the paper shows that the SL is a key
parameter in discerning about the product destination (sell it at
discounted prices as it has still a residual life, or dispose it off) and
the shortage determination. In this context, considering SL allows
us to carry out a more conscious decision-making process, as the
impact of such parameter on the warehouse management, for
given market demand, is taken into account. In fact, open-dating
products are suited for human consumption until the reaching of
the end of their SL. Thus, the model proposed allows us to remove
useless discarding costs arising from the neglecting of the SL.
Moreover, the model shows that the batch size can be more prop-
erly set in presence of information about the SL, as in fact taking
into account shortages due to the presence of outdated units in
stock (products that overcome the SL during the cycle time), allows
us to increase the EOQ to tackle such phenomenon. Managerial in-
sights have been derived from the sensitivity analysis, that allows
us to improve the retailer strategy depending on the cost/demand
parameter variations. Further studies in this context may relate
the introduction of a random SL, the optimization with respect to
the safety factor k and the introduction of stochastic lead time,
studying their impact on the optimality conditions. Moreover,
the impact of the price-discount policies on the demand rate of
products near to the end of their SL should be also investigated,
as the decrease on prices could lead to the demand increase, thus
contributing to the reduction of the deterioration costs.
References
Aggoun, L., Benkherouf, L., & Tadj, L. (2001). On a stochastic inventory model with
deteriorating items. International Journal of Mathematics and Mathematical Sci-
ences, 25(3), 197–203.
Alstrøm, P. (2001). Numerical computation of inventory policies. based on the
EOQ/σ x value for order point systems. International Journal of Production Eco-
nomics, 71(1-3), 235–245.
Annadurai, K., & Uthayakumar, R. (2010). Reducing lost sales rate in (T, R, L) in-
ventory model with controllable lead time. Applied Mathematical Model, 34,
3465–3477.
Avinadav, T., & Arponen, T. (2009). An EOQ model for items with a fixed shelf-life
and a declining demand rate based on time-to-expiry. Asia-Pacific Journal of Op-
erational Research, 26(6), 759–767.
395
Avinadav, T., Herbona, A., & Spiegel, U. (2013). Optimal inventory policy for a per-
ishable item with demand function sensitive to price and time. International
Journal of Production Economics, 144(2), 497–506.
Bai, R., & Kendall, G. (2008). A model for fresh produce shelf space allocation and
inventory management with freshness condition dependent demand. INFORMS
Journal on Computing, 20(1), 78–85.
Bakker, M., Riezebos, J., & H Teunter, R. (2012). Review of inventory systems
with deterioration since 2001. European Journal of Operational Research, 221(2),
275–284.
Begum, R., Sahoo, R. R., & Sahu, S. K. (2010). An EOQ model for deteriorating items
with Weibull distribution deterioration unit production cost with quadratic de-
mand. Applied Mathematical Sciences, 4(6), 271–288.
Brander, P., & Forsberg, R. (2006). Determination of safety stocks for cyclic schedules
with stochastic demands. International Journal of Production Economics, 104(2),
271–295.
Broekmeulen, R., & Van Donselaar, K. (2007). A replenishment policy for a perish-
able inventory system based on estimated aging and retrieval behavior. In BETA
working paper nr (p. 218).
Chang, C. T., Chen, Y. J., Tsai, T., & Wu, S. (2010). Inventory models with stock- and
price-dependent demand for deteriorating items based on limited shelf space.
Yugoslav Journal of Operations Research, 20(1), 55–69.
Chiu, H. ,N. (1995). An approximation to the continuous review inventory model
with perishable items and lead time. European Journal of Operational Research,
87, 93–108.
Chung, K., Ting, P., & Hou, K. (2009). A simple cost minimization procedure for the
(Q, r) inventory system with a specified fixed cost per stockout occasion. Applied
Mathematical Modelling, 33(5), 2538–2543.
Covert, R. P., & Philip, G. C. (1973). An EOQ model for items with Weibull distribu-
tion deterioration. AIIE Transactions, 5, 323–326.
Diwekar, U. (2014). Batch processing: modeling and design. CRC Press: Taylor and
Francis Group, NY.
Ghare, P. N., & Schrader, G. F. (1963). A model for exponentially decaying inventory.
Journal of Industrial Engineering, 15, 238–243.
Ghosh, S. K., & Chaudhuri, K. S. (2005). An EOQ model for a deteriorating item with
trended demand, and variable backlogging with shortages in all cycles. Advanced
Modeling and Optimization, 7(1), 57–68.
Goyal, S. K., & Giri, B. C. (2001). Recent trend in modeling of deteriorating items.
European Journal of Operational Research, 134(1), 1–16.
Hala, A. F., & EI-Saadani, E. (2006). Constrained single period stochastic uniform
inventory model with continuous distributions of demand and varying holding
cost. Journal of Mathematics and Statistics, 2(1), 334–338.
Halim, K. A., Giri, K. S. Chaudhuri, & S, K. (2008). Fuzzy Economic Order Quan-
tity model for perishable items with stochastic demand, partial backlogging
and fuzzy deterioration rate. International Journal of Operational Research, 3(1/2),
77–96.
Hayya, J. C., Harrison, T. P., & Chatfield, D. C. (2009). A solution for the intractable
inventory model when both demand and lead time are stochastic. International
Journal of Production Economics, 122(2), 595–605.
Karaesmen, I., Scheller-Wolf, A., Deniz, B., & Civelek, I. (2008). Inventory manage-
ment of perishables: new results. INFORMS MSOM 2008. conference proceedings.
June 5-6, 2008. College Park, MD: University of Maryland.
Khanra, S., Sana, S. S., & Chaudhuri, O. K. (2010). An EOQ model for perishable item
with stock and price dependent demand rate. International Journal of Mathemat-
ics in Operational Research, 2(3), 320–335.
Kilcast, D., & Subramaniam, P. (2000). The stability and shelf life of food, Cambridge,
UK: CRC Press, Woodhead Publishing Limited.
Li, L. (2007). Supply Chain Management. Concepts, techniques and practices. USA:
Word scientific publishing.
Li, R., Lan, H., & Mawhinney, J. R. (2010). A review on deteriorating inventory study.
Journal of Service Science and Management, 3, 117–129.
Mahata, G. C., & Goswami, A. (2009). Fuzzy EOQ models for deteriorating items with
stock dependent demand and non-linear holding cost. International Journal of
Applied Mathematics and Computer Sciences, 5(2), 94–98.
Manna, S. K., & Chaudhuri, K. S. (2006). An eoq model with ramp type demand rate,
time dependent deterioration rate, unit production cost and shortages. European
Journal of Operational Research, 171(2), 557–566.
Mishra, P., & Shah, N. (2008). Inventory Management of time dependent de-
teriorating items with salvage value. Applied Mathematical Sciences, 2(16),
793–798.
Mishra, S., & Singh, P. K. (2011). Computational approach to EOQ model with power
form stock-dependent demand and cubic deterioration. American Journal of Op-
erations Research, 1, 5–13.
Nahmias, S. (2011). Perishable inventory systems. Springer. International Series in op-
erations research & management science. New York Dordrecht Heilderberg Lon-
don.
Ou, J., & Min, J. (2010). Note on EOQ model with stock-dependent demand. In ICEE
(pp. 4196–4199).
Raafat, F. (1991). Survey of literature on continuous deteriorating inventory models.
Journal of Operational Research, 42(1), 27–37.
Piramuthu, S., & Zhou, W. (2013). RFID and perishable inventory management with shelf-
space and freshness dependent demand. International Journal of Production
Economics, 144(2), 635–640.
Roy, T., & Chaudhuri, K. S. (2011). An inventory model for Weibull distribution de-
terioration under price-dependent demand and partial backlogging with oppor-
tunity cost due to lost sales. International Journal of Modelling, Identification and
Control, 13(1/2), 56–66.
396 C. Muriana / European Journal of Operational Research 255 (2016) 388–396
Roy, M. D., Sana, S. S., & Chaudhuri, K. (2011). An optimal shipment strategy for
imperfect items in a stock-out situation. Mathematical and Computer Modelling,
54(9–10), 2528–2543.
Roychowdhury, S. (2009). An optimal ordering policy for a stochastic inventory
model for deteriorating items with time-dependent selling price. Advanced Mod-
eling Optimization, 11(3), 289–304.
Sana, S. S. (2011). Price-sensitive demand for perishable items – an EOQ model.
Applied Mathematics and Computation, 217(13), 6248–6259.
View publication stats
Søren, G., & Hill, R. (2000). The (r.Q) control of a periodic-review inventory system
with continuous demand and lost sales. International Journal of Production Eco-
nomics, 68(3), 279–286.
Wilson, R. H. (1934). A scientific routine for stock control. Harvard Business Review,
13, 116–128.
Yan, X. (2012). An EOQ model for perishable items with freshness-dependent de-
mand and partial backlogging. International Journal of Control and Automation,
5(4), 19–38.