Excel Based Monte Carlo Simulation for the (Q,r) Inventory Control Model

Banhan Lila*, Adisak Nowneow and Sanya Yimsiri

Department of Industrial Engineering, Faculty of Engineering, Burapha University,
Saensuk, Mueang, Chonburi, 20131, Thailand
*Corresponding Author E-mail: blila@eng.buu.ac.th
Received: Dec 23, 2021; Revised: Mar 29, 2022; Accepted: May 09, 2022

Abstract
In supply chain management, inventory plays a key role to deal with demand and supply uncertainty aiming to
guarantee a smooth flow of materials and products along the chain. This paper focuses on determining the suitable
values of Q and r in the (Q,r) inventory control policy model when it was applied to the situation of nonstationary but
known to be empirically discretely distributed of product demands and lead times. The total cost ( TC) and service level
(SL) were used to measure the policy performance using an Excel-based Monte Carlo Simulation (MCS) approach
with a set of actual historical demand data from an automotive tire service store. Results from the MCS indicated that
the (Q,r) model could lead to 12.36% lower TC with 2.31% higher SL, on average, when the Q and r values were
determined based on empirical discrete distribution compared to that of normal distribution. Therefore, the empirical
discrete distribution of demand and lead time should be utilized in a situation where the assumption of norm al and
other traditional distributions is invalid.
Keywords: Inventory Control Policy, Empirical Discrete Demand Distribution, Excel based Monte Carlo Simulation,
Automotive Tires

1. Introduction Numerous evidences from both practices and

As a member of the supply chain, a retailer buys researches [5] have indicated that difficulty in selecting
products from upstream members and sells them to its the suitable inventory control policy directly relates to
customers at a certain retail price. The difference between variation, especially for the circumstance with unknown
the retail price and cost that includes unit operations cost and nonstationary demand [6–9]. This is true for the
marks the profit per unit. To increase the profit and to situation of the automotive tire service store who orders
maintain, perhaps to gain, competitive advantage, the tires from manufacturers or wholesalers in advance,
retailer must try to reduce the cost of its operations stores them at a warehouse until they are sold to
related to holding of such products while maintaining customers. It is clear that the exact customer demands are
customer service [1–2]. Thus, adopting an appropriate not foreknown and normally are not constant over time
inventory control policy to match the variation of [10].
customer demands, then, has become vital and challenged Most continuous review of inventory control policies,
for the retailer [2–4]. This decision is, commonly, a part decisions on the order quantity (Q) and the reorder point
of inventory management strategy. (r) are needed for each product to meet customer
104 วิศวสารลาดกระบัง ปี ที่ 39 ฉบับที่ 2 มิถุนายน 2565

requirements with reasonable cost. The customer service To discuss mathematical models related to inventory
level (SL) is usually set as the target to achieve with control concepts throughout this paper, the following
minimum (or at reasonable) total cost (TC) [2],[8–12]. notations are defined.
Thus, SL and TC are generally used as the main indexes t Subperiod in a planning time
to evaluate the performance of an inventory policy and T Planning horizon
are significantly affected by the parameters Q and r. The Co Ordering cost (baht/unit)
appropriate values of Q and r closely relate to the Cc Carrying cost (baht/unit/period)
behavior of demands and lead times. Therefore, it is Cs Shortage cost (baht/unit)
important to statistically understand the demand and lead CEX Expediting cost in (baht/time)
time behaviors which can be pointed out when sufficient Dt Product demand for t
historical data is available. Much of the research about 𝑑̅ Average demand
inventory control policies have assumed that demands are Lt Ordering lead time t
normally distributed with a known mean (µ) and standard 𝐿̅ Average ordering lead time
deviation (σ) [8 ] ,[1 3 – 1 4 ] . Some recently applied Q Order quantity
researchers have extended their analysis to Poisson [9] Q* Optimal order quantity
and Exponential of demand distribution [1 5 ] with r Reorder point
constant lead time. However, some historical demand At Available inventory for t
patterns do not statistically match any of the traditional SS Safety Stock
distributions while the ordering lead times are not always SL Customer service level
constant [1 0 ] . Therefore, the empirical distribution SLtg Target customer service level
should be assumed in the evaluation of an inventory TC Total cost
control policy. SDt Satisfied product demand for t
This paper considers the situation of nonstationary Zt 1 if an order is placed at the end of t or 0 otherwise
but known to be empirically discretely distributed of Iavg,t Average inventory for t
product demands and lead times, focusing on determining Iint,t Inventory at the beginning of t
the suitable values of Q and r for the (Q, r) model. The Iend,t Inventory at the end of t
TC and SL of the model are determined using the Excel- Is,t Shortage units for t
based Monte Carlo Simulation (MCS) approach with a 𝐼𝑡∗ Dummy inventory at the end of t
set of actual historical demand data from an automotive CCt Carrying cost incurred for t
tire service. OCt Ordering cost incurred for t
SCt Shortage cost incurred for t
2. Backgrounds f(Dt,i) Probability mass function of Dt
This section is separated into 2 subtopics. The main Ft(At) Cumulative distribution function of At
performance indexes of inventory control policy are E(Dt) Expected value of Dt
discussed in Section 2.1. A brief literature review related ORt Order received at the beginning of t
to the problem scenario is discussed in Section 2.2. AVt Order arrival time for order placed at the end of t
Ladkrabang Engineering Journal, Vol. 39 No.2 June 2022 105

The dummy inventory at the end of period t, ( 𝐼𝑡∗ ), practice, shortage cost may occur when the condition in
is determined using Iend,t +Q.Zt to prevent reordering if Equation (2) is violated for some t’s which would pull the
the preceding order has not arrived. overall SL down and damage the image of a company (or a
retailer). Numerically, the shortage cost is added to the
2.1 Performance Indexes total inventory management cost to better enumerate the
The main reason to employ an inventory strategy is to situation and it is given in Equation (3).
meet customer demand, especially in a retail operation [1–
2],[16]. However, the demand is usually not known for 𝑇𝐶 = ∑𝑇𝑡=1(𝑍𝑡 𝐶𝑜 + 𝐼𝑎𝑣𝑔,𝑡 . 𝐶𝑐 + 𝐶𝑠 ) (3)
certainty. Thus, it is not always possible to hold exactly the
amount of product demanded. In practice, with selected Iavg,t can be calculated from (Iint,t + Iend,t) / 2. The value
inventory control policy the additional amount of of Cs may vary depending on how the shortage situation is
inventory, called safety stocks (SS), is kept on hand to handled. Assuming that there is no other constraints, the
increase the ability to meet the product demand in a timely, inventory control policy should be selected and
efficient manner. The higher expected SL requires a higher implemented in order to maximize the SL with the
inventory level and, thus, higher inventory costs. minimum TC or to minimize the TC with acceptable target
The actual SL of any product, as shown in Equation (1), SL, regardless of demand distribution.
is the ratio between the total satisfying demand and the
total demand for the entire controlling period (denotes by 2.2 Literature Review
T) [2],[9]. Some books [1–2],[16], and many research papers
described the importance and details of inventory
𝑆𝐿 =
∑𝑇
𝑡=1 𝑆𝐷 𝑡
(1) management. The recent versions tended to emphasize more
∑𝑇
𝑡=1 𝐷𝑡
on the supply chain [1],[3],[10]. Bedworth and Bailey [16]
The target (or planned) customer service level (SLtg) explained the principle of EOQ and how it can be
may be set at the beginning of the controlling horizon. The implemented with a variety of inventory control policies
demand for the product for subperiod t is a random under deterministic assumptions. They also have suggested
variable. SLtg is simply the probability of demand for how to use safety stock in dealing with stochastic conditions
subperiod t (Dt) being less than or equal to the available with an underlying normal probability of demand and lead
inventory for the same period (At). For period t, the time. Applications can be found in [17–19]. Similar
probability is given by the cumulative distribution function background can be found in Russell and Taylor III [1] with
Ft(At), which can be expressed in Equation (2). more emphasis on supply chain management. Lila [2]
provided detail regarding how to set up Excel spreadsheet for
𝑆𝐿𝑡𝑔,𝑡 = 𝐹𝑡 (𝐴𝑡 ) = 𝑃[𝐷𝑡 ≤ 𝐴𝑡 ] (2) inventory control using the Monte Carlo Simulation
approach. Implementation and adaptation of such technique
Where P represents the probability of the term in the can be seen in numbers of application researches, [8–
bracket. The expected service level for the entire planning 10],[12],[14],[20], for example. All of them tried to
horizon can be expressed as ∏𝑇𝑡=1 𝐹𝑇 (𝐴𝑡 ). Obviously, at determine the appropriate inventory control policies with
is a function of the order quantity Q and reorder point r. In uncertain demands. Silsat and Lila [8] applied to the case of
106 วิศวสารลาดกระบัง ปี ที่ 39 ฉบับที่ 2 มิถุนายน 2565

consumable items that their demands did not dependent on can help to minimize the costs while meeting acceptable SL.
finished products and are highly fluctuated. Mahitpan, Lila, The basic parameters needed are order size, reorder point and
and Kunadilok [9] and Pengsawat [12] applied to spare parts safety stock.
for maintenance of critical equipment. Smmutranukul and With known and constant demand, no shortage and
Phanvijitsiri [10] compared performance between the (Q, r) instantaneous assumptions, the optimal order quantity is
and (T, S, s) models for a tire service store while Leepaitoon, determined using Equation (4).
and Bunterngchit [20] studied similar cases but demands
were generated based on the maximum, minimum and 𝑄∗ = √
2.∑ 𝐷𝑡 .𝐶𝑜
𝐶𝑐
(4)
average of historical data. Ruanghiranwanich and Lila [14]
used the MCS to find a suitable order quantity of parts for The quantity Q* is enough to fulfill the constant demand
electronic products. Some other researches about inventory for 𝑄∗ /𝑑̅ Periods and minimizes the total cost. In this case TC
management have utilized high-level simulation programs to in Equation (3) can simply be written as shown in Equation (5).
study the performance of inventory policies with uncertain
demands [6–7],[11],[19]. Abuizam [6] studied the periodic 𝑇𝐶 =
𝐶𝑜 .∑ 𝐷𝑡
𝑄∗
+ 𝐶𝑐 . 𝐼𝑎𝑣𝑔 (5)
inventory model using Parasade@RISK simulation.
Cholodowicz and Orlowski [7] considered inventory policy The amount Q* is reordered when the inventory level is less
for perishable products that have Weibull demand than or equal to the reorder point r, that can be set using
distribution using System Dynamics. Limbuan and Lila [11] Equation (6).
used ARENA to simulate the (Q, r) and the (T, S, s) models
for uncertain but low demands for spare parts. These 𝑟 = 𝑑̅ . 𝐿 (6)
researches resulted in a saving of 7–25% of inventory
If the lead time is also constant, the time between each
management cost compared to the before-studied policies
order is placed, remains the same.
through numerical cases.
In reality, the deterministic condition about demand and
Based on the reviews, it is obvious that the simulation
lead time are usually not true. Thus, SS is enacted to deal with
technique is an appropriate tool for use in studying inventory
the stochastics conditions. This requires an analysis of
control policies with stochastic conditions. To overcome the
statistical behaviors for both demands and lead times through
burden of buying specific simulation programs (normally
historical data. In the case that demand and lead time are
with high cost) and learning how to use them, Excel is a
normally distributed, SS could be calculated using several
convenient choice since it is most certainly available with any
equations explained in [1],[2],[5],[16]. However, the focus of
computer.
this paper is on the case of an empirical discrete demand. In any
period, Dt can take any value in position ith in a discrete random
3. Inventory Control Concepts
variable having ft(Dt,i) as its probability mass function (PMF)
Inventory refers to items or products that are kept to satisfy
as depicted in Table 1. Dtm,i represents the mid-point of each
the future demands that are random variables [2]. Holding
class i. For subperiod t, Dt can be classified into classes, i starts
insufficient inventory may lower the SL while having too large
from 1 and increases upward. Each class i has probability of
inventory incurs costs. An appropriate inventory control policy
occurrence of f(Dt,i). The actual demand value (Dt,i) is assumed
Ladkrabang Engineering Journal, Vol. 39 No.2 June 2022 107

to be discretely uniformly distributed between the lower (LDt,i) In this paper, performance indexes (SL and TC) of
and upper (UDt,i) bounds of each class i. the (Q, r) model was investigated if it was implemented
as the control policy at a retailer for an automotive tire
Table 1 Probability mass function of Dt service where demand and lead time are empirically
Classi Dtm,i LDt,i UDt,i f(Dt,i) discrete random variables. The concept of the (Q,r)
1 Dtm1 LDt,1 UDt,1 f(Dt,1) model is the quantity Q is ordered and is reordered once
2 Dtm2 LDt,2 UDt,2 f(Dt,2) the inventory depletes to the level of r or lower. The
… … … … … process repeats for the entire planning horizon.
u Dtm3 LDt,u UDt,u f(Dt,u) The investigation was performed through the Monte
u+1 Dtm,u+1 LDt,u+1 UDt,u+1 f(Dt,u+1) Carlo Simulation on Microsoft Excel. The development
of such simulation model is explained in Section 4.
… … … …
4. Excel Based Monte Carlo Simulation
Thus, in any subperiod, holding inventory greater than Monte Carlo (MC) is a technique to generate random
or equal to E(Dtm), statistically, corresponding to ≥0.5 or variables of interest from a known behavior [2].
50% SL. The variable At in Equation (2) also represents the Subsequently, such variables can be used in the logic of
reorder point r to reach the SLtg. Based on this relationship, a simulation model or a system to evaluate its
the r is then derived and can be expressed in Equation (7). performance. This type of model is often referred to as
the Monte Carlo Simulation (MCS) [2]. This paper built
𝑟 = 𝐸(𝐷𝑡 ) + ∑𝑢𝑖=1 𝑓𝑡 (𝐷𝑡𝑖 ) (7) the MCS on Microsoft Excel Spreadsheet to simulate the
performance of the (Q,r) model. The MCS for T periods
The term ∑𝑢𝑖=1 𝑓𝑡 (𝐷𝑡,𝑖 ) is the cumulative probability was constructed for individual tire model having TC and
function of Dt with values ranging from class i = 1 to class SL as performance indexes, the logical concepts and their
i = u that enable r satisfies the SLtg. Thus, it also represents corresponding Excel formula are explained in section 4.3.
the SS.
The lead time is often easier to manage as it does not 4.1 Input Parameters
vary in a wide range [9],[10]. The probability mass In Fig. 1, the main inputs for the MCS model consist
function Lt can be defined using discrete probability as of the following.
shown in Table 2. 1) All cost parameters including Co, Cc, and Cs,
beginning inventory, CEX and SLtg. These
Table 2 Probability mass function of Lt parameters were placed on the top left, above the Logic
Lt f(Lt) model.
1 f(L1) 2) The starting values of Q and r which were
2 f(L2) determined using Equations (4) and (7), respectively.
… … 3) Distributions of demand and lead time in the form
L f(LL) of empirical discrete probability tables.
108 วิศวสารลาดกระบัง ปี ที่ 39 ฉบับที่ 2 มิถุนายน 2565

4.2 Generation of Random Variables a. CCt = Cc. Iavg,t

The demand (Dt) and the lead time (Lt) are the main b. OCt = Co. Zt
random variables. With the discrete form for Dt as shown in c. SCt = CEX + Cs.(Dt - At),
Table 1, the value was generated using a nested VLOOKUP() if Dt - At > 0 or 0 otherwise.
function that performs the following 2 steps; The concept used in calculating shortage cost was based on
1) Select the class i based on the given discrete actual practice that the owner would always seek available
probability function. products from a nearby store when he/she did not have what the
2) The value of Dt, was generated based on a discrete customer wanted. Therefore, the expediting cost (CEX) would
uniform distribution between the LDt,i and the UDt,i occur along with loss of some profit which refers to as Cs.
. The generated values must be rounded to the 4.4 Output Analysis
nearest whole number if the demand are integers. The TC and SL could be determined based on the
Similarly, but easier, the Lt was generated using a relationship in Equations (3) and (1), respectively. In the MCS,
VLOOKUP() function based on its probability mass function TC is simply the summation of CCt, OCt and SCt for all periods
when the order was placed. t = 1 to T. Similarly, the SL is the ratio between the summations
4.3 Logic of the Model of SDt and Dt for all periods t = 1 to T.
Logics of the MCS can be described as follow; The MCS was verified by comparing its TC and SL with
1) Generated the pseudorandom numbers (rnt) for all manual calculation when the known and constant demands and
t’s using a rand() function. lead times were entered. The result indicated that the MCS
2) Generate Dt using the nested VLOOKUP() function model worked properly and can be used for further analysis.
from its PMF, for all t’s. Equations (5) and (7) were utilized to determine the values
3) For t=1, of Q and r, respectively. The fact that the aspect of stochastic
a. Set Iint,1 = beginning inventory (Iint,0), and Iint,t lead time was not considered in Equations (7), the values of Q
= Iend,t-1 for the remaining periods. and r were only set as the starting values in the MCS. The TC
b. Set OR1 = 0, indicating a starting period. with values of Q and r varied from -50% to +100%, with 10%
4) Set At = Iint,t + ORt for all t’s step, from their starting values were run to find the minimum
5) Set SDt = Min(Dt, At) for all t’s TC with acceptable SL (SL ≥ SLt,g). For each combination of Q
6) Set Iend,t = At - SDt for all t’s and r values, the number of simulation runs was determined
7) Set Zt = 1, if Iend,t + Q.Zt ≤ r or 0 otherwise. according to the criteria that the error (half-width, HW) of 95%
8) If Zt = 1, generate Lt using VLOOKUP function confidence interval (CI) of the TC must be less than 10% of its
from its PMF, or 0 otherwise, for all t’s mean. After the Q and r values were set to use in the MCS
9) Set AVt = t + 1 + Lt, if Lt, ≠ 0 or 0 otherwise. model to investigate the performance of the (Q, r) control
10) For t > 1 to T, policy in the interested case.
a. Set ORt = countif(range from 1 to t, t)*Q
b. Set 𝐼𝑡∗ = 𝐼𝑡−1 ∗
- SDt + Q, if Zt-1 = 1 or 𝐼𝑡∗ = 5. Numerical Case
∗
𝐼𝑡−1 SDt otherwise. In this paper, the real historical weekly units sold of the
11) Costs calculation, highest demand tire model at an automotive tires retail
Ladkrabang Engineering Journal, Vol. 39 No.2 June 2022 109

service located in Chonburi province was collected Table 5 Relevant information (cont.)
between 2019 to 2020, along with its order lead times. Item value unit
Statistically, based on the sum of the square of error, at CEX 1000 Baht/time
0.05 significant level, the demand and lead time behaviors Annual Demand 3779 units
did not match with any traditional probability distribution. Beginning Inventory 322 units
Therefore, empirically discrete distributions seem to be the SLtg 0.85
best functions and are in Tables 3 and 4 for demand and
lead time, respectively. According to reasonable management of the store
From Table 3, the average (𝐸(𝐷𝑡 )) is 82.85 units and manager, TC for the 2019–2020 years were evaluated to be
standard deviation (SD) of Dt is 47.85 units. Similarly, 93,877 baht/year with 1.00 of SL. These numbers would be
from Table 4, the average ( 𝐸(𝐿𝑡 ) ) is 1.1 weeks and used as a baseline for comparison with the indexes getting
standard deviation (SD) of Lt is 0.3 weeks. Other relevant from the MCS.
information of this tire model was collected and is The starting values of Q and r were calculated to be
provided in Table 5. 297.31 units and 173.85 units, respectively (E(Dt) = 82.85,
from class i = 1 to u = 3, SS = 91 units). The SLtg was set
Table 3 Distribution of Dt (units/week) to 0.85 (with r = 173.85 units, according to Table 3, this
Classi Dtm,i LDt,i UDt,i f(Dt,i) value falls into class 6, thus the SLtg of 0.87 could be
1 15.150 0.000 30.300 0.08 expected). The values of Q and r were rounded to 300 units
2 45.451 30.301 60.601 0.27 and 170 units, respectively, and were used to run the
3 75.752 60.602 90.902 0.37 experiments by varying Q from 150 to 600 and varying r
4 106.053 90.903 121.203 0.13 from 85 to 340, with 10 runs for each combination which
5 136.354 121.204 151.504 0.02 was enough since the half-widths were between 3.78–
6 166.655 151.505 181.805 0.06 7.65% for TCs, and were between 1.55– 2.29% for SLs, for
7 196.956 181.806 212.106 0.08 all combinations. The outcomes showed that the minimum
TC occurred at the combination of Q = 330 and r = 204 and
Table 4 Distribution of Lt (weeks) 221, with SL at least 0.945. To confirm the outcome, r =
Lt f(Lt) 210 was chosen at set to run the MCS to find the optimal
1 0.90 Q. Results are shown in Fig. 1.
2 0.10

Table 5 Relevant information

Item value unit
Unit cost 2850 Baht/unit
Cc 2.39 Baht/unit/week
Co 2000 Baht/order
Cs 82 Baht/unit Figure 1 TC from r = 210 units
110 วิศวสารลาดกระบัง ปี ที่ 39 ฉบับที่ 2 มิถุนายน 2565

From Figure 1, the minimum TC could be found with result of the experiments discussed previously. The
Q between 300 to 400 units. The value of Q = 350 units statistical results are given in Table 8.
and r = 210 units were set to run the MCS again for 30
runs. Statistical results of TC and SL are given in Table 6. Table 8 TC and SL from the MCS between normal and
Half-widths of both TC and SL deviate from their discrete distribution
means of only 4.68% and 1.03%, respectively. Therefore, Normal Discrete
Statistics
results from 30 runs of the MCS were statistically accurate. TC SL TC SL
Mean 88640 0.946 77682 0.968
Table 6 TC and SL from the MCS SD 13504 0.027 7599 0.022
Statistics TC SL HW 5043 0.01 3638 0.01
Mean 77682 0.968
SD 7599 0.022 Half-widths for all values of TC and SL for both
HW 3638 0.01 distributions were within 6% of their respective means
indicating that 30 runs were reasonable accurate.
The analysis also indicated that if the (Q,r) model was The means of TC and SL getting from the use of normal
implemented, about 17.25% (93,877 - 77,682 = 16,195 and discrete Q,r parameters were compared using 2 unequal
baht/year) of the total inventory management cost could be variance t-tests. Results, as shown in Table 9, strongly
reduced with expected SL dropped from 1.00 to 0.968, or indicated that there were differences in both means at 0.05
around 3.2% decrement. significant level. The Q,r parameters from the discrete
Performances of the (Q,r) model were also compared distribution return, 10,958 baht/year or 12.36% lower TC and
based on normal and discrete distributions of demand and 0.022 or 2.31% higher SL than that of the normal distribution.
lead time. Input parameters that were used in 30 runs of
the MCS for the two underlying distributions are provided Table 9 Result of t-test of TC and SL from MCS between
in Table 7. normal and discrete distributions
TC SL
Items
Table 7 Input parameters for normal and discrete distribution Normal Discrete Normal Discrete
Mean 88640 77682 0.946 0.968
Parameters Normal Discrete
 (unit/week) 71.69 82.74 SD 13504 7599 0.027 0.022
 (unit/week) 46.27 47.77
Observations 30 30 30 30
Hypothesized Mean Diff. = 0 Mean Diff. =0
Q (unit) 300 350
p value 0.0003 0.0011
r (unit) 176 210

For normal distribution, the value Q from historical 6. Results and Conclusions
demand was 297.32 (rounded to 300) while the value r was This paper aimed to determine the suitable values of Q
calculated from 𝐷̅𝑡 . 𝐿̅𝑡 + 𝑧0.85 . 𝑆𝐷. √𝐿̅𝑡 . On the other and r in the (Q,r) inventory control policy model when it
hand, for discrete distribution, the values Q and r were the was applied to the situation of nonstationary but known to
be empirically discretely distributed of product demands
Ladkrabang Engineering Journal, Vol. 39 No.2 June 2022 111

and lead times. The product demand for each period (Dt) is Using Simulation-Based Optimization,” in Proc. the
defined in the form of an empirically discrete classes while 7th IFAC Conference on Manufacturing Modelling,
the discrete uniform distribution is assumed to represent Management, and Control International Federation of
the random distribution of demand within a class. The Automatic Control, Saint Petersburg, Russia, 2013, pp.
Excel-based Monte Carlo simulation (MCS) approach was 418–423, doi: 10.3182/20130619-3-RU-3018.00384.
described in detail and utilized to evaluate the performance [4] Y. Tao, L. H. Lee, E. P.Chew, G. Sun and V. Charles,
of the model in terms of TC and SL. Application to the “Inventory Control Policy for a Periodic Review
numerical case of a high-demand tire model of an System with Expediting,” Applied Mathematic
automotive tire service store demonstrated that the MCS Modelling, vol. 49, pp. 375–393, 2017, doi:
can be used to estimate the outcome of the (Q,r) model 10.1016/j.apm.2017.04.036.
efficiently. Based on the set of historical data used in this [5] N. C. Goncalves, M. S. Calvaho and P. Cortez,
case study, result from the MCS also showed that setting “Operations Research Models and Methods for Safety
of the inventory control parameters with the discrete Stock Determination: A Review,” Operations
distribution of demands and lead times of products led to Research Perspectives, vol. 7, pp. 1–14, 2020, doi:
significantly better performance in terms of TC and SL than 10.1016/j.orp.2020.100164.
that of the normal distribution. Thus, the method proposed [6] R. Abuizam, “Optimization of (s, S) Periodic
should be utilized in a situation that demand and lead time Inventory Model With Uncertain Demand and Lead
do not seem to be normally distributed or match with other Time Using Simulation,” International Journal of
traditional probability distributions. The MCS can also be Management and Information Systems, vol.15, no. 1,
modified to fit the concept of other similar control policies. pp. 67–78, 2011, doi: 10.19030/ijmis.v15i1.1597.
However, this paper did not perform the sensitivity [7] E. Cholodowicz and P. Orlowski, “Development of
analysis of the Co, Cc, CEX, and Cs parameters as they may New Hybrid Discrete-time Perishable Inventory
influence the performance of such control policies, since Model based on Weibull Distribution with Time-
they are generally fixed for a specific situation. varying Demand Using System Dynamics Approach,”
Computer and Industrial Engineering, vol. 154, pp.
References 1–13, 2021, doi: 10.1016/j.cie.2021.107151.
[1] S. R. Russell and B. W. Taylor III, “Role of inventory [8] S. Silsat and B. Lila, “A Study of Inventory
management,” in Operations Management, 10th ed., Management Policies for Consumable items with
Solaris, Singapore: John Wiley & Sons (Asia) Pte Non-Stationary Demand: A Case Study of Sanitary
Ltd., 2019, ch. 13, sec. 13.1, pp. 555–556. Ware Factory,” presented at the Conference of
[2] B. Lila, “Analysis and control of inventory,” in Industrial Engineering Network, Songkla, Thailand,
Production Planning and Control, Bangkok, May 5–7, 2021.
Thailand: Top Publishing Co. Ltd., 2010, ch. 5, sec. [9] K. Mahitpan, B. Lila and J. Kunadilok, “A Study of
5.6,, pp. 167–168. Inventory Management Policy of Spare Parts for
[3] M. Godichaud and L. Amodeo, “Comparing Machine Maintenance Operations,” in Proc. the 2017
Inventory Policies for Closed-Loop Supply Chain Technology Innovation Management and
112 วิศวสารลาดกระบัง ปี ที่ 39 ฉบับที่ 2 มิถุนายน 2565

Engineering Science International Conference, of LIFO with Expiration Date Products,” The Journal
Bangkok, Thailand, 2017, pp. 251–257. of Thai Operations Research, vol. 2, pp. 22–32, 2017.
[10] A. Smmutranukul and J. Phanvijitsiri, “Study of [16] D. D. Bedworth and J. E. Bailey, “Inventory analysis
Inventory Control Policies; Case Study of a Selling and control,” in Integrated Production Control
and Changing Tires Store,” B.Eng. project, Dept. Ind. Systems, 2nd ed., New York, NY, USA: John Wiley
Eng., Burapha Univ., Chonburi, Thailand, 2019. & Sons, 1987, ch. 6, sec. 6.3-6.4pp. 203–210.
[11] W. Limbuan and B. Lila, “A Study of Inventory [17] C. Kasemset and W. Chatchayangkul, “Inventory
Control Policy for Spare Parts,” presented at the Management Model: A Case Study of Chemical
Conference of Industrial Engineering Network, Ubon Fertilizer Store”, presented at the Conference of
Rachathani, Thailand, Jul. 23–26, 2018, pp. 1–5. National Operations Research, Bangkok, Thailand,
[12] W. Pengsawat, “A Study of Management Policies for Sep. 6–7, 2012.
a Composite Material Used in the Maintenance [18] S. Santi and H. B. Mursyid, “Comparison Continuous
Operations of the Aviation Engine”, M.Eng. Thesis, and Periodic Review Policy Inventory Management
Dept. Ind. Eng., Burapha Univ., Chonburi, Thailand, System Formula and Enteral Food Supply in Public
2018. Hospital Bandung,” International Journal of Innovation,
[13] D. Williams and T. Tokar, “A Review of Inventory Management and Technology, vol. 4 no. 2, pp. 253–258,
Management research in Major Logistics Journals 2013, doi: 10.7763/IJIMT.2013.V4.401.
Themes and Future Directions,” The International [19] T. Aouam, F. Ghadimi and M. Vanhoucke, “Finite
Journal of Logistics and Management, vol. 19, no. 2, pp. Inventory Budgets in Production Capacity and Safety
212–232, 2008, doi: 10.1108/09574090810895960. Stock Placement under the Guaranteed Service
[14] T. Ruanghiranwanich and B. Lila, “A Study of Plastic Approach,” Computer and Operations Research, vol.
Component Inventory Models for Electronic 131, pp. 1–18, 2021, doi: 10.1016/j.cor.2021.105266.
Products,” presented at the National Conference on [20] S. Leepaitoon and S Bunterngchit, “The Application
Administration and Management, Songkla, Thailand, of Monte Carlo Simulation for Inventory
Jun. 30, 2018, pp. 530–540. Management: a Case Study of a Retail Store,”
[15] W. Supitak and S. Pulivekin, “An Inventory International Journal of the Computer, the Internet
Management using Simulation Concepts for the Case and Management, vol. 27, no. 2, pp. 67–83, 2019.