VIETNAM NATIONAL UNIVERSITY – HOCHIMINH CITY

INTERNATIONAL UNIVERSITY
DEPARTMENT OF INDUSTRIAL ENGINEERING &
MANAGEMENT

DETERMINE OPTIMAL INVENTORY LEVELS

FOR HM COMPANY IN THE CASE OF
MULTI-PRODUCT WITH STOCHASTIC
DEMANDS, BACKLOGGED SHORTAGES, AND
CAPITAL CONSTRAINT

INVENTORY MANAGEMENT COURSE

Lecturer: Assoc.Prof. Nguyen Van Hop
Group Number: 16
Class: IEIE22IU01

Student ID Member name % Contribution

IELSIU22302 Nguyen Duy Thien 100%
IELSIU22287 Ngo Nhat Bich Tram 100%
IELSIU22174 Doan Hong Ngoc 100%
IELSIU22280 Tran Ngoc Thu 100%

December/2024
 ABSTRACT

In this Inventory Management course’s project, the requirement for a real case study of

an enterprise is met by the current inventory system of HM company who does not

have any specific inventory policies to manage stocks. This project adopted a model

which is multi-item with stochastic demands, shortages considered as backlogged to

clarify the initial inventory levels needed for each of four products at the beginning of

inventory cycle with the capital witnessed as the only constraint, and the objective is to

optimize the inventory costs and profits. Data sources include historical data of

demands, financial parameters of products provided by the company. The analyst is

also carried out to determine the statistical characteristics of each product. Demand

behavior over periods is simulated by Pareto Distribution, which allows to establish

and examine the model for more stability. The findings help the company to set

reasonable inventory levels with minimized inventory costs and maximized profit in

consideration of budget limitation. This model is more applicable for both online and

offline sales where the products are not withdrawn immediately at the time it is

purchased but after a period counted from the day when sales occur.

Keywords: Multi-item, Stochastic demands, Capital constraint, Pareto Distribution,

Inventory levels

B
 TABLE OF CONTENTS

ABSTRACT.......................................................................1.B
TABLE OF CONTENTS............................................................................................1.C
CHAPTER 1 INTRODUCTION..............................................................................1
1.1 System Description........................................................................................1
1.2 Problem Statement........................................................................................2
1.3 Scope and Limitations...................................................................................2
CHAPTER 2 METHODOLOGY..............................................................................4
2.1 Mathematical model formulation.................................................................4
2.1.1 Assumptions................................................................................................4
2.1.2 Notations......................................................................................................4
2.1.3 Mathematical formulas................................................................................6
2.2 Algorithm.....................................................................................................14
CHAPTER 3 RESULTS.........................................................................................16
3.1 Data Collection and Processing..................................................................16
3.1.1 Data Collection..........................................................................................16
3.1.2 Data Processing.........................................................................................16
3.2 Results Discussion........................................................................................18
3.2.1 Optimal Inventory Levels..........................................................................19
3.2.2 Cost Breakdown........................................................................................19
3.3 Sensitivity Analysis......................................................................................21
3.3.1 Impact of Holding Costs............................................................................21
3.3.2 Impact of changing Budget (B).................................................................25
3.3.3 Impact of Purchasing Costs (A).................................................................25
3.3.4 Conclusion of sensitivity analysis.............................................................26
CHAPTER 4 CONCLUSIONS...............................................................................27
REFERENCES............................................................................................................28
APPENDIX A. COMPUTER CODE.........................................................................4.A
APPENDIX B. DEMAND DATA..............................................................................4.D

C
 CHAPTER 1 INTRODUCTION

1.1 System Description

The configuration of this model is for an inventory system handling multiple periods

and multiple products to optimize three features: sum of costs incurred as a total in

inventory, overall profits derived, and demand uncertainties with respect to budget

limits or backlogged shortages associated with it. The input data used in the present

implementation is demand planning estimates per product, cost parameters of buying,

holding, ordering, backlog, and financial limitation constraints. The decision factors

are aimed at determining the best beginning inventory level for the four selected

products over the specified inventory cycle. It uses the probabilistic (s, S) inventory

model along with Pareto Distribution to simulate demand fluctuation and enhance the

precision of prediction. The Python-based algorithms are used in computing the

appropriate level of stock using the iterative method; Lagrangian multipliers and the

bisection methodology ensure the most cost-effective solution within resource

restrictions. The goal is to figure out the initial inventory levels at the beginning of

each inventory cycle which reduces total inventory costs (buying, holding, and

ordering), while increasing service levels and maximizing expected profits. The

outputs include optimal inventory levels, cost breakdowns, and profitability analyses,

as well as sensitivity tests to account for demand swings and budget adjustments. This

system offers a strong, scalable solution for enterprises dealing with nonperishable

commodities, such as medical equipment, where cost effectiveness, demand

fluctuation, and service reliability are crucial to operational success.

1.2 Problem Statement
Company HM, a supplier of medical devices, faces challenges in managing its

inventory efficiently. Due to the short lead times for restocking products, the company

currently orders inventory based on intuition and immediate needs rather than a

structured, data-driven approach. This informal method results in inconsistent

inventory levels, leading to issues such as frequent stockouts and overstocking.

Stockouts prevent the company from fulfilling customer orders on time, potentially

decreasing customer service level. On the other hand, overstocking leads to high

holding costs, as extra inventory ties up the company's budget in storage and

maintenance rather than other growth or operational needs.

Additionally, Company HM operates with a limited budget for inventory, which makes

it vital to carefully balance inventory levels and costs. Without an appropriate

inventory policy, the company struggles to optimize its available budget while keeping

products in stock. This lack of a system for monitoring demand patterns, setting

reorder points, and planning for potential backorders increases the risk of costly

inefficiencies.

1.3 Scope and Limitations

This project addresses these issues by developing a systematic inventory management

approach using a periodic review (s, S) model tailored to the company’s operational

requirements. This model will incorporate a multi-item probabilistic inventory

framework that considers the stochastic demand under Pareto Distribution, the

possibility of backordering and the limited budget of the company. By implementing

this model, we aim to achieve an optimal initial inventory level of each of four

2
products at the start of each inventory cycle which minimizing the total cost of

inventory management.

In spite of the effort of trying to tackle the problem thoroughly, some limitation still

exist:

 Limited scope of product analysis: Although Company HM manages a large

volume of Stock keeping units (SKUs), this case study only focuses on four

types of products. This limited selection may limit the model’s applicability to

other products within the company’s broader inventory.

 Assumption of constant lead times: The model assumes that lead times for all

products are constant. In reality, lead times can vary due to factors such as

supplier delays, transportation issues, or production disruptions. Variable lead

times can affect inventory levels and the effectiveness of the (s,S) model,

which leads to inaccuracies in reorder decisions.

 Single-echelon inventory systems: The model assumes a single-echelon

inventory system, focusing solely on Company HM’s inventory without

considering upstream suppliers or downstream distributors. In a more complex

supply chain, interactions with suppliers and customers could influence

inventory decisions, necessitating more comprehensive multi-echelon models

3
 CHAPTER 2 METHODOLOGY

2.1 Mathematical model formulation

2.1.1 Assumptions
The model deals with a multi-item inventory system in which the demand for products

is uncertain. This system follows these assumptions:

- The inventory system considers four different items.

- Items are jointly replenished every 45 days.

- The lead time is the same for four items.

- Shortages are allowed and fully backlogged.

- Demands for items are random variables that follow probability distributions.

- A probabilistic multi-item system is analyzed, in which the optimal inventory

level at the beginning of the inventory cycle must be determined.

- Three costs are considered: the expected holding cost, the expected shortage

cost and the expected replenishing cost.

- The aim is to maximize the expected profit per unit time.

2.1.2 Notations
N : The number of items considered in the model

T0 : The length of the inventory cycle. Items are jointly replenished every

T0 time units

Qi : The lot size of the ith item, with i = 1, 2,…, N. This is the demand

during the inventory cycle

Si : The inventory level of the item i at the beginning of each inventory

cycle, with i = 1, 2,…, N. The stock of each item i goes up to the level Si after

4
adding the replenishment size Qi in the inventory. These Si are variables that need

to be determined to maximize the expected profit per unit time.

si : The reorder point of the product i, with i = 1, 2,…, N. si = Si – Qi

hi : The holding cost per unit time of a unit in stock of the ith item, with i

= 1, 2,…, N

A : The ordering cost. This is a constant and independent of the order

quantity for each item

wi : The unit backlogging cost per unit time of the ith item, with i= 1, 2,…

pi : The unit selling price of item i, with i = 1, 2,…,N

ci : The unit purchasing price of item i, with i = 1, 2,…,N. ci < pi

B : Budget limitation

Xi : The total demand along the inventory cycle T0 of the ith item (i = 1, 2,

… N). It is a non-negative random variable with probability density function fi(xi)

defined on the interval [ai, bi), with and

. The cumulative distribution function of Xi is denoted by

Fi(xi)

μi : The average demand of the item i during the period T0

Di(t) : Demand accumulated up to time t for the ith item with 0 < t < T0 and

1, 2,…, N

AQi(Xi, Si) : The average quantity held in stock of the item i during the period T0

RC : The replenishing cost per unit time

ABi(Xi, Si) : The average backlogging of the item i during the period T0

5
 EQi(Si) : The expected average quantity held in stock of the item i during the

period T0

EBi(Si) : The expected average backlogging of the item i during the period T0

EHC(S1, S2, . . ., SN ): The expected total holding cost per unit time

EBC(S1, S2, . . ., SN): The expected total backlogging cost per unit time

ERC : The expected replenishing cost per unit time

EC(S1, S2, . . ., SN ): The expected total inventory cost per unit time

EP(S1, S2, . . ., SN ): The expected profit per unit time

2.1.3 Mathematical formulas

It is assumed that Di(t) is time-dependent and is given by

(1)

The ni represents the index of the demand pattern, with ni > 0. If the demand pattern

index of the ith item is larger than one, a greater part of the demand occurs at the

beginning of the period. If the demand pattern equals one, the demand rate is constant

throughout the period. When the demand pattern is smaller than one, a larger part of

the demand occurs toward the end of the inventory cycle.

a. Inventory model for multiple items with stochastic demands, power demand

patterns, fixed scheduling period, and fully backlogged shortages

The average quantity in inventory for item i is

(2)

6
The average shortage is

(3)

From (2), the expected average amount in the inventory EQi(Si) is

(4)

and from (3), the expected average shortage EBi(Si) is

(5)

For each item i, with i = 1, 2, …, N, the holding cost per unit time is hiEQi(Si) and the

backlogging cost per unit time is wi EBi(Si). Thus, the expected total holding cost is

(6)

and the expected total backlogging cost is

(7)

The expected replenishing cost is

(8)

The sum of holding cost, backlogging cost, and replenishing cost is the expected total

inventory cost per unit time of the system C(S1,…, SN)

7
 (9)

The expected profit EP(S1,…, SN) per unit time during each inventory cycle is

(10)

The goal is to maximize the profit per unit time, corresponding to minimizing the

expected inventory cost per unit time. Therefore, the objective is to determine the

optimal initial inventory levels Si (i = 1, 2,…, N) that minimize the cost function C(S1,

…, SN), as defined in equation (10), within the region

with i = 1, 2,…, N}.

Proposition 1.

For each item i, with i = 1, 2, …, N, the function Zi(Si) defined on the interval [0, ∞) is

a continuous function, which takes value in the interval (0, 1]. In addition, this function

decreases on [0, bi).

(11)

Theorem 1.

(i) The expected inventory cost C(S1,…, SN) proposed in (9) is a strictly convex

function on the set with i = 1, 2,…, N}.

(ii) The optimal inventory levels , with i = 1, 2,…, N that minimize the inventory

cost function C(S1,…, SN) must satisfy the constraint

8
 , i = 1, 2,…, N (12)

From Proposition 1, equation (12) is equivalent to

, i = 1, 2,…, N (13)

and from Theorem 1, equation (13) has a unique solution , for each i = 1, 2,…, N.

Thus, the optimal inventory levels are given by

, i = 1, 2,…, N (14)

From (6) and (14), the minimum expected holding cost per unit time is

(15)

From (7) and (14), and considering that the average demand for item i is μi the

minimum expected backlogging cost per unit time is

(16)

From (9) and (14), the minimum expected inventory cost per unit time is given by

9
 (17)

From (10) and (14), the maximum expected profit EP0 per unit time is given by

(18)

Thus, when the budget is unlimited, the inventory policy suggested by (14) and (17) is

the optimal way to handle the inventory problem.

b. Inventory model for multiple items with stochastic demands, power demand

patterns, backlogged shortages, and limited budget

In this section, we study the optimal policy for the probabilistic multi-item inventory

system with power demands and full backlogging, assuming that the inventory cycle is

fixed and the total budget is limited.

Let B be the total budget available for all N items. Considering that ci represents the

unit cost of item i, with i = 1, 2,…, N, and Si is the initial inventory level of item i,

then the total cost of the N items must be less than or equal to the available budget.

Thus, the below condition must be satisfied

(19)

The objective function C(S1,…, SN) for this system is given in (9), but now we consider

the constraint of limited budget in (19). Thus, the new inventory problem is

subject to

(20)

10
Optimal policy for the multi-item system with limited budget

If the inventory level , with i = 1, 2,…, N, determined by (14), satisfied the

constraint (19), then they will be the optimal inventory levels.

On the other hand, as the cost function C(S1,…, SN) is a convex function, if the level

do not satisfy the limited budget constraint, then the optimal inventory levels for this

system with limited budget must hold the equality in (19). Thus, for i = 1, 2,..., N, we

can now get the optimal inventory levels Si∗ using the Lagrangian multipliers

technique. The following expression provides the Lagrangian function L from problem

(20), where is the Lagrangian multiplier.

(21)

For an optimal solution, we use partial derivatives of and obtain

(22)

Given a value of , for each item i, let Si( ) be the inventory level that solves the

equation (22)

(23)

If λ = 0, the inventory level Si(0) obtained from (23) are equal to levels given in

(14), with i = 1, 2,…, N. The inventory levels Si(λ) given in (23) must meet the

condition

Let g(λ) be the function defined on [0, ∞), with I(λ) and 1≤i≤N}.

11
 (24)

Theorem 2.

The continuous and strictly decreasing function g(λ) is provided by (24). Moreover, if

g(0) > 0, then the function g(λ) has a unique positive root λ* and, from (23), the

optimal inventory level of item i is Si(λ*).

We assume in this section that g(0) > 0 since the inventory levels presented in (14),

with i = 1, 2,..., N, do not satisfy constraint (19). Therefore, from Theorem 2, the

equation g(λ) =0 always has a unique positive solution λ* on the interval

. Once the value of λ* has been determined, the optimal

inventory levels , for i = 1, 2,..., N can be calculated from (23).

Suppose that the optimal inventory level of item i is , the minimum

expected holding cost per unit time is

(25)

The minimum expected backlogging cost per unit time is

(26)

12
The minimum expected inventory cost per unit time is given by

(27)

The maximum expected profit EP* per unit time is

(28)

c. Inventory model for multiple items with Pareto Distribution, power demand

patterns, backlogged shortages, and limited budget

The probability density of the demand fi(xi) follows a Pareto distribution as follows

(29)

for all with , i = 1, 2,…, N, where is the scale parameter and is the

shape parameter.

The average demand of the item i during the period T0 are

(30)

For each item i = 1, 2,…, N, the function Zi(Si) given in (2) depends on the relative

values of Si, and . Thus, we have

(31)

The maximum expected profit per unit time is

13
 (32)

MATHEMATICAL MODEL

Objective function: Minimize expected inventory cost

(33)

subject to

(34)

(35)

2.2 Algorithm
This is the algorithm of finding the optimal inventory level for each of four products.

Firstly, we will find the optimal inventory level without constraint through inverse Z-

function:

0
hi
Si = Z-1 ( )
hi +w i

Then, we put the value of S0i into the g function with respect to S0i :

4
g( S0i ¿ = ∑ ci Si −B
0

i=1

At this stage, we examine the value of g function, which leads to two cases: if the

value of the examined function is less than or equal 0, we can conclude the Lagrangian

multiplier λ ¿=0 , optimal inventory level of each product is S¿i =S0i , and go to step 5 to

14
calculate related inventory costs and end the algorithm. On the other hand, we will use

wi
Lagrangian multiplier λ , where λ ≤ , with the function:
ci

4
g( λ ¿ = ∑ ci Si (λ)−B
i=1

The value of λ is calculated by taking the inverse g-function with bisection method

wi
from 0 to max{ }. After figuring out the λ ¿, we will consider whether it is less than
ci

wi
or equal to : if it is not, then we conclude the optimal inventory level of product i
ci
¿
Si =¿0; otherwise, we can find the optimal inventory level through inverse function of
¿
Z with λ by using bisection method:
¿
0 hi + λ ci
-1
S =Z (
i )
hi + w i

Finally, we can figure out the inventory cost after computing all optimal inventory

levels of four products.

Table 2. 1. Algorithm for finding optimal initial inventory level for beginning of each
inventory cycle
0 -1
hi
Step 1. Calculate Si = Z ( h +w )
i i

4
Calculate g( S ¿ = ∑ ci Si −B
0 0
i
i=1

¿
Step 2. If g( S0i ¿ ≤ 0 , then λ =0 and S¿i =S0i for all i = 1, 2, 3, 4. Go to step 5

Else, go to step 3.
4

Step 3. Calculate λ ¿=g−1 ( λ )from g( λ ¿ = ∑ ci Si (λ)−B

i=1

¿
wi ¿ h + λ ci
Step 4. If λ∗≤ , then Si = Z-1 ( i ) . Go to step 5
ci hi + w i

15
 ¿
Else, Si =0and go to step 5.

Step 5. Calculate total inventory cost, expected profit, remaining budget.

16
 CHAPTER 3 RESULTS

3.1 Data Collection and Processing

3.1.1 Data Collection
This case study focuses on four types of products, and we have collected the real data

from the enterprise’s inventory system as the table below:

Table 3. 1. Parameters of four products

CMDP.50 OA1 OA2 OA3
Selling price VND 901.325 VND 390.243 VND 390.243 VND 390.243
Purchasing price VND 889.000 VND 356.328 VND 356.328 VND 356.328
Inventory cycle
1,5 1,5 1,5 1,5
(month)
Order quantity/lot
3 30 40 30
size
Holding cost VND 22.225 VND 8908 VND 8908 VND 8908
Ordering cost VND 1.000.000
Budget VND 30.000.000
Total inventory cycles 8
Backlogging cost (B2) VND 22.225 VND 7575 VND 7575 VND 7575
Total demand along
24 584 872 320
the inventory cycles
Lead time (HN) 1 week 1 week 1 week 1 week
Lead time (HCM) 3 days 3 days 3 days 3 days

3.1.2 Data Processing

First, we calculate the power demand pattern indices. The Python code for this part

aims to estimate the demand pattern index (n i) for different products based on their

cumulative demand data over time. The ni value provides insights into the rate at which

demand grows over time. Hence, we obtain the result below:

Figure 3. 1. The result of power demand pattern indices

17
According to the result, we can see that product 1 and 3 exhibit sublinear growth,

suggesting potential market saturation. Next, product 2 shows promising growth,

warranting investment in strategies to accelerate demand. Finally, product 4

demonstrates stable linear growth, emphasizing the importance of consistent supply

and marketing efforts.

Then, our group used Pareto Distribution to estimate the scale (eta) and shape (alpha)

parameters from cumulative demand data for four products over eight periods. The

code first calculates the scale parameter (eta) as the minimum non-zero cumulative

demand value. Then, it estimates the shape parameter (alpha) using the method of

maximum likelihood. Hence, the results show that the scale and shape parameters vary

across the four products, indicating differences in their demand distributions. Product 1

has a relatively high scale parameter, suggesting a higher starting point for demand.

Product 2 has a lower scale parameter but a higher shape parameter, indicating a

steeper distribution. Products 3 and 4 have similar scale parameters but different shape

parameters, suggesting variations in the rate of demand growth.

Figure 3. 2. The result of Pareto distribution

For the next part, we use the data collection from the enterprise to fill in the parameters

for calculating the optimal inventory level:

18
 Figure 3. 3. Parameters data

Finally, after applying Python code, we can easily determine the best inventory level

for a product with uncertain demand using the newsvendor model. The code considers

important costs like ordering, holding, and backlogging. Our code starts by finding the

expected demand and then uses the newsvendor formula to calculate the optimal

inventory level. After that, we use it to compute the expected inventory level, holding

costs, and backlogging costs. Finally, it adds up the ordering, holding, and backlogging

costs to find the total cost. Therefore, the obtaining results show the best inventory

level for different situations, along with the expected inventory level, holding costs,

backlogging costs, and total costs. The next part will discuss and explain the result of

our project.

3.2 Results Discussion

By using Python, we obtain the results for our case:

3.2.1 Optimal Inventory Levels

Figure 3. 4. Optimal inventory level after coding

19
  According to the output, the result shows a higher allocation to items with

likely greater demand or profitability, for example, OA1 and OA2.

 These results represent the ideal quantities to maintain for each item to balance

costs and expected profits, and the optimal inventory levels are approximately

given as:

o CMDP.50: 1.98 – round up to 2

o OA1: 36.86 – round up to 37

o OA2: 23.79 – round up to 24

o OA3: 18.61 – round up to 19

3.2.2 Cost Breakdown

Here is the code’s result of our group:

Figure 3. 5. Coding result

The holding cost is 19,451,552 VND, which indicates the cost of storing the inventory,

which includes expenses like warehousing, depreciation, and insurance. A significant

part of total costs, holding costs can be optimized by maintaining efficient stock levels.

Then, we note that the holding cost is considerably high compared to the ordering cost,

indicating that maintaining inventory incurs significant expenses. Therefore, this could

stem from factors such as storage fees, insurance, or potential spoilage.

20
A backlogging cost of zero suggests no unmet demand or delays in fulfilling customer

orders. Hence, this is a positive outcome, as it indicates high service levels.

The ordering cost is calculated at 125,000 VND. This relatively low cost shows that

the number of replenishment orders is minimized, likely due to effective inventory

policies such as bulk ordering.

30,000,000 VND of purchasing cost is the primary expense, as purchasing inventory

typically forms the bulk of the total cost. It reflects the cumulative cost of acquiring the

optimal inventory levels. From the coding, we can see that the type of cost is

extremely high, indicating that the purchased products are expensive. This could be

due to high raw material costs, complex manufacturing processes, or high demand for

the products.

Total Inventory Cost is the combined total of holding, ordering, and purchasing costs.

It represents the overall expense of managing inventory and serves as a benchmark for

evaluating cost efficiency. According to the code’s result, the total inventory cost is

substantial, which is likely driven by the high purchasing and holding costs.

We can see that there is a nearly balanced budget, with a negative number, suggesting

our model has fully utilized available resources to achieve optimal outcomes without

significant overspending. Also, the expected profit represents the estimated profit after

considering all costs. It is a substantial value, indicating that the chosen inventory

levels are highly profitable and align with business objectives. Despite the high costs,

the expected profit is positive, showing sales revenue exceeds total expenses.

21
3.3 Sensitivity Analysis

3.3.1 Impact of Holding Costs

3.3.1.1 Change in h1

Figure 3.6. Sensitivity analysis result when changing h1

As h1 increases, the inventory level of CMDP.50 exhibits a significant inverse

relationship, with changes ranging from -16.16% for a 40% increase to +23.73% for a

40% decrease, highlighting its high sensitivity. In contrast, the OA 1, OA2, and OA3

inventory levels are less affected, showing minimal percentage changes of less than

2%. Expected profit also reveals low sensitivity, with changes remaining below 0.6%

and slightly increasing with higher h1. Meanwhile, inventory costs decrease moderately

as h1 rises, with percentage changes confined to within ±0.25%. Hence, this analysis

emphasizes that h1 substantially impacts CMDP.50 inventory levels while exerting

relatively minor effects on profit and costs. Moreover, this suggests that optimizing h 1

is essential for achieving a balance between inventory and cost efficiency.

22
3.3.1.2 Change in h2

Figure 3.7. Sensitivity analysis result when changing h2

For CMDP.50, the percentage change in inventory levels ranges from an increase of

+5.56% due to a +40% rise in h2 to a decrease of -7.58% resulting from a -40%

decline, indicating moderate sensitivity. The inventory levels of OA 1, OA2, and OA3

exhibit differing responses, with OA1 experiencing a significant decrease (for instance,

-10.42% with a +40% increase), while OA2 and OA3 show relatively smaller

fluctuations, highlighting their mixed sensitivity. Expected profit is inversely related to

h2, facing considerable declines of up to -12.35% with a +40% increase, demonstrating

high sensitivity. On the other hand, inventory costs decline steadily as h 2 rises, with

reductions of as much as -6.54% for a -40% decrease. In summary, h 2 has significant

impacts on profit and cost dynamics, indicating that prudent adjustments to h 2 are

crucial for maintaining a balance between profitability and inventory efficiency.

23
 3.3.1.3 Change in h3

Figure 3.8. Sensitivity analysis result when changing h3

We note that the inventory level of CMDP.50 illustrates a moderate sensitivity, with

percentage fluctuations ranging from +4.55% for a +40% increase in h 3 to -6.56% for a

-40% decrease. Then, OA1 demonstrates a proportional response, showing smaller

variations (e.g., +5.59% for a +40% increase and -7.65% for a -40% decrease),

indicating low to moderate sensitivity. In contrast, OA 2 exhibits significant sensitivity,

with changes reaching as high as -14.54% and +19.79%, reflecting a substantial impact

of h3. OA3 presents mixed responses, with percentage changes spanning from +6.23%

to -8.49%. Expected profit declines sharply with an increase in h 3, plunging by as much

as -11.13% at a +40% rise, while it improves by +12.00% with a -40% reduction,

highlighting its strong sensitivity to changes in h 3. Conversely, inventory costs behave

inversely, rising by +5.50% at a +40% increase and decreasing by -5.93% at a -40%

decline. According to these findings, we can see that h 3 has a significant influence on

OA2's inventory levels, expected profit, and inventory cost, necessitating careful

consideration to optimize overall system performance.

24
 3.3.1.4 Change in h4

Figure 3.9. Sensitivity analysis result when changing h4

We can see that the sensitivity analysis of h 4 (the holding cost of product 4) highlights

its influence on inventory levels, expected profit, and inventory costs. For CMDP.50,

the inventory level exhibits moderate sensitivity, with percentage changes ranging

from +3.03% at +40% to -4.55% at -40%. OA1 and OA2 display a similar pattern,

demonstrating moderate variations; for instance, OA1 fluctuates between +3.72% and -

5.26%, while OA2 changes from +4.83% to -6.73%, indicating their dependency on h 4.

OA3, however, shows significant variability, decreasing by -14.40% at +40% and

increasing by +20.15% at -40%, reflecting a high sensitivity to h 4. Expected profit

exhibits a strong negative response to increases in h 4, declining as much as -6.88% at

+40%, but improving notably (+7.76%) when h 4 is reduced by -40%. Conversely,

inventory costs rise moderately (+3.40% at +40%) and decline significantly (-3.88% at

-40%). These findings indicate that h 4 has a substantial impact on OA 3's inventory

levels and expected profit, necessitating careful adjustments to achieve cost-

effectiveness and maintain system stability.

25
3.3.2 Impact of changing Budget (B)

Figure 3.10. Sensitivity analysis result when changing B

As B increases by +40%, the inventory levels of CMDP.50 and OA 1 see significant

rises of +29.80% and +35.73%, respectively. OA2 and OA3 experience even more

noticeable changes, with increases of +48.38% and +40.35%, indicating a strong

positive dependence on budget increases. On the other hand, when B decreases by -

40%, inventory levels dive, with CMDP.50 dropping by -30.30%, OA 1 by -37.49%,

OA2 by -45.49%, and OA3 by -40.62%. Additionally, expected profit is highly

sensitive to changes in B, declining by -48.65% with a +40% increase and surging by

+48.50% when B decreases by -40%. Therefore, this highlights a strong negative

correlation between budget and profit margin. Furthermore, inventory costs rise

considerably (+24.06%) when B increases by +40%, while a reduction in B leads to a

notable decrease (-23.98%) in costs.

3.3.3 Impact of Purchasing Costs (A)

Figure 3.11. Sensitivity analysis result when changing A

26
The sensitivity analysis of changing purchasing cost A reveals minimal impact on

inventory levels and costs, showcasing a highly stable system. Across all variations

from +40% to -40%, the inventory levels of CMDP.50, OA 1, OA2, and OA3 remain

unchanged, indicating that purchasing costs have no direct influence on inventory

quantities. Expected profit shows a slight proportional relationship with A, where an

increase of +40% reduces profit by -0.2039%, while a decrease of -40% increases

profit by +0.2039%. Similarly, inventory cost is directly proportional to A, with a

+40% increase raising costs by +0.1009% and a -40% decrease reducing costs by -

0.1009%.

3.3.4 Conclusion of sensitivity analysis

Among the factors analyzed, holding costs (h3 and h4) and budget (B) show significant

influence on inventory levels, expected profit, and inventory costs. Adjustments to

holding costs can substantially affect both profitability and cost efficiency, with lower

holding costs leading to higher profits and reduced inventory costs. Similarly, changes

in budget significantly impact inventory levels and financial outcomes, where a larger

budget increases inventory levels but reduces expected profit due to higher costs. On

the other hand, purchasing costs (A) have a negligible effect on the system, with

inventory levels remaining unchanged and only having a minimal influence on profit

and costs. In summary, the analysis concludes that the system is most sensitive to

holding costs and budget, requiring careful optimization of these variables to balance

profitability and operational efficiency. In contrast, its robustness against purchasing

cost variations ensures stability in procurement processes.

27
 CHAPTER 4 CONCLUSIONS

The current study addresses the inventory problem of HM Company and provides a

structured methodology to optimize inventory management under random demand,

financial restrictions, and backlogged shortages. The applications of a multi-item

probabilistic inventory model led to identifying the most advantageous levels of stocks

for the four products, minimizing the total costs, and improving the profits. With the

adopted model, results show that the use of the available budget can be maximized, as

nearly all the available budget is used, and there is an expected profit realized.

According to the results, a correct balance of purchasing cost, holding cost, and

ordering cost would maintain service levels without relating to the financial resources.

In particular, the higher allocation to products with greater demand or profitability

underlines the importance of prioritizing inventory items strategically. Besides, there

are no costs of backlogging, which means that the model was effective in aligning the

inventory level with demand, hence the timely fulfillment of customer orders. In

general, though the project reveals a substantial benefit, this analysis is conducted on a

subset of few products, keeping the lead times constant, and a single-echelon inventory

system. The extension of this methodology to incorporate variable lead times, the

inclusion of other categories of products, and multi-echelon systems are scopes for

further work that might lead to higher applicability in the model with more complex

supply chains. The implementation of the proposed recommendations will enable HM

Company to ensure more consistent inventory performance, less inefficiency, and

long-term operational and financial objectives.

28
 REFERENCES

[1].Gunawan, C. T. S., Lesmono, D., & Salim, D. (2024). MULTI-ITEM

PROBABILISTIC INVENTORY MODELS CONSIDERING DISCOUNT,
EXPIRATION, WAREHOUSE CAPACITY, AND CAPITAL
CONSTRAINTS. BAREKENG: Jurnal Ilmu Matematika Dan Terapan, 18(3),
2001–2014. https://doi.org/10.30598/barekengvol18iss3pp2001-2014
[2].Maitra, S. (2024). Inventory Management Under Stochastic Demand: A
Simulation-Optimization Approach. ArXiv (Cornell University).
https://doi.org/10.48550/arxiv.2406.19425
[3].Sicilia, J., San-José, L. A., Alcaide-López-de-Pablo, D., & Abdul-Jalbar, B. (2022).
Optimal policy for multi-item systems with stochastic demands, backlogged
shortages and limited storage capacity. Applied Mathematical Modelling, 108,
236–257. https://doi.org/10.1016/j.apm.2022.03.025
[4].Zhang, G. (2010). The multi-product newsboy problem with supplier quantity
discounts and a budget constraint. European Journal of Operational
Research, 206(2), 350–360. https://doi.org/10.1016/j.ejor.2010.02.038

29
APPENDIX A. COMPUTER CODE
B
C
 APPENDIX B. DEMAND DATA

Time t CMDP.50 OA1 OA2 OA3

1
2
3 8 3
4
5
6
7 15 15 3
8
9
10
11
12
13
14 5 5 3
15
16
17
18
19 10 10
20 2
21
22
23
24
25
26
27
28
29 3 5
30
31
32
33 1
34 10 10
35 10
36
37
38 10 10 10
39

D
40 10 10
41
42
43
44 5
45
46
47 5 5
48 1
49
50
51
52
53
54
55
56 1 5
57 10
58
59
60
61
62
63 3
64 10
65
66 20 20
67 1
68
69
70 10
71
72
73
74
75
76
77 10
78 2
79
80
81
82

E
 83
84
85
86
87
88
89
90
91
92
93
94 5 10 10
95
96
97
98
99
100
101 2 14 5
102
103
104
105
106
107 10 10
108 10 10
109
110 3 5 3
111
112
113 3 5 5
114
115
116
117 5
118
119
120
121
122 10 20 5
123
124 3
125

F
126
127
128
129
130
131
132 1
133
134
135
136 1 10 15
137 5 10 5
138
139 14 25
140
141
142
143
144
145
146
147
148
149
150 6
151
152
153
154 10 5
155 20
156
157
158
159 20 5
160
161
162
163 5
164
165
166
167
168 2 8

G
169
170
171
172
173
174 3
175
176
177
178 1
179
180 10 10
181 5 5
182 4 5
183
184
185
186
187
188
189
190
191
192 5
193
194
195
196
197 4
198
199 3
200
201 15 10 8
202
203
204
205
206
207 10 10 15
208 6 7
209
210 5
211

H
212
213
214
215
216
217 3 5
218
219
220
221
222
223
224
225
226
227 10 20 10
228 1 5 8
229
230
231
232
233
234 5 4
235 1 3
236 3 3
237 10 20
238
239
240
241 3 3
242 10 10 5
243
244
245
246 10 10
247
248
249
250
251 3 8 3
252
253
254

I
255
256
257
258 10
259
260
261
262 10 4
263
264 5 5 3
265 3 2 1
266 15 20
267 2 1
268
269
270 2 4
271
272 10
273 2 3 2
274 5 5
275
276
277
278
279
280
281
282
283 2
284 1
285
286 13 8
287
288 10 15 10
289
290
291
292
293
294 20 10 10
295 5
296
297 13 12 2

J
298
299 3 3 4
300
301
302
303
304 3
305 5
306
307
308
309 5
310
311
312 2
313
314 1 3
315
316
317
318 2
319
320
321 4 4 2
322 5 5 3
323 3 3 3
324
325 1 20
326
327 1
328 5 6
329 5
330 6 6
331
332
333
334
335
336
337
338
339 3 3
340 5 2

K
341
342
343 10
344
345
346
347 5 5
348 3
349
350
351
352
353 5 5 3
354 6 7
355
356
357 2
358
359
360
361 10 20
362 3
363 2 8 8 6
364
365
366
367 2 1

L
M