Winning Strategy for the Littlefield Simulation Game

A System Dynamics Approach

A Major Qualifying Project /Interactive Qualifying Project Report
Submitted to the Faculty
Of the
WORCESTER POLYTECHNIC INSTITUTE
In partial fulfillment of the requirements for the
Degree of Bachelor of Science
In Economic Sciences
By
__________________________
Qiying Fan
__________________________
Mengjie Liu
__________________________
Murtaza Turab Thahirally
__________________________
Siqi Wang
Submitted 10/14/2011

Approved:

____________________

Professor Khalid Saeed, Advisor

_____________________

Professor Oleg Pavlov, Advisor

Abstract

Littlefield Simulation game is an important learning tool for understanding operations

principles in production environments, and therefore it is widely used by many leading business

schools. This project attempts to model this game using system dynamics approach, which

allows realistic representation of the production system of Littlefield Simulation. It also creates

an environment for comparison of the consequences of various policies.

Page | 1
TABLE OF CONTENTS

ABSTRACT ................................................................................................... 1
1.0 INTRODUCTION ..................................................................................... 5
2.0 BACKGROUND ....................................................................................... 7
2.1 LITTLEFIELD TECHNOLOGIES AS A TEACHING TOOL ............................................. 7
2.2 STUDENT RESEARCH AND EXPERIENCE ............................................................ 7
2.3 LIMITATIONS OF PREVIOUS RESEARCH AND STRATEGIES...................................... 8
2.4 SYSTEM DYNAMICS METHOD AND ITS BENEFITS ............................................... 8
2.5 OPERATIONS MANAGEMENT TECHNIQUES AS POLICIES ...................................... 9
3.0 PROBLEM EXAMINATION .................................................................... 11
3.1 GAME-PLAY EXPERIENCE AND PROBLEM DESCRIPTION ..................................... 11
3.2 DATA REVIEW ......................................................................................... 12
3.2.1 Station Queues .............................................................................. 12
3.2.2 Station Utilization ......................................................................... 13
3.2.3 Inventory ....................................................................................... 14
3.2.4 Completed Jobs ............................................................................. 15
3.2.5 Lead Time...................................................................................... 16
3.2.6 Revenue ........................................................................................ 17
3.3 DYNAMIC HYPOTHESIS .............................................................................. 18
4.0 SYSTEM DYNAMICS MODEL DESCRIPTION AND TESTING .................... 21
4.1 OVERVIEW OF THE SECTORS ....................................................................... 21
4.2 SECTOR DESCRIPTIONS .............................................................................. 22
4.2.1 Input to the system ....................................................................... 22
4.2.2 Cash Constraints ........................................................................... 25
4.2.3 Financials ..................................................................................... 29
4.2.4 Contracts ...................................................................................... 33
4.2.5 Average Lead time........................................................................ 36
4.2.6 Machines ...................................................................................... 37
4.2.7 Stations ......................................................................................... 40
4.3 MODEL TESTING ...................................................................................... 46
4.3.1 Extreme Value Testing .................................................................. 46
4.3.1.1 Incoming Customer Order .............................................................. 46
4.3.1.2 Desired Number of Machines......................................................... 49
Page | 2
 4.3.1.3 Reorder Level and Reorder Quantity .............................................. 50
4.3.2 Input Response Testing ................................................................. 53
4.3.2.1 Constant Incoming orders .............................................................. 53
4.3.2.2 Step Response ................................................................................ 54
4.3.2.3 Sine Wave Input ............................................................................. 55
4.3.2.4 Random Input Response ................................................................ 56
5.0 GAME-PLAYING STRATEGIES FOR MAXIMIZING PROFIT ...................... 58
5.1 ANALYSIS OF CURRENT STRATEGY ................................................................ 58
5.1.1 Initial Condition and Input to the System ...................................... 58
5.1.2 Contract Selection ......................................................................... 59
5.1.3 Ideal Number of Machines to Purchase......................................... 60
5.1.4 Re-order Level............................................................................... 61
5.1.5 Re-order Quantity ......................................................................... 62
5.1.6 Machine Priority............................................................................ 63
5.2 PREDICTIONS OF MODEL BEHAVIOR ............................................................. 63
5.3 PROFIT-MAXIMIZING STRATEGY IMPLEMENTATION AND TESTING ....................... 65
5.3.1 Initial Run with Reactive Strategy ................................................. 66
5.3.2 Switching to Contract 3 ................................................................. 67
5.3.3 Implementing optimal re-order level and re-order quantity .......... 70
5.3.3.1 Optimum Reorder Level ................................................................. 71
5.3.3.2 Optimum Reorder Quantity ........................................................... 72
5.3.3.3 Refining reorder level and reorder quantity for the last 50 days .... 73
5.3.4 The Number of Machines .............................................................. 75
5.3.5 Verifying Optimal Strategy with the Random Input ...................... 77
6.0 LIMITATIONS AND FUTURE WORK ...................................................... 78
6.1 LIMITATIONS ....................................................................................... 78
6.2 FUTURE WORK ........................................................................................ 78
7.0 CONCLUSION ....................................................................................... 80
8.0 WORK CITED ....................................................................................... 81
9.0 APPENDIX ........................................................................................... 83
9.1 EQUATIONS ............................................................................................ 83
9.2 USER INTERFACE ...................................................................................... 90
9.3 LITTLEFIELD GAME PARAMETERS ................................................................. 91

Page | 3
TABLE OF FIGURES
FIGURE 1-DAILY AVERAGE NUMBER OF KITS FOR STATIONS ......................................................................................................12
FIGURE 2-UTILIZATION FOR STATION1 ..................................................................................................................................13
FIGURE 3 INVENTORY LEVEL................................................................................................................................................14
FIGURE 4 NUMBER OF JOBS COMPLETED FOR EACH CONTRACT.................................................................................................15
FIGURE 5 NUMBER OF JOBS COMPLETED ...............................................................................................................................15
FIGURE 6 DAILY AVERAGE JOB LEAD TIME BY CONTRACT TYPE ..................................................................................................16
FIGURE 7 DAILY AVERAGE REVENUE BY CONTRACT TYPE ..........................................................................................................17
FIGURE 8 DAILY AVERAGE REVENUE .....................................................................................................................................17
FIGURE 9 DYNAMIC HYPOTHESIS MODEL ..............................................................................................................................19
FIGURE 10 OVERALL SECTOR DIAGRAM ................................................................................................................................21
FIGURE 11 INPUT OF THE SYSTEM ........................................................................................................................................23
FIGURE 12 EFFECT OF RAW MATERIAL RATIO ON JOB RELEASE .................................................................................................25
FIGURE 13 CASH CONSTRAINTS ...........................................................................................................................................26
FIGURE 14 CASH CONSTRAINT ON RAW MATERIALS................................................................................................................28
FIGURE 15 CASH CONSTRAINTS ON MACHINES ......................................................................................................................29
FIGURE 16 FINANCIAL SECTOR ............................................................................................................................................30
FIGURE 17 CONTRACTS SECTOR ..........................................................................................................................................33
FIGURE 18 -CONTRACT TYPE 1 AND LEAD TIME......................................................................................................................35
FIGURE 19- CONTRACT TYPE 2 AND LEAD TIME......................................................................................................................35
FIGURE 20 -CONTRACT TYPE 3 AND LEAD TIME......................................................................................................................36
FIGURE 21 LEAD TIME SECTOR ............................................................................................................................................36
FIGURE 22 MACHINES SECTOR............................................................................................................................................38
FIGURE 23 STATIONS SECTOR .............................................................................................................................................41
FIGURE 24 CASH LEVEL FOR EXTREME VALUES OF CUSTOMER ORDERS .......................................................................................48
FIGURE 25 -LEAD TIME FOR EXTREME VALUES OF CUSTOMER ORDERS .......................................................................................48
FIGURE 26 CASH LEVEL FOR EXTREME VALUES OF DESIRED MACHINES .......................................................................................50
FIGURE 27 LEAD TIME FOR EXTREME VALUES OF DESIRED MACHINES ........................................................................................50
FIGURE 28 CASH LEVEL FOR ZERO REORDER LEVEL AND REORDER QUANTITY ...............................................................................52
FIGURE 29 LEAD TIME FOR ZERO REORDER LEVEL AND REORDER QUANTITY ................................................................................52
FIGURE 30 CASH LEVEL WITH CONSTANT INCOMING CUSTOMER ORDERS ...................................................................................54
FIGURE 31 CASH LEVEL STEP RESPONSE................................................................................................................................55
FIGURE 32 CASH LEVEL SINE WAVE RESPONSE ......................................................................................................................56
FIGURE 33 CASH RESPONSE TO A RANDOM INPUT ..................................................................................................................57
FIGURE 34 INCOMING CUSTOMER ORDERS PER DAY ...............................................................................................................58
FIGURE 35-MEAN, STANDARD DEVIATION AND RANGE OF INCOMING CUSTOMER ORDER PER DAY .................................................59
FIGURE 36- ORDER DISTRIBUTION .......................................................................................................................................59
FIGURE 37 CASH LEVEL WITH A REACTIVE STRATEGY ................................................................................................................66
FIGURE 38 CONTRACT TYPE CHOSEN WHEN POLICY IS NOT USED .............................................................................................67
FIGURE 39 POLICY SECTOR .................................................................................................................................................68
FIGURE 40 CONTRACT TYPE 1 REVENUE GRAPH .....................................................................................................................68
FIGURE 41 CONTRACT TYPE 3 REVENUE GRAPH .....................................................................................................................69
FIGURE 42 CASH LEVEL WHEN CONTRACT 3 IS CHOSEN ...........................................................................................................70
FIGURE 43- IMPLEMENTING ECONOMIC REORDER QUANTITY AND ECONOMIC REORDER LEVEL .......................................................71
FIGURE 44- CHANGING THE REORDER LEVEL..........................................................................................................................72
FIGURE 45 CHOOSING THE OPTIMUM REORDER QUANTITY........................................................................................................73
FIGURE 46 REFINING REORDER LEVEL AND REORDER QUANTITY FOR THE LAST 50 DAYS...................................................................74
FIGURE 47- AVERAGE LEAD TIME BEFORE CHANGING DESIRED NUMBER OF MACHINES ...................................................................75
FIGURE 48 LEAD TIME BEFORE AND AFTER CHANGING THE NUMBER OF MACHINES ........................................................................76
FIGURE 49 CASH LEVEL BEFORE AND AFTER CHANGING THE DESIRED NUMBER OF MACHINES ...........................................................76
FIGURE 50 CASH LEVEL WHEN CUSTOMER ORDERS ARE RANDOM ...............................................................................................77

Page | 4
1.0 Introduction

Littlefield Simulation is a game widely used in management courses that replicates a

manufacturer’s decision making mechanism. In the game, teams are challenged to optimize the

system and maximize cash flow for Littlefield Technologies, a factory that assembles Digital

Satellite System Receivers from electronic kits. The electronic kits are acquired from the

supplier. The assembly process consists of four steps carried out at three stations. The last step is

inspection. The configurations that participants could adjust include the contract type, inventory

reorder point, reorder quantity, number of machines on each of the three stations, and the

machine strategies, which is the priority of the workflow. The system receives a random number

of customer orders each day, which ranges between 3 and 24. The revenue is affected by the

average lead time to process the customer order, contract type the player chooses, and the

inventory reordering settings.

A System Dynamics method is used to analyze the game, where the data retrieved from

the game is used to create a reference mode. The next step is to create a dynamic hypothesis,

which is a set of casual feedback loops, to represent the production system and the players’

decision mechanisms. A System Dynamics model is built to incorporate the dynamics hypothesis

and to replicate the reference mode. The last step is to design policies to achieve optimized

sustainable profit growth through analyzing trends in the game and patterns generated by the

model.

The paper elaborates on the System Dynamics model and the steps we went through to

design the winning policy. Considering the scale and level of complexity of our model, certain

statistical tools were utilized to generate optimized input parameters, which helps us efficiently

Page | 5
test and experiment with the model, and also to produce more realistic patterns of various

policies.

The paper first analyzes the data retrieved from the game and discusses the chosen

methodology with justifications. Afterwards, the model is described in detail, along with the tests

performed to verify the validity. Thereafter the policy is described and tested again. The final

section of the paper outlines future work that can be done and the limitations of Littlefield

Simulation and the model we constructed.

Page | 6
2.0 Background

2.1 Littlefield Technologies as a Teaching Tool

In 1996, Stanford University adopted Littlefield Technologies in its MBA courses.

"Littlefield Technologies has it all --- the good, the bad, and the lessons most managers can only

learn the hard way," according to the article "Factory Simulation Program Wins Wide

Recognition" on the Stanford Graduate School of Business' website. Compared to the traditional

teaching methods, Littlefield Technologies is more affordable, accessible and effective.

“…the simulation games are strongly linked to class concepts- the students are forced to

think logically about the problems that they are facing and they learn from iterative

experimentation. “ said Julia Miyaoka, professor at San Francisco State University, who used

Littlefield Technologies in both her undergraduate and MBA classes.

Littlefield Technologies draws knowledge on topics such as capacity, utilization, queuing,

scheduling and inventory. It is currently adopted by institutions globally and students often

compete as groups to run the factory in the most efficient and profitable manner.

2.2 Student Research and Experience

Many students enjoyed the experience of playing Littlefield Technologies. Winner of the

2007 MIT Sloan Operations Simulation Competition, the “yalewillwin” team built a demand

model to forecast the demand trend and a queuing model to experiment with different settings,

such as increasing the number of machines, changing some of the lot sizes. Other winning

strategies mentioned by the team members include re-evaluating final inventory level to avoid

waste or shortage of inventory, regular attention to the daily operations of manufacturing facility

and team dynamics.

Page | 7
 Another team shared their lessons learned from Littlefield Technologies. They performed

detailed analysis of demand, contract types, capacity settings and reorder point using basic

operations and mathematical formulas. The four major mistakes they concluded are: the

incorrect documentation, which leads to over-purchasing of machines, the waste of order fees

due to frequent orders, the lack of inventories in some of the days due to missing reorder time

and the failure to evaluate the inventory in the last few days.

2.3 Limitations of Previous Research and Strategies

After taking a closer look at the existing reports on strategies done by other students, we

concluded that these strategies, regardless of results, are experience-based, reached through trial

and error, or in other words “reactive” instead of “proactive”. Although many operations

management concepts and mathematical formulas can provide guidelines for decision making, it

is difficult to apply a rigid algorithm to respond to changes in the complex system.

During the research process, our team tried to understand the mechanisms contained in the

game by dividing the mechanism or model into sectors. After studying the reactions among the

sectors, each sector is analyzed individually.

2.4 System Dynamics Method and Its Benefits

System dynamics is a methodology and a mathematical modeling technique for framing,

understanding, and discussing complex issues and problems. It was created in the mid-1950s by

Professor Jay Forrester of MIT during his research on how science and engineering can

contribute in determining the success or failure of a corporation. Casual loop diagrams, stock and

flow diagrams and equations are the basic components of a system dynamic model.

System dynamics has a wide range of applications as it can usually be applied whenever a

problem can be expressed as variable behaviors through time. Its applications can be expanded
Page | 8
into many fields such as medicine, urban studies, global studies, environmental studies,

information science, history, economic behavior, physics and many more to help with policy

analysis and design.

Applying System Dynamics approach doesn’t mean abandoning all traditional operations

management techniques in this research project. In fact, we have used certain operations

management techniques to design policies, which work as the rules that govern our decisions and

strategies. In other words, system dynamics is tool we use to understand how policies affect

behaviors of a complex system.

Jay Forrester states in Designing the Future, “The model generates streams of decisions

controlled by policies built into the model. The policies make all the decisions step-by-step in

time as the simulation unfolds. Then, if the resulting behavior is undesirable, one searches for a

better set of policies that yield improved resulting.”

2.5 Operations Management Techniques as Policies

There are two major operations management techniques we have used: queuing theory

and inventory management theories, which include Just-In-Time and Economic Order Quantity.

Queuing theory is the mathematical study developed to provide models to predict the

behavior of systems that attempt to provide service for randomly arising demands. In Littlefield

Technologies, queuing theory is used to assess the waiting time of arriving jobs, frequency with

which the queue length exceeds a given limit, the proportion of the machine’s idle time and the

duration of a period during which a machine is continuously occupied. The limitation of using

classical queuing theory is that it is based on data provided in this game. So it may be too

restrictive to be able to model real-world situations exactly. The complexity of production lines

with product-specific characteristics cannot be handled with those models.

Page | 9
 Inventory management includes inventory control and lead-time management. In Littlefield

Technologies, it is used to minimize costs and therefore maximize profit associated with

different contract types. In this project, our team also adopts Just-In-Time, an inventory strategy

that strives to improve the return on investment by reducing in-process inventory and associated

carrying costs.

To control certain inventory level, Economic Order Quantity is used as the optimal order

quantity that minimizes total inventory holding costs and ordering costs. We calculate the

economic order quantity with the assumption that the ordering cost is constant, the rate of

demand is constant, lead time is fixed for each contract type and only ordering and holding cost

are associated with ordering. Because there’s no holding cost in this Littlefield Technologies

simulation game, the EOQ number we get can be as large as infinity. So, it is the best choice to

order only once and order as many as possible to fulfill the need to save the ordering cost.

We also use other operations management techniques such as process analysis, bottleneck

analysis, lean manufacturing, sensitivity analysis, but only as supplements in evaluating the

process, policies and strategies.

Page | 10
3.0 Problem Examination

This section examines the problem in detail. First, an overview of our game-play

experience and the problem is presented. Then, the data acquired from the game-run is described

and explained. Lastly, a dynamic hypothesis of the problem is proposed based on the patterns

observed from the data.

3.1 Game-play Experience and Problem Description

Littlefield Simulation is an open-ended interactive game. The game reveals limited data

to players, cash level and inventory levels are two indicators provided by system. The game

challenges the players to react to changes in the system spontaneously and process new data as

they make decisions, instead of executing a rigid decision model throughout the simulation. In

our game-play, winning strategies were often achieved through trial and error.

The variation of indicators imposes great complexity, which makes it difficult for players

to quickly respond to the changes in the system. The reaction of players is more reactive than

proactive. For instance, when an increase of lead time is observed, we would directly purchase

more machines, instead of tracing the cause of the fluctuation.

The total time span of the game-play was almost five whole days. The time requirement

makes it very difficult to repeat the game.

Page | 11
3.2 Data Review

The data is extracted from the two rounds of Littlefield Technology Simulation we

participated in fall 2010. Three project members played the game in two teams; therefore we

have two sets of data retrieved from the simulation.

We chose to use the data from Team 2’s game-play in our analysis and model calibration,

because the data gives better insight into the difference between contract 1’s and contract 2’s

impact on cash, since Team 2 changed contract type to contract 3 earlier in the game. Also, Team

2 purchased sufficient amount of machines, while the utilization rates of Team 1 peak at 100%

due to lack of capacity. Therefore, Team 2’s data gives better insight into the ideal number of

machines.

3.2.1 Station Queues

2500

2000

1500
Kits queued at Station 1
Kits queued at station 2
1000
Kits queued at Station 3

500

0
145

241
113
129

161
177
193
209
225

257
1

49
17
33

65
81
97

Figure 1-Daily Average Number of Kits for Stations

The graph in Figure 1 shows the kits queued for stations. The queues for three stations go

up steadily as long as we have enough kits. The queues decrease if we run out of inventory.

Page | 12
However the queues slope down dramatically after Day 67 since we purchased more machines at

that point in time. We use this data to study the ideal number of machines.

3.2.2 Station Utilization

Utilization (10-day moving average)

1.2

1 10 days moving
average of utilization
0.8 of machine 1

0.6 10 days moving

average of utilization
0.4 of machine 2
10 days moving
0.2 average of utilization
of machine 3
0
113
129
145
161
177
193
209
225
241
257
1
17
33
49
65
81
97

Figure 2-Utilization for Station1

As shown in Figure 2, utilization of stations fluctuates near 100% before day 69. The

main reason is the shortage of machine capacity. The fluctuations in utilization are due to lack of

inventory (before day 69) and the randomness of order (after day 69). After day 69 we purchased

more machines and the utilization fluctuated at a much more reasonable level than before. We

use 10 day-moving average so we can have a better picture of how the data looks from a general

perspective. This data is used to study the ideal number of machines, discussed later in the

strategy section.

Page | 13
3.2.3 Inventory

Inventory Level in Kits

14,000

12,000

10,000

8,000 Inventory Level in Kits

6,000 Poly. (Inventory Level in

Kits)
4,000

2,000

0
103
120
137
154
171
188
205
222
239
256
273
290
307
1
18
35
52
69
86

Figure 3 Inventory Level

The red dotted line in Figure 3 describes the process of inventory depletion and re-

ordering.

Due to inadequate setting of inventory re-order level; there is a consistent shortage of

inventory before day 69. The re-order level of inventory was reset at day 69, so the shortage of

kits was alleviated. The re-order quantity was raised at day 150, inventory level increased.

Generally speaking the graph is running smoothly after day 69 re-order level set. This dataset is

used to study the re-order point and re-order quantity.

Page | 14
3.2.4 Completed Jobs

Number of Completed Jobs for Each Contract

35
30
25 Number of Completed Jobs
Each Day Contract 1
20
Number of Completed Jobs
15
Each Day Contract 2
10
Number of Completed Jobs
5 Each Day Contract 3
0
113
127
141
155
169
183
197
211
225
239
253
267
1

71
15
29
43
57

85
99

Figure 4 Number of Jobs Completed For Each Contract

Number of Jobs Completed

35

30

25

20 Number of Jobs Completed

15
Poly. (Number of Jobs
Completed)
10

0
106
121
136
151
166
181
196
211
226
241
256
1
16
31
46
61
76
91

Figure 5 Number of jobs completed

Figure 4 and Figure 5 show the number of jobs completed. Due to inventory depletion,

number of completed job at day 17 reached zero. Number of completed job reached peak at day

68 as a result of purchasing new machines, which increased the productivity. The normal

Page | 15
fluctuation of completed job reflects the random number of job arrivals. We use this data to

study the number of jobs arrived each day.

3.2.5 Lead Time

4 Daily Average Job Lead

Time Contract1
3 Daily Average Job Lead
Time Contract2
2
Daily Average Job Lead
1 Time Contract3

0
121
106

136
151
166
181
196
211
226
241
256
1
16
31
46
61
76
91

Figure 6 Daily Average Job Lead Time by Contract Type

Lead time increases when any of the machines reaches 100% utilization. We can see the

lead time has an increasing trend at first due to the lack of machines. The work flow becomes

much smoother after day 69 because of the purchases of new machines. The decrease in the lead

time corresponds to the decreases in machine utilization. We use this data to analyze the required

lead time for each contract and to confirm our estimates of ideal number of machines.

Page | 16
3.2.6 Revenue

1400
1200
1000 daily average revenue
per job set1
800
daily average revenue
600
per job set2
400
daily average revenue
200 per job set3
0

113
129
145
161
177
193
209
225
241
257
1
17
33
49
65
81
97

Figure 7 Daily Average Revenue by Contract Type

2500

2000

1500

1000

500

0
201
101
111
121
131
141
151
161
171
181
191

211
221
231
241
251
261
1

71
11
21
31
41
51
61

81
91

Figure 8 Daily Average Revenue

Figure 7 and Figure 8 show the daily average revenue. The dash line in Figure 8 is the

actual number of daily average revenue generated. It is sum of daily average revenue per job

set1, daily average revenue per job set2 and daily average revenue per job set3.

The solid line in Figure 8 is the 10 day-moving average of the daily average revenue. The

revenue reached 0 on day 17 and day 72. It peaked at day 89 because of the change in contract.

The revenue chart is steadier than most other graphs we have presented because revenue

generated per job is fixed, which implies that when we have enough kits in the inventory and

enough productivity at each station, the daily revenue would not change abnormally. Revenue

Page | 17
generated by each job completion is a built-in parameter and cannot be changed. We use this

data to study the revenue generated from each contract type.

3.3 Dynamic Hypothesis

Figure 9 shows a Causal Loop Diagram (CLD) that will be used as the basis for modeling

the Littlefield simulation. A CLD helps in visualizing how interrelated variables affect one

another. The CLD consists of a set of nodes representing the variables connected together. The

relationships between these variables, denoted by arrows, can be labeled as either positive or

negative.

The Littlefield Simulation CLD consists of seven feedback loops, including six negative

feedback loops, or balancing loops (labeled as “B”) and one positive feedback loop, or

reinforcing loops (labeled as “R”). In a reinforcing loop, each action produces a result which

promotes more of the same action. In a balancing loop, each action attempts to bring two things

to agreement.

The loops are described in detail after Figure 9. The CLD contains the aggregation of two

series of stocks: the Machines stocks and Raw Materials stocks.

Page | 18
 -

Machines
Lead Time
+
B1
B3 -
+
B4
-
Incoming Customer + Customer +
Orders Order
Backlog2 R1 Contract Revenue

Cash
B2 +

- B6 -
B5
-
+
Raw
Materials

Figure 9 Dynamic Hypothesis Model

Explanation of the Feedback Loops

Loop Number B1: Customer Order BacklogMachinesCustomer Order Backlog

Increase in Machines accelerates the depletion of Customer Order Backlog. Increase in

Customer Order Backlog motivates the decision maker to purchase more Machines.

Loop Number B2: CashRaw MaterialsCustomer Order Backlog2MachinesCash

Increase in Cash enables the decision maker to purchase more Raw Materials. Sufficient Raw

Materials enables the Customer Order Backlog to deplete at a predictable rate. Increase in

Customer Order Backlog motivates the decision maker to purchase more Machines. Payment for

Machines depletes Cash.

Page | 19
Loop Number B3: CashMachinesCash

Increase in Cash enables the decision maker to purchase more Machines. Payment for new

Machines depletes Cash.

Loop Number B4: CashMachinesLead TimeContract RevenueCash

Increase in Cash enables the decision maker to purchase more Machines. Increase in Machines

leads to decrease in Lead Time. Less Lead Time induces higher Contract Revenue. Higher

Contract Revenue means higher rate of increase for Cash.

Loop Number B5: CashRaw MaterialsCash

Increase in Cash enables the decision maker to purchase more Raw Materials. Payment for Raw

Materials depletes Cash.

Loop Number B6: CashRaw MaterialsLead TimeContract RevenueCash

Increase in Cash enables the decision maker to purchase more Raw Materials. Sufficient Raw

Materials ensures a certain Lead Time. Less Lead Time induces higher Contract Revenue. Higher

Contract Revenue means higher rate of increase for Cash.

Loop Number R1: CashRaw MaterialsCustomer Order Backlog2MachinesLead

TimeContract RevenueCash

Increase in Cash enables the decision maker to purchase more Raw Materials. Sufficient Raw

Materials ensures a certain Lead Time. Less Lead Time induces higher Contract Revenue. Higher

Contract Revenue means higher rate of increase for Cash.

Page | 20
4.0 System Dynamics Model Description and Testing

This section describes the Littlefield model as it was built using a System Dynamics

approach. The section starts with an overview of the sector diagram. Thereafter, each sector is

described in detail, with justification for the approach chosen to model that sector. The sector

description ends with explanation of all the equations used in the sector.

4.1 Overview of the Sectors

Incoming Stations Lead Time Contracts

Customer
Orders

Cash Financial

Constraints

Machine

Figure 10 Overall Sector Diagram

Figure 10 is an overview of the model for the Littlefield Simulation. The diagram shows

the top level relationships between the different sectors in the model. The input to the game is in

the incoming customer orders sector. The job orders are released to the station sector. The station

sector keeps track of the all the inventory from the raw material to finished goods. When the

order is finished, the number of orders in the system is decreased and the appropriate lead time is

calculated in the lead time sector. Based upon the perceived lead time the contracts sector

chooses the contract which will give the highest possible revenue. Thereafter, the revenue from

Page | 21
the sale of the good is calculated in the financial sector based upon the lead time and the contract

chosen.

The cash constraints sector places a restriction on the number of machines and raw

materials that can be bought based upon the amount of cash available. Whenever raw materials

and inventory are purchased there is a decrease in the amount of cash.

The game player has the ability to control 3 aspects of the game in order to maximize

profit. First, the game player can change the reorder level and reorder quantity of the raw

materials. Second, the game player can increase or decrease the desired lead time of the different

machines by increasing or decreasing the number of machines. Last, the game player can change

the priority between stage 2 and stage 4 of the production process.

4.2 Sector Descriptions

Each sector shown in Figure 10 is described in this section. The incoming customer

orders are the input to the system. Therefore, this is the section described first.

4.2.1 Input to the system

Figure 11 shows the Input of the System sector which consists of two stocks. Customer

Order Queue represents the Customer Order backlog, i.e. the Customer Orders waiting to be

processed throughout the system. A Customer Order is processed when a Job is released to

station 1 queue. Customer Order in the System represents the total amount of Jobs in the system.

When a Job is released, the backlog increases. One constraint setting in the Littlefield Simulation

is that the total amount of customer orders being processed in the system, i.e. Customer Order in

the System, cannot exceed 100.

Page | 22
 The Customer Order Queue and Raw Material Inventory both affect Job Release. A low

Raw Material Inventory to Customer Orders Queue ratio would slow down Job Release. The

impact of the ratio on Job Release is defined with a graphical function, which slopes linearly

from 0 to 1 for ratios less than 1, and stays at 1 afterwards. The graphical function is then

multiplied to Job Release in the Stations sector.

Figure 11 Input of the System

Page | 23
Equations

Stock 1: Customer Order Queue

This stock accumulates all the customer orders waiting to be released to the factory.

Customer_Order_Queue(t) = Customer_Order_Queue(t - dt) +

(Customer_Orders_Incoming - Customer_Orders_Processed) * dt

INITIAL VALUE of Customer_Order_Queue = 0

Inflows: Customer_Orders_Incoming = 12

Outflows: Customer_Orders_Processed = Job_Release

Stock 2: Customer Orders in the system

This stock measures the total orders being processed in the factory.

Customer_Orders_in_the_system(t) = Customer_Orders_in_the_system(t - dt) +

(Job_release_into_the_system - Completed_orders) * dt

INITIAL VALUE of Customer_Orders_in_the_system = 0

Inflows: Job_release_into_the_system = Job_Release

Outflows: Completed_orders = completion_rate

Variable: Raw_Material_to_Customer_Order_Ratio

This variable measures whether there are enough raw materials to process customer

orders

Page | 24
 Raw_Material_to_Customer_Order_Ratio =

Raw_Material_Inventory/Customer_Order_Queue)

Graphical Function: Effect of Raw Material on Job Release

Figure 12 shows the graphical function which shows the effect the ratio of raw materials

to customer orders has on job release. If there is not enough raw materials the customer orders

are released based upon the amount of raw materials available. If the ratio is greater than 1 all the

customer orders are released into the factory.

Effect_of_raw_material_ratio_on_job_release =

GRAPH(Raw_Material_to_Customer_Order_Ratio)

Figure 12 Effect of Raw Material Ratio on Job Release

4.2.2 Cash Constraints

Figure 13 depicts the Cash Constraints sector. It shows the different expenses that will be

encountered in the system, which includes Total Cost of Order and expenses to purchase

Machinery. The constraints affect the Raw Materials and Machines separately. If the Cash

Constraints limit purchase of machines, the system still allows the maximum number of kits to

Page | 25
be bought based upon the present Cash. The consideration is that when raw material is

insufficient, the whole production process stops, and backlog will keep increasing, which would

greatly affect lead time. The purchase of Machinery is only allowed when the Cash Constraint is

not invoked by the algorithm, which has an implicit preference for Raw Materials. This decision

is rather discrete. All Cash Constraints are implemented by calculating the Ratio of Purchase vs.

Cash, and Ratio of Machine Cost vs. Cash. Then, the calculated values are used in graphical

functions. The slopes of the graphical functions describe the Effect of Cash on Purchasing Raw

Materials and Effect of Cash Ratio on Machine, these two have different values.

Figure 13 Cash Constraints

Equations

The two main equations in this sector calculate the ratio of the cost of machines and the cost

of raw materials with the amount of cash available. The machine purchase constraint also has the
Page | 26
cost of raw materials included in it because a machine can be bought only if there is enough cash

for raw materials. This places a priority for raw materials because if there are no raw materials

production will stop.

a) Raw Material Purchase Constraints

Variables

Ratio_of_purchase_vs_cash = Cash/Total_cost_of_order

Total_cost_of_order = fixed_cost+cost_per_job*raw_material_reorder_quantity

Constants

fixed_cost = 1000

cost_per_job = 600

b) Machine Purchase Constraints

Variable

ratio_of_machine1_cost_vs_cash = Cash / (cost_of_machine_1+Total_cost_of_order)

Graphical Functions

a) Raw Material Purchase Constraints:

Effect_of_cash__on_purchases_on_raw_materials =

GRAPH(Ratio_of_purchase_vs_cash)

Page | 27
 Figure 14 Cash Constraint on Raw Materials

Figure 14 shows the linear function that models how the purchase of raw materials is

affected as the ratio of the cost of raw materials to cash changes. The function is a straight line

from 0 to 1 for a ratio of less than 1 and 1 for higher ratios. This is because raw materials are a

high priority investment because without it production will stop. Therefore if there is not enough

cash to buy all the raw materials required, the cash will be used to buy as many as possible.

b) Effect of Cash Ratio on Machine

Effect_of__cash_ratio_on_machine1 = GRAPH(ratio_of_machine1_cost_vs_cash)

Figure 15 shows the graphical function that models the purchase of machine based upon

the ratio of the cost of buying the machines to cash available. Since machines are lower priority

than raw materials the machine will be bought only if there is enough cash. Therefore the graph

is 0 for ratios less than 1 and 1 for ratios equal to or greater than 1.

Page | 28
 Figure 15 Cash Constraints on Machines

4.2.3 Financials

The Financial Sector, shown in Figure 16, has one stock called Cash, which stands for

the total amount of Cash at any given time. The inflow to Cash, Revenues per Day, includes

Total Machine Sales, Revenue from Completing Orders, and Revenue from Interest. Interest is

earned from present Cash.

The Total Machine Sales is calculated by multiplying Prices of Machines Sold with the

Total Machine Sales.

The Interest Rate on present Cash is 10%. This Interest is compounded per day. In order

to calculate the Interest Rate per Day the following equation is used:

(1+x)^365=1.0026.

This gives the Interest Rate per Day of 0.0026%. This is multiplied to present Cash and added

to the Cash Value every day.

Page | 29
 The Revenue from Completing Orders is calculated by multiplying the revenue per batch

with completion rate, which is explained in the contract sector.

The Outflows of Cash include expenses from Ordering Raw Material and Total Machine

Expenses. An if-then-else is used in the Constant Change in Machine to check whether a

machine is being bought or sold.

The Total machine expenses are calculated by multiplying the different machine prices

with the number of machines bought. The variable, Expenses from Ordering Raw Materials, is

calculated by multiplying the raw material purchase rate with the fixed cost over the raw material

reorder quantity and adding cost per job.

Figure 16 Financial Sector

Page | 30
Equations

Stock: Cash

The financial sector has 1 stock which is for the amount of cash. The two flows are

revenues and expenses. The main revenues are from completing orders and sale of machines.

Cash can also increase from interest earned on it. The expenses are from the purchase of raw

materials and machines.

Cash(t) = Cash(t - dt) + (Revenue_per_day - Expenses) * dt

INITIAL VALUE of Cash = $1,000,000

Inflow: Revenue_per_day =

Revenue_from_completing_orders+Revenue_from_Interest+Total_Machine_Sale

Outflow: Expenses =

Expenses_from_Ordering__Raw_Material+Total_Machine_Expenses

Variables

Revenue_from_completing_orders = completion_rate*Revenue_per_batch

Revenue_from_Interest = Cash*Interest_rate

Total_Machine_Sales = Machine_1_Sales+Machine_2_Sales+Machine_3_Sales

Total_Machine_Expenses = Machine1_expense +Machine_2_Expense +

Machine_3_Expense

Expenses_from_Ordering__Raw_Material = raw_material_purchase_rate *

(fixed_cost/raw_material_reorder_quantity+cost_per_job)

Page | 31
Constants

Interest_rate = .0026

Price_of_machines_sold = 10000

cost_of_machine_1 = 90000

cost_of_machine_2 = 80000

cost_of_machine_3 = 100000

Functions

Machine_1_Sales =IF(Change_in_Machine_1<0) THEN

SQRT((Price_of_machines_sold*Change_in_Machine_1)^2) ELSE

0Machine1_expense = IF(Change_in_Machine_1>0) THEN cost_of_machine_1

* Change_in_Machine_1 ELSE 0

Machine_2_Expense = IF(Change_in_Machine_2>0) THEN cost_of_machine_2

*Change_in_Machine_2 ELSE 0

Machine_2_Sales = IF(Change_in_Machine_2<0) THEN

SQRT((Price_of_machines_sold*Change_in_Machine_2)^2) ELSE 0

Machine_3_Expense = IF(Change_in_Machine_3>0) THEN cost_of_machine_3

*Change_in_Machine_3 ELSE 0

Machine_3_Sales = IF(Change_in_Machine_3<0) THEN

SQRT((Price_of_machines_sold*Change_in_Machine_3)^2) ELSE 0

Page | 32
 These discrete functions above change from 0 to 1 based upon whether a machine is

being sold or bought.

4.2.4 Contracts

Figure 17 shows the contracts sector. This sector is used to select the contract type based

upon the present Average Lead Time. Current Average Lead Time affects Perceived Lead Time,

which affects the contract type chosen. The process utilizes if-then-else functions to determine a

contract type based on the range of Perceived Lead Time. Perceived Lead Time is a smooth

function using Average Lead Time and is basically used to signal the delay it takes for people to

perceive the actual Average Lead Time. The contracts types have either a value of 1 or 0,

depending on if it is chose or not.

The Average Lead Time also affects the Revenue per Batch as the lead time effects the

revenue on the order based upon the contract chosen.

Contracts

Perceiv ed lead time's

ef f ect on Contract Ty pe1 Contract Ty pes 1

Average Lead Time

Contract Ty pe 2
Perceiv ed Lead Time Perceiv ed Lead Time's
ef f ect on Contract Ty pe2

Rev enue per batch

Contract Ty pe 3

Perceiv ed Lead Time's

ef f ect on Contract Ty pe3
~
Average Lead Time Contract Ty pe 1
~
rev enue per batch
Contract Ty pe 2
Rev enue per Batch

~
Contract Ty pe 3
Rev enue per Batch

Figure 17 Contracts Sector

Page | 33
Equations

This sector models the human reactions on changing contracts based upon lead time.

Revenue_per_batch =

Contract_Types_1*Contract_Type_1_revenue_per_batch+Contract_Type_2*Cont

ract_Type_2_Revenue_per_Batch+Contract_Type_3*Contract_Type_3__Revenu

e_per_Batch

The revenue per batch calculates the revenue on each order based upon the contract

quoted and its lead time and the actual lead time the contract was completed in.

Perceived_Lead_Time = SMTH3(average_lead_time,3.5,1.5)

The perceived lead time variable is modeled as a delay of average lead time using the

Smooth function.

Perceived Lead Time's Effect on Contract Type

The discrete functions of the perceived lead time’s effect on the contract decide which

contract to choose based on the perceived lead time.

Perceived_lead_time's_effect_on_Contract_Type1 = IF (Perceived_Lead_Time>1.2)

THEN 1 ELSE 0

Perceived_Lead_Time's__effect_on_Contract_Type2 = IF(Perceived_Lead_Time>.6)

AND (Perceived_Lead_Time<=1.2) THEN 1 ELSE 0

Page | 34
 Perceived_Lead_Time's_effect_on_Contract_Type3 = IF(Perceived_Lead_Time<=.6)

THEN 1 ELSE 0

Revenue per Batch

The three graphical functions (Figure 18, Figure 19 and Figure 20) help to calculate the

revenue earned for each order based upon the actual lead time and the contract selected.

Contract_Type_1_revenue_per_batch = GRAPH(average_lead_time)

Figure 18 -Contract Type 1 and Lead Time

Contract_Type_2_Revenue_per_Batch = GRAPH(average_lead_time)

Figure 19- Contract Type 2 and Lead Time

Page | 35
 Contract_Type_3__Revenue_per_Batch = GRAPH(average_lead_time)

Figure 20 -Contract Type 3 and Lead Time

4.2.5 Average Lead time

Figure 21 shows the Average Lead Time sector. This sector shows the calculation of

Average Lead Time. It adds the actual delays in stations 1 to 4, or Actual Time 1 to 4 and the time

spent in customer backlog, Job Release Lead Time. The sum of all these is the Average Lead

Time.
Lead Time

Job Release
Lead TIme

Actual Time 1

actual time 2 Av erage Lead Time

actual time 3

actual time 4

Figure 21 Lead Time Sector

Page | 36
Variable:

This sector calculates the actual average lead time based upon the lead time from all the

stations and machines.

Average_Lead_Time = Job_Release__Lead_Time + Actual_Time_1 +actual_time_2

+actual_time_3 + actual_time_4

4.2.6 Machines

The machine sector is shown in Figure 22. This sector has 3 stocks; one stock for each

machine type. The machine stocks are named Machine 1, Machine 2 and Machine 3

The number of Machines will change based upon the Desired Number of Machines over a

constant Adjustment Time. The Adjustment Time is an arbitrarily chosen constant through game

play experience.

Positive Change in Machine can only happen if there is enough Cash. This is done

through the Cash Constraints for machines, which is explained in the Cash Constraints sector.

Page | 37
 Figure 22 Machines Sector

Equations

This sector models the machines and purchase and sale of them. The numbers of

machines are modeled as stocks with their initial values based upon the parameters played in the

actual game. The change in machines is based upon the difference between desired machines and

actual number of machines and based upon cash constraints.

Page | 38
Stock: Machine

machine1(t) = machine1(t - dt) + (Change_in_Machine_1) * dt

INITIAL VALUE of machine1 = 3

INITIAL VALUE of machine 2=1

INITIAL VALUE of machine 3=1

Constant

Desired Machines= (User Chosen Value)

Adjustment_Time = 1

Variable

Difference in Machines= Desired_Machines-machine1

Function

Change_in_Machine_1 = Change_in_Machine_1 = IF (TIME <51) OR (TIME >218)

THEN 0

ELSE(Difference_in_Machines/Adjustment_Time)*Efffect_of__cash_ratio_on_

machine1

The above function shows the change in machine which represent whether a machine is

bought or sold. A positive value means a machine is bought and vice-versa. It is an if-then-else

function as the Littlefield game has constraints on when machines can be bought and sold: No

machines can be bought before day 50 and after day 218. Furthermore the machines can be

Page | 39
bought only if there is enough cash in the cash sector. This effect is shown by the multiplication

of the variable Effect of cash ratio on machine

4.2.7 Stations

The stations sector is shown in Figure 23. It consists of 6 stocks. The Stations sector is

the main sector in the model. The stocks show the pipeline of the production process from the

Order of Raw Material, Raw Material Inventory, and the Queues at each of the 4 steps.

The order for raw material is affected by Raw Material Purchasing Rate and raw

material arriving rate. As long as there is enough Cash to support the Customer Order, The Raw

Material Purchase Rate equals to Reorder Quantity when the sum of Raw Material Inventory

and Raw Material Order is less than Reorder Level. And the stock of Order for Raw Material

depletes over the time of Order Delay.

Arriving raw material orders add to Raw Material Inventory. A Job is Released when

there is Customer Order in the queue, the Raw Material Inventory is sufficient and the Total

Orders In The System do not exceed the maximum level. The time it takes to release a job is

dependent on Job Process Time, which is a constant value, and the Queue’s Effect On Job

Process Lead Time.

Thus, the Job is Released into Queue for Step 1. The Processing Rate for each step

depends on the Queue length for that particular queue, and the Actual Time needed to process a

Job, which is affected by the Queue length and a Normal Processing Time. The queue will not

go negative. A Normal Processing Time is determined by each Machine’s Capacity and the

number of that particular type of Machines. The Machine Capacity fluctuates around an

Average Service Rate with a certain Standard Deviation.

Page | 40
 Figure 23 Stations Sector

Equations

Stock 1: Order for Raw Material

The stock measures the number of raw materials ordered and in the pipeline to be

delivered to the factory.

Order_for_Raw_Material(t) = Order_for_Raw_Material(t - dt)

+(raw_material_purchase_rate - Raw_Material_Arriving_Rate) * dt
Page | 41
 INITIAL VALUE of Order_for_Raw_Material = 0

Inflows: raw_material_purchase_rate = raw_material_reorder_quantity *

Inventory_Ratio's_effect_on_Reordering * Effect_of_cash__on_purchases

Outflows: Raw_Material_Arriving_Rate = Order_for_Raw_Material/order_delay

Variable: Inventory Ratio

The inventory ratio measures whether the current stock of raw materials and the raw materials

waiting to be delivered are greater than or below the reorder level.

Inventory__Ratio =

(Raw_Material_Inventory+Order_for_Raw_Material)/Reorder__Level

Constants

Reorder__Level = 24

Raw_Material_Reorder_Quantity = 120

Stock 2: Raw Material Inventory

This stock measures the total number of raw materials in the system. Jobs are released

into station 1 if the total number of orders in the factory is less than 100 and if there are enough

raw materials.

Raw_Material_Inventory(t) = Raw_Material_Inventory(t - dt) +

(Raw_Material_Arriving_Rate - Job_Release) * dt

INITIAL VALUE of Raw_Material_Inventory = 160

Inflows: Raw_Material_Arriving_Rate = Order_for_Raw_Material/order_delay

Page | 42
 Outflows: Job_Release =

Customer_Order_Queue*Total_orders'_effect_on_Job_Release

*Effect_of_raw_material_ratio_on_job_release/Job_Release__Lead_Time

Variable

Job_Release__Lead_Time = (Customer_Order_Queue+1) *Job_Processing_Time

Constants

The Job Processing Time and Job Release Lead Time are used to calculate the total time

an order has to wait before it is released.

The order delay is set to 3 as it takes 3 days for the raw material orders to get delivered.

Job_Processing_Time = 1/48

order_delay = 3

Stock 3: Queue for Step 2

This stock shows the material from station 1 waiting to be processed at station 2. Once

jobs are processed they are released to station 3.

queue_for_step2(t) = queue_for_step2(t - dt) + (step1_process_rate- step2_process_rate)*

dt

INITIAL VALUE of queue_for_step2 = 0

Inflows: step1_process_rate = queue_for_step_1/Actual_Time_1

Outflows: step2_process_rate = queue_for_step2/actual_time_2

Page | 43
Variables

The normal processing time (NPT) measures the time it takes the station to process an

order. The actual time measures the time taken for the order waiting in queue and for the time it

takes to be processed. The priority for step 2 variable shows how much of the capacity of

machine 2 is set for queue 2 as opposed to queue 4. Below are formulations for station 1 and

station 2. Variables machine1 and machine2 are the number of machines in each station.

Actual_Time_1 = Normal_Processing_Time_1*effect_of_capacity__utilization_1

effect_of_capacity__utilization_1 = 1+queue_for_step_1

Normal_Processing_Time_1 = 1/ (machine1*machine1_capacity)

actual_time_2 = effect_on_cap_ut_2*NPT_2

effect_on_cap_ut_2 = 1+queue_for_step2

NPT_2 = 1/ (machine2*machine2_capacity*priority_for_step_2)

The machine1_capacity variable is a variable used to calculate the capacity for machine 1

and is used similarly for other machine. The NORMAL function is a statistical function which

outputs random numbers of the mean specified (average service rate) and specified standard

deviation.

machine1_capacity = NORMAL(average_service_rate1,std_dev1, 5)

Constants

std_dev1 = 4

average_service_rate1 = 12

Page | 44
 std_dev2 = 10

Average_Service__Rate_2 = 30.6

priority_for_step_2 = 0.5

Function: Inventory Ratio’s Effect on Re-ordering

The inventory ratio’s effect on reordering variable is a discrete function which checks

whether reorder level is less than the number of raw materials in the inventory and order

pipeline.

Inventory_Ratio's_effect_on_Reordering=IF(Inventory__Ratio>1) THEN(0) ELSE (1)

Page | 45
4.3 Model Testing

We use extreme value testing and input response testing to verify the robustness of the

model. We use sensitivity analysis to evaluate the model’s accuracy. A well-constructed model

should yield sensible and rational results under extreme conditions.

Most model behaviors are predictable. For example, if the machine purchase dramatically

increases, the cash level should drop significantly, and the average lead time should decrease as

well. Extreme value testing gives confidence of the model’s validity.

All tests in this section follow the rule of the simulation that all settings could not be

changed until 50 days in the game.

4.3.1 Extreme Value Testing

Extreme value testing establishes confidence of a model’s validity. The parameters

chosen for extreme value testing include: incoming customer orders, desired number of

machines, re-order level and quantity. All four parameters are exogenous variables. In Littlefield

Simulation game, the number of incoming customer orders per day is generated by the system.

Players can change the number of machines bought and sold, the re-order level, and quantity.

4.3.1.1 Incoming Customer Order

The average number of incoming customer orders per day is around 12. The range for the

customer orders is from 3 to 24. Extreme values for incoming customer orders are used to test

the model. If the model generates plausible behavior when tested with extreme values for

incoming customer orders, then this builds confidence in the model’s corresponding elements.

Page | 46
 The values tested are 0 and 100. All other parameters are kept to their initial values at the

start of gameplay. In Figure 24, the blue line shows the cash level over time when the number of

incoming customer orders per day equals to 0. The cash level only increased slightly, which is

due to the revenue earned from interest. The red line shows the model behavior when the number

of incoming customer orders per day equals 100.

The cash level increases quickly at first when the orders are fulfilled within the lead time

required by then contract. As customer order backlog increases and the lead time increases as a

result, the cash level starts to increase at a lower rate which is only slightly higher than the slope

of the blue line. The reason for the difference in slope is that in the second case, the cash level is

higher than in the first case, which brings in more interest.

Figure 25 shows the trends of lead time when the number of incoming customer order

equals zero and 100. When the value equals zero, lead time is constantly zero, since there are no

jobs being processed. When incoming customer order is 100, the lead time steadily increases as

the customer order backlog increases.

Page | 47
 Cash: 1 - 2 -

1: 1250000

1: 1125000 2

1
1
1
1
1: 1000000 1
1.00 54.40 107.80 161.20 214.60 268.00
Page 1 Day s 4:35 PM Fri, Oct 14, 2011
Cash

Figure 24 Cash Level for Extreme Values of Customer Orders

Perceiv ed Lead Time: 1 - 2 -

1: 200

1: 100
2

1: 0 1 1 1 1
50.00 104.50 159.00 213.50 268.00
Page 2 Day s 4:37 PM Fri, Oct 14, 2011

Figure 25 -Lead Time for Extreme Values of Customer Orders

Page | 48
4.3.1.2 Desired Number of Machines

The initial numbers of machines for station 1, station 2 and station 3 respectively are

three, one, and one. Extreme value testing for desired machine levels sets these values at greater

and lower levels than the initial values. If in response to the test the model still generates

reasonable behavior, then we build confidence in the model’s corresponding elements.

The values selected for testing are zero and 20, for each machine. All other variables are

kept to the same initial values at the start of the gameplay. The blue line in Figure 25 shows the

cash level when desired number of machines equals zero; the red line shows cash level when

desired number of machines equals 20.

When desired number of machines is zero, cash level keeps on gradually increasing

because of the revenue from machines sales in first 30 days and the interest earned throughout

the game. When the desired number of machines is set to 20 at day 50, the cash level

immediately drops, and stays constant at a low value throughout the game-run. The reason is that

purchasing machines takes considerable amount of cash. The value that cash level stays constant

at is affected by cash constraint settings, which limit investment in machines based on current

cash level.

Figure 27 shows the lead time’s responsive behavior for extreme values of desired

number of machines, the blue line correspond to zero and the red line correspond to 20.

When there is not a single machine in the system, jobs cannot be processed due to the

lack of capacity. As a result, lead time keeps increasing linearly as the customer order backlog

increases. When the machines are being purchased too quickly, the lead time still increases

linearly as the aggressive investment in machines hinders the purchase of raw material, and thus

stops production. This increases both customer order backlog and in the lead time.
Page | 49
 Figure 26 Cash Level for Extreme Values of Desired Machines

Figure 27 Lead Time for Extreme Values of Desired Machines

4.3.1.3 Reorder Level and Reorder Quantity

The initial values for re-order level and re-order quantity, in the game, were 24 and 160

correspondingly. Extreme value testing for re-order level and re-order quantity requires values

much greater than and much less than the initial values. However, high values for re-order
Page | 50
quantity are part of the profit maximizing strategy as discussed in section 5.1.5. High values for

re-order level do not have significant impact on the system. Therefore, only extremely low

values for re-order level and re-order quantity need to be tested in extreme value testing.

Zero is selected for both tests for re-orders level and re-order quantity. All other variables

are kept to the same initial values at the start of the simulation, while the two parameters tested

are set to zero at separate times. Figure 28 and Figure 29 show cash level and lead time’s

response behavior. The blue lines show the model’s reaction when re-order level equals zero, and

the red lines show the model’s reaction when re-order quantity equals zero.

As it takes four days for the ordered raw materials to arrive, lead time increases linearly

due to the insufficient raw materials when re-order level is set to zero. When re-order quantity is

set to zero, production would stop due to the lack of raw material. In both cases, customer order

backlog will keep increasing and lead time will keep increasing linearly. The cash earned

through interest contributes to cash level in both cases. The cash level increases slowly when re-

order level equals zero since only a certain amount of jobs can be processed. The cash decreases

Page | 51
slowly since there are no jobs.

Figure 28 Cash Level for Zero Reorder Level and Reorder Quantity

Figure 29 Lead Time for Zero Reorder Level and Reorder Quantity

Page | 52
4.3.2 Input Response Testing

This section tests the model’s responsive behavior to different input patterns. In the

Littlefield game the incoming customer orders are random values that follow a normal

distribution, with a mean of approximately 12 and a standard deviation of 3. As the standard

deviation is of statistical significance it is prudent to test the model with different inputs. The

rationale is that the intrinsic behavior of the model should not be affected by variation in inputs.

The testing starts with constant incoming customer orders. Afterwards, we test the step input,

sine wave input, and random input. In each case, the response of cash level is compared to the

parameters’ response to constant input. The similarity of the responses establishes confidence in

the model, and serves as a basis for strategy design.

4.3.2.1 Constant Incoming orders

Figure 30 shows the cash response with constant incoming customer orders. This is the

baseline run and the other input responses are compared to it. Figure 30 shows that not much

cash is earned throughout the simulation.

Page | 53
 Cash: 1 -

1: 1500000

1
1
1
1

1: 1250000

1: 1000000
1.00 54.40 107.80 161.20 214.60 268.00
Page 1 Day s 2:52 PM Sat, Oct 15, 2011
Cash

Figure 30 Cash Level With Constant Incoming Customer Orders

4.3.2.2 Step Response

The step response shows how quickly the system returns to steady state, and if there is

unwanted endogenous relationships in the model which may stop the system from returning to a

steady state.

The red line in Figure 31 shows the cash level’s response to a step input. In the base run,

the cash level increases at a steady but slow rate after 50 days. Cash level’s response of the step

input of an equal magnitude as the base run, is to start rising at the same steady slow rate after

the adjustment period.

In the testing, incoming customer orders are changed to the following equation:

0+STEP (AMPLITUDE= 12, TIME = 52)

Page | 54
 Cash: 1 - 2 -

1: 1500000

1
1
1 2
2
1
2

2
1: 1250000

1: 1000000 2
1.00 54.40 107.80 161.20 214.60 268.00
Page 1 Day s 2:29 PM Sat, Oct 15, 2011
Cash

Figure 31 Cash Level Step Response

4.3.2.3 Sine Wave Input

The red line in Figure 32 shows the system’s response to a sine wave input. The

incoming customer orders equation is changed to:

12+12*SIN (2*pi*TIME/5)

The sine wave has the amplitude of 12 and a period of 5. The value varies between 0 and

24.

The sine wave response deviates only slightly from the constant input response;

therefore, we conclude that the system is very stable.

Page | 55
 Figure 32 Cash Level Sine Wave Response

4.3.2.4 Random Input Response

The input response testing culminates with the random input response testing. The

random input response is important as it resembles the actual situation in Littlefield simulation,

where the input is random. The red line in Figure 33 shows the response of the system to a

random input. The incoming customer order was changed to:

(NORMAL (12, 3) +RANDOM (-6, 9))

The red line shows little deviation from the constant input response; therefore, we

conclude that the cash level is almost independent from the type of input as long as the mean of

the input is constant. This conclusion simplifies later tests as the incoming customer orders can

be kept constant.

Page | 56
Figure 33 Cash Response to a Random Input

Page | 57
5.0 Game-playing Strategies for Maximizing Profit

The game player has control over 3 aspects of the game to maximize profit. Firstly, the

reorder level and reorder quantity of the raw materials can be changed. Secondly, desired lead

time of the different machines could be changed; this decision directly affects purchasing and

selling of machines. Lastly, priority between stage 2 and stage 4 of the production process can be

changed.

5.1 Analysis of Current Strategy

5.1.1 Initial Condition and Input to the System

Figure 34 Incoming Customer Orders per Day

Understanding the trend of incoming customer orders (Figure 34) is essential to decision-

making. In the simulation, the volume of incoming customer orders greatly affects decision-

Page | 58
making on items such as machines purchases, re-order level and quantity settings, and the

priority of stations. These decisions are often based on the forecast of incoming customer order

in the model. As listed in Figure 35, means, standard deviations, and range are calculated. The

mean shows that the volume of incoming customer orders fluctuates around 12 per day.

Parameter Value
Average 11.91
Standard Deviation 3.28
Range 3~24

Figure 35-Mean, Standard Deviation and Range of Incoming Customer Order per Day

Range Frequency Weight

15 < orders 59 22.01%
8<=orders<=15 149 55.60%
orders < 8 60 22.39%

Figure 36- Order Distribution

Figure 36 shows order distribution. The bulk of orders are between 8 and 15 units. The

distribution itself appears to be symmetrical.

5.1.2 Contract Selection

There are three contracts available in the Littlefield game; for each contract selected,

revenue per job decreases as the lead time increases. There are two feasible strategies: alternating

contract during the gameplay, and selecting one contract throughout the gameplay. If the

alternate contract strategy is chosen, purchasing additional machines would be unnecessary, but

the latest trend of incoming customer orders needs to be tracked closely: when the incoming

customer orders is low, and machine utilization rates are low as a result, contract 3 should be

selected; when the incoming customer orders is high and the machine utilization rates are high,

Page | 59
contract 1 or contract 2 should be selected, based on the current number of machines. It is

difficult to determine the timing of contract change as well.

If a fixed contract is chosen, more cash would be spent on machine purchasing.

Aggressive machine-purchasing in the initial stage leads to relatively low utilization rates. As a

result, fewer adjustments of the number of machines are required during the gameplay. It is

noticeable that contract 3 grants the highest profitability and contract 1 offers the lowest. The

daily revenue from contact 1, 2 and 3 are $750, $1000, and $1250 respectively, when the orders

are completed within required lead time.

Based on the analysis above, we recommend choosing a fixed contract strategy with

contract 3.

5.1.3 Ideal Number of Machines to Purchase

The maximum lead time without losing revenue for contract 3 is one day. The loss is in

both in cash and opportunities to generate profit. Therefore, it is recommended to keep the lead

time below one day.

According to the forecast mentioned in the section 5.1.1, the possible highest mean (+1

standard deviation) of the number of incoming customer orders is 15 per day, there are 22% of

the days when the number of incoming customer orders is greater than 15, and the maximum

value is 20 per day.

Therefore, when selecting fixed contract strategy with contract 3, the ideal capacity

should be around 17 to 18 jobs per day (+2 standard deviation). The utilization rates will be

around 70% under this setting, according to the data from game-run.

Page | 60
 A high process rate of station 1 is desired to guarantee sufficient time for stations 2 and 3

to process their respective queues.

In the game-run, we had five machines and the peak utilization level was around 70%

(Refer to the graph Station Utilization Station 1 from Sec. 3.2). Reducing the number of

machines leads to an increase in utilization rate. Specifically, when reducing the number of

machines at station 1 to four, the utilization starts to fluctuate around 90%, and peaks at 100%. It

increases lead time, and therefore affects revenue. We conclude that the ideal number of

machines at station 1 is five.

Similarly, we analyzed station 2 and station 3 utilization rates from our Team 2 data. The

conclusion is that the ideal number of machines for station 1, station 2 and station 3 is five, three

and two, respectively.

5.1.4 Re-order Level

Since fixed contract strategy with Contract 3 is selected, raw material shortage would

lead to a loss in profit. Safety stock is a level of extra stock that is maintained to lessen the

danger of this raw material shortage due to uncertainties in customer orders. A commonly used

approach for the calculation of the safety stock is based on the average customer orders, the lead

time it takes for raw materials to arrive after the order is placed, the service level-the desired

probability that the chosen level of safety stock will not lead to a raw material shortage- and the

standard deviation of customer orders.

Page | 61
 The safety stock is calculated below:

Z=1.64 is selected to achieve 95% service level.

√ =

√ jobs

The factors that determine the appropriate re-order point are the safety stock, average

incoming customers and the lead time. The re-order point calculation is shown below:

jobs

We chose 60 jobs as re-order level.

5.1.5 Re-order Quantity

In the game, cash earns 10% per year. Accordingly, the daily interest rate is:

Therefore, the break-even point for ordering raw material each day is:

When there is $3.85 million cash, the revenue from interest equals to the cost of an order.

It means if cash is over $3.85 million, there will be an opportunity cost resulting from ordering

raw material. Considering actual situations of the game-play, the opportunity cost is not

significant.

The Economic Order Quantity (EOQ) is calculated below:

Page | 62
 √

Since there is no holding cost in this game, the theoretical value for EOQ is infinitely

large. Therefore, it is recommended to set a large re-ordering quantity to reduce cost.

5.1.6 Machine Priority

Experiment result with the model indicates that the priority of station 2 machines will

only affect the lead time when capacity is insufficient. Since both the utilization of station 2 and

of station 3 stay steadily below 80% in our gameplay, we concluded that the priority of station 2

won’t affect our lead time and revenue. Therefore it’s recommended to use default setting of

priority, which is first-in-first-out.

5.2 Predictions of Model Behavior

Calculation indicates that the maximum ending cash is $3.5 million. The calculation

process is described in the following paragraph.

Daily job arrival has an average of 12 and a standard deviation of 3. Revenue per job for

contract 3 is $1,250. Cost for kits required per job is $600. Daily profit is calculated as:

( ) ( )

The maximum profit that can be generated is:

The initial capital is $1 million, and the revenue generated from first 50 days is around

$0.8 million. Adding all three values up, we conclude that the maximum ending cash is around

$3.5 million.

Page | 63
 Due to the complexity in calculations and model structure, small discrepancies between

expected value and actual value are inevitable.

Page | 64
5.3 Profit-Maximizing Strategy Implementation and Testing

This section tests the strategy designed in the previous section. First, the model is

changed appropriately to include a policy sector. Thereafter, incremental changes are made until

all the policy changes take effect. For each change made, the model is simulated so that effects

of the change can be seen. The model testing section showed that the system response is almost

independent of the type of input; therefore, in order to simplify analysis only constant orders will

be used. When one parameter is changed, all the other parameters are kept constant. Unless

stated, the parameter values are always the same as during the initial Littlefield game. These

parameters are shown in the Appendix Section 9.3 The following strategies are tested:

1) Initial run with reactive strategy

2) Changing contract type to contract 3

3) Implementing optimum re-order level and re-order quantity

4) Changing the number of machines for maximum utilization

In each case the effectiveness of the change in the overall policy design is verified by

seeing the increase in cash level at the end of the game. In the end, the strategy is tested with

random input for verification. The random input is similar to the input in Littlefield Simulation.

Note that the settings cannot be changed until day 50 according to the rules of Littlefield

Simulation.

Page | 65
5.3.1 Initial Run with Reactive Strategy

As shown in Figure 37, when the model runs with the initial settings given in the game,

the ending cash level on day 268 is $1. 478 so the profit earned is $0.478 million

Cash: 1 -

1: 4720000

1: 2860000

1 1
1

1
1: 1000000
1.00 67.75 134.50 201.25 268.00

Page 1 Days 2:46 PM Sun, Oct 09, 2011

Cash

Figure 37 Cash level with a reactive strategy

During the test-run, contract 1 is selected for the first 50 days and after 50 days. Selecting

contract 1 is one of the initial settings of the game. Figure 38 below shows the 3 different

contract types. When the value for the contract type is 1 that particular contract has been chosen.

When the value for the contract type is 0; that contract has not been chosen. After 50 days, the

model automatically selects the contract 1, which is a result from the built-in algorithm that

selects the most profitable contract type based on perceived lead time.

Page | 66
 1: Contract Types 1 2: Contract Type 2 3: Contract Type 3

1: 1 1 1 1 1
2:
3:

1:
2: 1
3:

1:
2:
3: 0 2 3 2 3 2 3 2 3
1.00 67.75 134.50 201.25 268.00

Page 2 Days 3:07 PM Sun, Oct 09, 2011

Untitled

Figure 38 Contract Type Chosen When Policy Is Not Used

Since the objective is to achieve the maximum cash level at the end of the game, it is

recommended to choose contract 3 from day 51 to day 268.

5.3.2 Switching to Contract 3

The analysis of revenue per contract, suggests that contract 3 brings in the highest profit.

In this section, we are testing selecting contract 3 as a strategy.

The model is modified to include a policy sector. The goal of the sector is to choose one

contract type after day 50. The sector is used to select contract 3 throughout the game. The

policy sector is shown in Figure 39.

Page | 67
 Figure 39 Policy Sector

Policy Sector Description

The lead time is smoothed for a 1 day period since the fluctuations in the lead time cause

errors in the calculations of the order revenue:

smooth_of_average_lead_time = SMTH1(Average_Lead_Time,1)

Contract type 1 revenue per batch

Figure 40 Contract Type 1 Revenue Graph

A lookup function in Figure 40 is used to calculate the revenue for the first 50 days when

contract type 1 is selected

Page | 68
Contract type 3 revenue per batch

Figure 41 Contract Type 3 Revenue Graph

A lookup function in Figure 41 is used to calculate the revenue from day 51 to day 268.

The following formula calculates the revenue per order:

Revenue from orders= IF (TIME<50) THEN Contract_Type_1_revenue_per_batch_2

ELSE Contract_Type_3__Revenue_per_Batch_policy.

Page | 69
 Results:

Cash: 1 - 2 -

1: 2000000

1: 1500000 2
2 1
2 1
1

1: 1000000
1.00 67.75 134.50 201.25 268.00

Page 1 Days 3:48 PM Sun, Oct 09, 2011

Cash

Figure 42 Cash Level when Contract 3 is Chosen

Figure 42 shows that proactively choosing contact 3 increases the cash level to

$1,504,409 (red line). It is $26,845 higher than reactively choosing contracts, the strategy

presented in last sector.

5.3.3 Implementing optimal re-order level and re-order quantity

As discussed in section 5.1.5, the optimal values for the re-order level and re-order

quantity can be determined. The theoretical optimal value for the re-order level is 60 jobs, or

3600 kits. That calculation, however, did not include purchasing’s effect on the revenue

generated from interest. The theoretical optimal value for the re-order quantity is infinitely large.

However, since the game has a limited number of days and the goal is to maximize cash at the

end of the simulation, over-purchasing raw materials is not recommended. Therefore, the optimal

re-orders level and re-order quantity should be determined through experimentations with the
Page | 70
model. The experiment results showed that the optimal value for the re-order level is 100 and the

optimum value for re-order quantity is 2500.

Cash: 1 - 2 -

1: 2500000

1: 1750000 2

1
1
1
2
2

1: 1000000
1.00 67.75 134.50 201.25 268.00

Page 1 Days 4:03 PM Sun, Oct 09, 2011

Cash

Figure 43- Implementing Economic Reorder Quantity and Economic Reorder Level

In Figure 43, the blue line shows the level of cash when choosing contract 3, and using

the initial settings with 24 as re-order level and 160 as re-order quantity, while the red line shows

the level of cash over time when the optimum values are implemented: 100 as re-order level and

2500 as re-order quantity. The optimum value’s ending cash level is $1.896 million which is

$0.391 million higher than the result of strategy in 5.2.2. There is a dip is cash level on the red

line because of a large amount of raw material is bought

5.3.3.1 Optimum Reorder Level

This experiment shows that increasing the re-order level above 100 does not lead to an

increase in cash and decreasing it leads to decreases in cash. This is shown in Figure 44. All four

lines are generated when contract 3 is chosen and re-order quantity is 2500 jobs. Only re-order
Page | 71
level is changed. The blue line shows cash over time when re-order level is 100 jobs, which leads

to an ending cash level of $1.895 million (blue line). Decreasing the value to 60 jobs leads to a

decrease of cash level to $1.565 million (red line). Decreasing the value to 90 jobs leads to a

decrease of cash level to $1.894 million (pink line). Increasing the value to 110 jobs also

decreases cash level to $1,847,550 (green line). Therefore, 100 jobs is the optimum value for re-

order level.

Cash: 1 - 2 - 3 -

1: 2000000

2
1: 1500000 1
2
2
3
3 1

1: 1000000
1.00 67.75 134.50 201.25 268.00

Page 1 Days 4:20 PM Sun, Oct 09, 2011

Cash

Figure 44- Changing the Reorder Level

5.3.3.2 Optimum Reorder Quantity

Proving that 2,520 kits is the optimum reorder quantity is more challenging than proving

that 100 is the optimum reorder level. This is because the deviation in the graph when changing

the reorder level is less noticeable. Therefore, Figure 45 is zoomed into the last few days of the

game. All three lines are run when contract 3 is chosen and when the re-order level is set to 100

jobs. The blue line shows cash over time when re-order quantity is set to 2,500 kits; and it gives

Page | 72
the highest cash level. The red line and pink line are cash levels over time when the re-order

quantity is set to 3,000 kits and 2,000 kits respectively. Both of them have lower cash level by

around $50,000.

Cash: 1 - 2 - 3 -

1: 1901000

1: 1800500
3

1: 1700000
1.00 54.40 107.80 161.20 214.60 268.00

Page 1 Days 4:30 PM Sun, Oct 09, 2011

Cash

Figure 45 Choosing the optimum reorder quantity

5.3.3.3 Refining reorder level and reorder quantity for the last 50 days

The re-order level and re-order quantity chosen works well, but do not grant the

maximum cash level in the end. Specifically, buying 2,500 kits during last 50 days is over-

purchasing. So the value should be reduced for the last 50 days. The optimum values for the re-

order level and re-order quantity for the last 50 days are 40 jobs and 500 jobs respectively.

Figure 46 shows the graph for the refinement.

Page | 73
 Figure 46 Refining reorder level and reorder quantity for the last 50 days

Both lines are generated when contract 3 is chosen and re-order level and re-order

quantity is set to 100 jobs and 2,500 jobs for days 50 to 218. For the blue line, the values

remained constant throughout the run. For the red line, the values changed to 40 for re-order

level and 500 for re-order quantity respectively by day 218. The latter setting grants $1.932

million of ending cash, which is $36,165 higher than its alternative.

Page | 74
5.3.4 The Number of Machines

Besides contract type, lead time is a main determining factor for revenue per job. Figure

47 shows how the average lead time changes from day 50 to day 268.

1: smooth of average lead time

1: 40

1: 20

1 1
1: 0
50.00 104.50 159.00 213.50 268.00

Page 2 Days 5:12 PM Sun, Oct 09, 2011

Figure 47- Average Lead Time before changing desired number of machines

Figure 47 shows that between days 50 and 150 the lead time is very high. Therefore, the

revenue earned from completing orders is minimal. After day 150, the lead time decreases and

the revenue earned from orders increases.

In order to reduce lead time, new machines are purchased. As discussed in section 5.1.3,

the desired numbers of machines are 5-3-2 for stations 1, 2 and 3 respectively. Figure 48 shows

the average lead time for the base run (blue line) and the result of implementing the desired

number of machines (red line). Note that the lead time for the red line is less than for the blue

line throughout the test-run.

Page | 75
 smooth of average lead time: 1 - 2 -

1: 40

1: 20

2
1

2
1 1
2 2
1: 0
50.00 104.50 159.00 213.50 268.00

Page 2 Days 5:20 PM Sun, Oct 09, 2011

Figure 48 Lead time before and after changing the number of machines

The ending cash for the proposed strategy is $3.511 million, which is $1.579 million

higher than in the case of the base strategy. The cash levels are shown in Figure 49.

Cash: 1 - 2 -

1: 3600000

1: 2300000

2
1
1

1
1 2 2
2

1
1: 1000000
1.00 54.40 107.80 161.20 214.60 268.00

Page 1 Days 4:56 PM Sun, Oct 09, 2011

Cash

Figure 49 Cash Level before and after changing the desired number of machines

Page | 76
5.3.5 Verifying Optimal Strategy with the Random Input

In Littlefield Simulation, the number of incoming customer orders per day is random with

a mean of 12 and a standard deviation of 3. When inputted with the more realistic data, the

strategy proposed in previous section should grant the same result, since the system’s intrinsic

behavior should be independent of deviations in inputs. Figure 50 shows the cash level when the

number of incoming customer orders per day is random with a mean of 12 and a standard

deviation of 3.

Cash: 1 - 2 -

1: 4000000

1
1: 2500000

2
1

1 2 1 2
2
1
1: 1000000
1.00 54.40 107.80 161.20 214.60 268.00

Page 1 Days 5:35 PM Sun, Oct 09, 2011

Cash

Figure 50 Cash Level when customer orders are random

In Figure 50, the blue line shows the cash level over time when customer orders are

constant. The red line shows the cash level over time when the orders are random. The difference

between the ending cash for the modes is only $1,000, which is relatively small in comparison to

the ending cash level.

Page | 77
6.0 Limitations and Future Work

6.1 Limitations

A significant limitation of the winning strategy devised is that it could not be tested with

the Littlefield game due to time constraints. However, we are confident that the strategy would

perform well because it consistently gives optimal cash in the iThink model. The model’s

robustness has been tested as well.

The model we constructed for this report is a continuous time system dynamics model.

Using continuous time removes some of the distortions seen in the Littlefield discrete time

model. For most cases, the removal of such distortions would be desirable. However, using the

same policy devised on the continuous time system may not be as robust on the Littlefield

discrete time system. For the most part this can be overlooked since modeling a process in

continuous time corresponds better with reality.

One of the main shortcomings of the Littlefield Simulation is the lack of significant

holding cost for raw materials, which deviates from reality. Furthermore, completed orders do

not affect future orders in the game, since incoming customer orders are treated as an exogenous

variable. While in reality, if a company performs well in terms of the lead time, it will certainly

receive more orders as its relationship with customers strengthens and reputation grows.

6.2 Future Work

In future work, our optimal strategy should be tested in the Littlefield Simulation game.

The results can be used to amend the strategy and the model.

Future work should also include the development of a more dynamic policy where the re-

order level, re-order quantity and the number of machines to be bought changes automatically

Page | 78
when the parameters of the game change. The future model should also encompass holding costs

for raw materials in the inventory. This is also a consideration that Littlefield Technology can

take into account.

Lastly, our system dynamics model can be used to simulate an industry with many

competing firms.

Page | 79
7.0 Conclusion

The Littlefield game is a discrete time simulation. In order to design a winning strategy

for Littlefield game, a continuous System Dynamics model was constructed to represent the

discrete time simulation.

The initial model building phase of the project involved replicating the trends seen in the

original game. In order to model these initial trends we used the same decision rules that were

used in the Littlefield game, such as changing contracts based upon lead time. The decision rules

used in the original game were reactionary rather than proactive. Such decision rules do not help

in maximizing profit at the end of the game. Therefore, a more proactive strategy was needed to

maximize profit.

The main finding of the project is that to maximize the cash flow the most profitable

contract type should be chosen (contract 3). The player needs also to balance two strategic

investments: investments in raw materials and machines. The reorder level and reorder quantity

chosen was the economic reorder quantity and economic reorder point. The desired number of

machines selected based upon the number of machines which would give the maximum

utilization of machines.

Our ability to identify a winning strategy suggests that a system dynamics model can be used to

identify and test strategies for the Littlefield simulation. Our system dynamics model can also be

used in advanced system dynamics classes to teach elements of manufacturing.

Page | 80
8.0 Work Cited

Stanford Graduate School of Business website

http://www.gsb.stanford.edu/news/research/mfg_kumar_factorygame.shtml

Miyaoka, J. (2005), "Making Operations Management Fun: Littlefield Technologies," INFORMS

Transactions on Education, Vol. 5, No 2, http://ite.pubs.informs.org/Vol5No2/Miyaoka/

“Yale SOM Team Wins International Operations Competition” News of Yale school of

Management.

http://mba.yale.edu/news_events/CMS/Articles/5952.shtml retrieved 08.27.2011

Littlefield Simulation 2 Report by Ge Yin, Joshua Stephens, Michael Buchanan, Mingjing Chen

and Tian Bai.

http://www.scribd.com/doc/51139499/Littlefield-Simulation-2-Report Retrieved 08.27.2011

Michael J. Radzicki and Robert A. Taylor (2008). "Origin of System Dynamics: Jay W. Forrester

and the History of System Dynamics” In: U.S. Department of Energy's Introduction to System

Dynamics

http://www.systemdynamics.org/DL-IntroSysDyn/start.htm Retrieved 08.27.2011

Forrester, Jay (1998) Designing the Future. Page 5.

http://clexchange.org/ftp/documents/whyk12sd/Y_1999-03DesigningTheFuture.pdf Retrieved

08.27.2010

Gross, Donald, John F.. Shortle, and James M. Thompson. Fundamentals of queuing theory. e

d. ed. Hoboken, NJ: J. Wiley Sons, 2008. rint. Chapter 1, age

Page | 81
Krajewski, Lee J., Larry P. Ritzman, and Manoj K. Malhotra. Operations management:

Processes and supply chains.. 9. ed. Upper Saddle River: Pearson Education, 2010. Print. Page

437

"Causal Loop Diagram." Wikipedia, the Free Encyclopedia. Web. 13 Oct. 2011.

<http://en.wikipedia.org/wiki/Causal_loop_diagram>.

"Safety Stock Calculations | Lean Sigma Supply Chain." Resource Systems Consulting. Web. 23

Oct. 2011. <http://www.resourcesystemsconsulting.com/blog/archives/90>.

"Re-order Level Calculation." Managerial Accounting - Free Cost and Management Accounting

Articles. Web. 23 Oct. 2011.

<http://www.accountingformanagement.com/reordering_level.htm>.

Page | 82
9.0 Appendix

9.1 Equations

Cash(t) = Cash(t - dt) + (Revenue_per_day - Expenses) * dt

INIT Cash = 1000000
INFLOWS:
Revenue_per_day =
Revenue_from_completing_orders+Revenue_from_Interest+Total_Machine_Sales
OUTFLOWS:
Expenses = Expenses_from_Ordering__Raw_Material+Total_Machine_Expenses
Customer_Orders_in_the_system(t) = Customer_Orders_in_the_system(t - dt) +
(Job_release_into_the_system - Completed_orders) * dt
INIT Customer_Orders_in_the_system = 0
INFLOWS:
Job_release_into_the_system = Job_Release
OUTFLOWS:
Completed_orders = completion_rate
Customer_Order_Queue(t) = Customer_Order_Queue(t - dt) + (Customer_Orders_Incoming -
Customer_Orders_Processed) * dt
INIT Customer_Order_Queue = 1
INFLOWS:
Customer_Orders_Incoming = 26
OUTFLOWS:
Customer_Orders_Processed = Job_Release
machine1(t) = machine1(t - dt) + (Change_in_Machine_1) * dt
INIT machine1 = 3
INFLOWS:
Change_in_Machine_1 = IF (TIME <51) OR (TIME >236) THEN 0 ELSE IF
(Difference_in_Machines < 0) THEN 0 ELSE
Efffect_of__cash_ratio_on_machine1*Difference_in_Machines/Adjustment_Time

Page | 83
machine2(t) = machine2(t - dt) + (Change_in_Machine_2) * dt
INIT machine2 = 1
INFLOWS:
Change_in_Machine_2 = IF (TIME <51) OR (TIME >236) THEN 0 ELSE IF
(Difference_in_Machine_2< 0) THEN 0 ELSE
Effect_of_cash_ratio_on_machine_2*Difference_in_Machine_2/Adjustment_Time
machine3(t) = machine3(t - dt) + (Change_in_Machine_3) * dt
INIT machine3 = 1
INFLOWS:
Change_in_Machine_3 = IF (TIME <51) OR (TIME >236) THEN 0 ELSE IF
(Difference_in_Machine_3 < 0) THEN 0 ELSE
Effect_of_cash_ratio_on_machine_3*Difference_in_Machine_3/Adjustment_Time
Order_for_Raw_Material(t) = Order_for_Raw_Material(t - dt) + (raw_material_purchase_rate -
Raw_Material_Arriving_Rate) * dt
INIT Order_for_Raw_Material = 0
INFLOWS:
raw_material_purchase_rate =
raw_material_reorder_quantity*Inventory_Ratio's_effect_on_Reordering*Effect_of_cash__on_p
urchases_on_raw_materials
OUTFLOWS:
Raw_Material_Arriving_Rate = Order_for_Raw_Material/order_delay
queue_for_step2(t) = queue_for_step2(t - dt) + (step1_process_rate - step2_process_rate) * dt
INIT queue_for_step2 = 0
INFLOWS:
step1_process_rate = queue_for_step_1/Actual_Time_1
OUTFLOWS:
step2_process_rate = queue_for_step2/actual_time_2
queue_for_step3(t) = queue_for_step3(t - dt) + (step2_process_rate - step3_process_rate) * dt
INIT queue_for_step3 = 0
INFLOWS:
step2_process_rate = queue_for_step2/actual_time_2

Page | 84
OUTFLOWS:
step3_process_rate = queue_for_step3/actual_time_3
queue_for_step4(t) = queue_for_step4(t - dt) + (step3_process_rate - completion_rate) * dt
INIT queue_for_step4 = 0
INFLOWS:
step3_process_rate = queue_for_step3/actual_time_3
OUTFLOWS:
completion_rate = queue_for_step4/actual_time_4
queue_for_step_1(t) = queue_for_step_1(t - dt) + (Job_Release - step1_process_rate) * dt
INIT queue_for_step_1 = 0
INFLOWS:
Job_Release =
Customer_Order_Queue*Total_orders'_effect_on_Job_Release*Effect_of_raw_material_ratio_o
n_job_release/Job_Release__Lead_TIme
OUTFLOWS:
step1_process_rate = queue_for_step_1/Actual_Time_1
Raw_Material_Inventory(t) = Raw_Material_Inventory(t - dt) + (Raw_Material_Arriving_Rate -
Job_Release) * dt
INIT Raw_Material_Inventory = 160
INFLOWS:
Raw_Material_Arriving_Rate = Order_for_Raw_Material/order_delay
OUTFLOWS:
Job_Release =
Customer_Order_Queue*Total_orders'_effect_on_Job_Release*Effect_of_raw_material_ratio_o
n_job_release/Job_Release__Lead_TIme
Actual_Time_1 = Normal_Processing_Time_1*effect_of_capacity__utilization_1
actual_time_2 = effect_on_cap_ut_2*NPT_2
actual_time_3 = NPT_3*effect_on_cap_ut_3
actual_time_4 = npt_4*effect_on_cap_ut_4
Adjustment_Time = 1

Page | 85
Average_Lead_Time =
Job_Release__Lead_TIme+Actual_Time_1+actual_time_2+actual_time_3+actual_time_4
Average_Service3_Rate_3 = 15.91
average_service_rate1 = 12
Average_Service__Rate_2 = 30.6
Contract_Types_1 = Perceived_lead_time's_effect_on_Contract_Type1
Contract_Type_2 = Perceived_Lead_Time's__effect_on_Contract_Type2
Contract_Type_3 = Perceived_Lead_Time's_effect_on_Contract_Type3
cost_of_machine_1 = 90000
cost_of_machine_2 = 80000
cost_of_machine_3 = 100000
cost_per_job = 600
Desired_Machines = .33
Desired_Machines_2 = .33
desired_machines_3 = .33
Difference_in_Machines = Desired_Machines-machine1
Difference_in_Machine_2 = Desired_Machines_2-machine2
Difference_in_Machine_3 = desired_machines_3-machine3
effect_of_capacity__utilization_1 = 1+queue_for_step_1
effect_on_cap_ut_2 = 1+queue_for_step2
effect_on_cap_ut_3 = 1+queue_for_step3
effect_on_cap_ut_4 = 1+queue_for_step4
Expenses_from_Ordering__Raw_Material = IF (raw_material_reorder_quantity= 0) THEN 0
ELSE raw_material_purchase_rate*(fixed_cost/raw_material_reorder_quantity+cost_per_job)
fixed_cost = 1000
Initial_Reorder_Level = 24
Initial_Reorder_Quantity = 160
Interest_rate = 0.00026
Intermediate_Reorder_Level = 24
Page | 86
Intermediate_Reorder_Quantity = 160
Inventory_Ratio's_effect_on_Reordering = IF(Inventory__Ratio>1) THEN (0) ELSE (1)

Inventory__Ratio = IF (Reorder__Level=0) THEN 0 ELSE

(Raw_Material_Inventory+Order_for_Raw_Material)/Reorder__Level
Job_Processing_Time = 1/48
Job_Release__Lead_TIme = (Customer_Order_Queue+1) *Job_Processing_Time
machine1_capacity = NORMAL(average_service_rate1,std_dev1, 5)
Machine1_expense = IF(Change_in_Machine_1>0) THEN
SQRT((cost_of_machine_1*Change_in_Machine_1)^2) ELSE 0
machine2_capacity = NORMAL(Average_Service__Rate_2,std_dev2, 5)
machine3_capacity = NORMAL(Average_Service3_Rate_3,std_dev3, 5)
Machine_1_Sales = IF(Change_in_Machine_1<0) THEN
SQRT((Price_of_machines_sold*Change_in_Machine_1)^2) ELSE 0
Machine_2_Expense = IF(Change_in_Machine_2>0) THEN
SQRT((cost_of_machine_2*Change_in_Machine_2)^2) ELSE 0
Machine_2_Sales = IF(Change_in_Machine_2<0) THEN
SQRT((Price_of_machines_sold*Change_in_Machine_2)^2) ELSE 0
Machine_3_Expense = IF(Change_in_Machine_3>0) THEN
SQRT((cost_of_machine_3*Change_in_Machine_3)^2) ELSE 0
Machine_3_Sales = IF(Change_in_Machine_3<0) THEN
SQRT((Price_of_machines_sold*Change_in_Machine_3)^2) ELSE 0
maximum_orders__to_orders_ratio = Customer_Orders_in_the_system/Maximum_order_limit
Maximum_order_limit = 100
Normal_Processing_Time_1 = 1/(machine1*machine1_capacity)
NPT_2 = 1/(machine2*machine2_capacity*priority_for_step_2)
NPT_3 = 1/(machine3*machine3_capacity)
npt_4 = 1/(machine2*machine2_capacity*(1-priority_for_step_2))
order_delay = 3
Perceived_Lead_Time = SMTH3(average_lead_time,3.5,1.5)

Page | 87
Perceived_lead_time's_effect_on_Contract_Type1 = IF (TIME< 50) THEN 1 ELSE IF
(Perceived_Lead_Time>2) THEN 1 ELSE 0
Perceived_Lead_Time's_effect_on_Contract_Type3 = IF (TIME <50) THEN 0 ELSE
IF(Perceived_Lead_Time<=.6) THEN 1 ELSE 0
Perceived_Lead_Time's__effect_on_Contract_Type2 = IF (TIME <50) THEN 0 ELSE
IF(Perceived_Lead_Time>1) AND (Perceived_Lead_Time<=2) THEN 1 ELSE 0
Price_of_machines_sold = 10000
priority_for_step_2 = 0.5
ratio_of_machine1_cost_vs_cash = Cash/(cost_of_machine_1+Total_cost_of_order)
ratio_of_machine2_cost_vs_cash = Cash/(cost_of_machine_2+Total_cost_of_order)
ratio_of_machine3_cost_vs_cash = Cash/(cost_of_machine_3+Total_cost_of_order)
Ratio_of_purchase_vs_cash = IF( Total_cost_of_order =0) THEN 0 ELSE
Cash/Total_cost_of_order
raw_material_reorder_quantity = IF (TIME <50) THEN Initial_Reorder_Quantity ELSE
Intermediate_Reorder_Quantity
Raw_Material_to_Customer_Order_Ratio = IF (Customer_Order_Queue=0) THEN 0 ELSE
Raw_Material_Inventory/Customer_Order_Queue
Reorder__Level = IF (TIME <50) THEN Initial_Reorder_Level ELSE
Intermediate_Reorder_Level
Revenue_from_completing_orders =
completion_rate*Revenue_per_batch+Revenue_from_orders*completion_rate
Revenue_from_Interest = Cash*Interest_rate
Revenue_from_orders = IF (TIME < 50) THEN Contract_Type_1_revenue_per_batch_2 ELSE
Contract_Type_3__Revenue_per_Batch_policy
Revenue_per_batch =
Contract_Types_1*Contract_Type_1_revenue_per_batch+Contract_Type_2*Contract_Type_2_
Revenue_per_Batch+Contract_Type_3*Contract_Type_3__Revenue_per_Batch
smooth_of_average_lead_time = SMTH1(Average_Lead_Time,1)
std_dev1 = 4
std_dev2 = 10
std_dev3 = 5
Total_cost_of_order = fixed_cost+cost_per_job*raw_material_reorder_quantity
Total_Machine_Expenses = Machine1_expense+Machine_2_Expense+Machine_3_Expense
Page | 88
Total_Machine_Sales = Machine_1_Sales+Machine_2_Sales+Machine_3_Sales
Contract_Type_1_revenue_per_batch = GRAPH(average_lead_time)
(0.00, 750), (2.00, 750), (4.00, 0.00)
Contract_Type_1_revenue_per_batch_2 = GRAPH(smooth_of_average_lead_time)
(0.00, 750), (2.00, 750), (4.00, 0.00)
Contract_Type_2_Revenue_per_Batch = GRAPH(average_lead_time)
(0.00, 1000), (1.00, 1000), (2.00, 0.00)
Contract_Type_3__Revenue_per_Batch = GRAPH(average_lead_time)
(0.00, 1250), (0.5, 1250), (1.00, 0.00)
Contract_Type_3__Revenue_per_Batch_policy = GRAPH(smooth_of_average_lead_time)
(0.00, 1250), (0.5, 1250), (1.00, 0.00)
Effect_of_cash_ratio_on_machine_2 = GRAPH(ratio_of_machine2_cost_vs_cash)
(0.00, 0.00), (0.2, 0.00), (0.4, 0.00), (0.6, 0.00), (0.8, 0.00), (1.00, 1.00), (1.20, 1.00), (1.40,
1.00), (1.60, 1.00), (1.80, 1.00), (2.00, 1.00)
Effect_of_cash_ratio_on_machine_3 = GRAPH(ratio_of_machine3_cost_vs_cash)
(0.00, 0.00), (0.2, 0.00), (0.4, 0.00), (0.6, 0.00), (0.8, 0.00), (1.00, 1.00), (1.20, 1.00), (1.40,
1.00), (1.60, 1.00), (1.80, 1.00), (2.00, 1.00)
Effect_of_cash__on_purchases_on_raw_materials = GRAPH(Ratio_of_purchase_vs_cash)
(0.00, 0.00), (0.2, 0.2), (0.4, 0.4), (0.6, 0.6), (0.8, 0.8), (1.00, 1.00), (1.20, 1.00), (1.40, 1.00),
(1.60, 1.00), (1.80, 1.00), (2.00, 1.00)
Effect_of_raw_material_ratio_on_job_release =
GRAPH(Raw_Material_to_Customer_Order_Ratio)
(0.00, 0.00), (0.2, 0.2), (0.4, 0.4), (0.6, 0.6), (0.8, 0.8), (1.00, 1.00), (1.20, 1.00), (1.40, 1.00),
(1.60, 1.00), (1.80, 1.00), (2.00, 1.00)
Efffect_of__cash_ratio_on_machine1 = GRAPH(ratio_of_machine1_cost_vs_cash)
(0.00, 0.00), (0.2, 0.00), (0.4, 0.00), (0.6, 0.00), (0.8, 0.00), (1.00, 1.00), (1.20, 1.00), (1.40,
1.00), (1.60, 1.00), (1.80, 1.00), (2.00, 1.00)
Total_orders'_effect_on_Job_Release = GRAPH(maximum_orders__to_orders_ratio)
(0.00, 1.00), (0.1, 1.00), (0.2, 1.00), (0.3, 1.00), (0.4, 1.00), (0.5, 1.00), (0.6, 1.00), (0.7, 1.00),
(0.8, 1.00), (0.9, 1.00), (1, 1.00), (1.10, 0.00), (1.20, 0.00), (1.30, 0.00), (1.40, 0.00), (1.50, 0.00),
(1.60, 0.00), (1.70, 0.00), (1.80, 0.00), (1.90, 0.00), (2.00, 0.00)

Page | 89
9.2 User Interface

Page | 90
9.3 Littlefield Game Parameters

Page | 91
Page | 92