An analysis of the effect of supply chain and manufacturing parameters on inventory cost reduction for push type manufacturing systems
iii TABLE OF CONTENTS PAGE LIST OF TABLES vii LIST OF FIGURES ix CHAPTER I. INTRODUCTION 1 Statement of the Problem 3 Statement of the Purpose 3 Importance of the Research 3 Research Questions 4 Assumptions 6 Limitations 6 Definition of Terms 7 CHAPTER II. LITERATURE REVIEW 8 Inventory Control 9 The Economic Order Quantity (EOQ) Model 9 Dynamic Lot Sizing 12 The News Vendor Model 13 The Base Stock Model 16 The (Q,r) Model 19 Supply Chain Management 22 Supply Chain Management (SCM) Practices 24 Performance Measurement of Supply Chains 26
iv PAGE Discrete-event System Simulation 28 When to Use Simulation? 29 Advantages and Disadvantages of Simulation 31 System in Simulation 32 Discrete and Continuous Systems 33 Steps in a Simulation Study 35 Inventory Control and Supply Chain Simulation 40 CHAPTER III. RESEARCH DESIGN AND METHODOLOGY 43 Research Design 43 Initial Information 43 Field Study 45 Regression Analysis 46 Independent Variables 47 Dependent Variable 50 Simulation Study 51 The Simulation Model 51 Independent Variables 51 Dependent Variables 54 Model Parameters 55 CHAPTER IV. RESULTS AND DISCUSSION 58 Initial Research 58
V PAGE Simulation 61 Preliminary Analysis 61 Simulation Model 68 Validation and Verification 71 Design of Experiments 72 Statistical Analysis of the Simulation 74 Regression Analysis 82 CHAPTER V. SUMMARY, CONCLUSIONS AND RECOMMENDATIONS 85 Summary 85 Conclusion 86 Recommendations 88 REFERENCES 91 APPENDIX A: STATISTICAL PROGRAM AND OUTPUTS FOR REGRESSION ANALYSIS (INDEPENDENT VARIABLES: QUALITY AND DELIVERY NONCONFORMANCES DEPENDENT VARIABLE: INVENTORY ON HAND) 94 APPENDDC B: STATISTICAL PROGRAM AND OUTPUTS FOR REGRESSION ANALYSIS (INDEPENDENT VARIABLES: % OF SCHEDULE CHANGES AND ESTIMATED DELAY DEPENDENT VARIABLE: % OF QUALITY NONCONFORMANCES) 99 APPENDDC C: STATISTICAL PROGRAM AND OUTPUTS FOR REGRESSION ANALYSIS (INDEPENDENT VARIABLES: % OF SCHEDULE CHANGES AND ESTIMATED DELAY DEPENDENT VARIABLE: % OF DELIVERY NONCONFORMANCES) 105 APPENDDC D: STATISTICAL PROGRAM AND OUTPUTS FOR 3-WAY ANOVA USING THE SIMULATION DATA 112
vi PAGE APPENDIX E: STATISTICAL PROGRAM AND OUTPUTS FOR 2-WAY ANOVA USING THE SIMULATION DATA 120 APPENDIX F: STATISTICAL PROGRAM AND OUTPUTS FOR REGRESSION MODEL USING THE SIMULATION DATA 128
vii LIST OF TABLES TABLE PAGE 1 Supplier classification guideline 54 2 Values of reference state input parameters 69 3 Values of reference state output indicators 70 4 Validation results for model behavior 71 5 The Levels of the Parameters 73 6 Summary of Tukey's pairwise comparison test 77 7 Summary of Tukey's pairwise comparison test 79 8 Summary of Tukey's pairwise comparison test 81 Al SAS Regression output for initial analysis 96 Bl SAS Regression output 101 CI % Schedule changes, estimated delays, and % delivery nonconformances for 33 companies 107 C2 SAS Regression output 108 Dl SAS 3-WAY ANOVA table for simulation data 114 El SAS 2-WAY ANOVA table for simulation data 122 E2 Tukey's studentized range test for average inventory on hand 123 E3 SAS 2-WAY ANOVA table for simulation data 124 E4 SAS 2-WAY ANOVA table for simulation data 125 E5 SAS 2-WAY ANOVA table for simulation data 126 E6 SAS 2-WAY ANOVA table for simulation data 127
Vl l l TABLE PAGE Fl SAS Regression output 130 F2 Forward selection method summary for the suggested regression model 137 F3 Backward selection method summary for the suggested regression model. .137 F4 Stepwise selection method summary for the suggested regression model.... 138 F5 SAS Regression output with parameter estimates 139 F6 R Square and MSE values for selected variables 140
ix LIST OF FIGURES FIGURE PAGE 1 EOQ Inventory model 11 2 Flow diagram of steps in a simulation study 37 3 Data and material flow 45 4 Sample MCT analysis 48 5 MCT mapping tool 49 6 MCT map 50 7 Simulation model 52 8 Plot for average inventory on hand vs. % of quality nonconformance 60 9 Plot for average inventory on hand vs. % of delivery nonconformance 60 10 Representation of the simplified simulation model 62 11 Plot of quality nonconformance percentages versus the percentage of schedule changes of each supplier 66 12 Plot of quality nonconformance percentage versus the estimated delays for each supplier 66 13 Plot of delivery nonconformance percentage versus the percentage of schedule changes of each supplier 67 14 Plot of delivery nonconformance percentage versus the estimated delay for each supplier 67 15 Average days on hand values for model validation 72 Al Residuals vs Predicted value plot for the regression model 97 A2 Normal probability plot of residuals for the regression model 98 B1 Residuals vs Predicted value plot for the regression model 102
X PAGE B2 Normal probability plot of residuals for the regression model 103 B3 Cook's distance plot showing the most influential data points 104 CI Residuals vs Predicted value plot for the regression model 109 C2 Normal probability plot of residuals for the regression model 110 C3 Cook's distance plot showing the most influential data points I l l Dl Normal probability plot of residuals for the 3-WAY ANOVA model 116 D2 Residuals vs. predicted values for 3-WAY ANOVA model 117 D3 Main effects plot for average inventory on hand 118 D4 Interaction effects plot for average inventory on hand 119 Fl Plot for average inventory on hand vs. QPPM 131 F2 Plot for average inventory on hand vs. DPPM 132 F3 Plot for average inventory on hand vs. number of suppliers 133 F4 Resuduals vs the predicted values for the regression model 134 F5 Normal probability plot of the residuals 135 F6 Cook's distance plot for most influential data points 136
AN ANALYSIS OF THE EFFECT OF SUPPLY CHAIN AND MANUFACTURING PARAMETERS ON INVENTORY COST REDUCTION FOR PUSH TYPE MANUFACTURING SYSTEMS An Abstract of a Dissertation Submitted In Partial Fulfillment of the Requirements for the Degree Doctor of Industrial Technology Approved: Dr. Mohammed Fahmy, Committee Chair Dr. Sue A Joseph Interim Dean of the Graduate College Mehmet Emre Bahadir University of Northern Iowa July, 2009
ABSTRACT In the global network of businesses, supply chain and order fulfillment managements are the most critical functional departments to determine the winner of the global competition. In this research a network of companies that are flowing information, product and services between providers and a receiver is investigated in order to gain a better insight of the current situation. Analyses, explanations and solutions were developed through responding to the following research questions: 1. What are the most important variables that affect the quality and delivery performances of a supply chain? 2. What are the most important variables that affect the service rate or fill rate of a supply chain of a manufacturing company? 3. What levels of the selected variables could be used in order to minimize inventory on hand? The research was based on the analysis of a supplier network of a midwestern manufacturing company. Initial study verified that there was no company policy established to prevent stock-outs resulting from late deliveries or quality nonconforming parts. In order to investigate the effects of existing company policies and guidelines a discrete event simulation model was developed. During the model building phase historic data was utilized to create simulation parameters. Analysis of the historic data revealed
that neither the production lead time nor the schedule changes affect the quality or delivery performance of suppliers. The results of the simulation confirm the importance of the number of suppliers in a supply chain. The number of suppliers negatively affects the efficiency of the order fulfillment process and high numbers of suppliers require higher inventory levels. The company's supplier classification guideline was also validated for delivery performance ratings by the simulation model. However, the supplier classification based on the quality performance was not found to be practically significant.
1 CHAPTER I INTRODUCTION Inventory control is the activity which organizes the availability of items to the customers. It coordinates the purchasing, manufacturing and distribution functions to meet marketing needs. This role includes the supply of current sales items, new products, consumables, and all other supplies. Inventory enables a company to support its customer service, logistic or manufacturing activities in situations where purchase or manufacture of the items is not able to satisfy customer demand. The aim of the inventory control is not to make all items available at all times as this may be detrimental to the finances of the company. Wild (1997) defines the normal function for stock control as meeting the required demand at a minimum cost possible. The aim of long term profitability of an organization has to be translated into operational and financial targets which can be applied to daily operations of the organization. On the other hand, the purpose of the inventory control function is to support business activities to optimize three main functions: inventory cost, customer service, and operating costs. Inventory levels in a company are driven by the company's sales and marketing strategy for its product lines, an understanding of customer buying patterns, and the competitive and economic environment. These factors are all external to the inventory management department in a company. How they are translated into inventory levels and availability is the function of the inventory strategy as translated into internal planning and control processes and procedures.
2 The purpose of this research is to investigate internal and external factors and relevant parameters that affect inventory level, service rate and cost variations in final assembly lines. In order to do that, a model that captures all the cited parameters of interest is proposed by the researcher. Later, this model is tested by a discrete event simulation technique using ARENA ® simulation software. At the end of the study, the results will be analyzed for their sensitivity to explain the variations under real life conditions. The vision behind the current production strategies in many production settings is to have the target inventory, at the target time, at the target place, in the target quality, in the target orientation with zero deviation from target. However, from an absolutely practical perspective, zero-deviation performance for all parts across all dimensions all the time is impossible to achieve. This vision is different than an "all inventory is waste" vision, which is supported by Just-in-Time (JIT) and Toyota Production Methods; Bernard (1999) suggests an integral strategy that is based on the recognition that a given level of inventory is necessary to the effective operation of the business. This level is a function of business conditions which existed at the time the inventory was ordered and which are forecasted to exist through the duration of the stocking horizon. Ensuring that the target level of inventory is available to support the needs of the business is the mission of inventory management.
3 Statement of the Problem The problem of this research was to develop a simulation model to analyze the effects of lead time, order schedule changes, number of suppliers, and delivery and quality related problems on safety stock levels in order to minimize inventory amount and reduce cost. Statement of the Purpose Like all other activities in a manufacturing company, inventory management has to contribute to the welfare of the whole organization. Therefore, the expected results of this research will allow organizations to align their suppliers and their suppliers' resources and capabilities, thereby create a competitive advantage and provide value to their customers. In order to do that, the goal of this research is to identify key inventory control parameters, and develop a mathematical model based on the factors that are being employed at the company under study. Importance of the Research Inventory cost reduction should be one of the prime goals of all manufacturing companies. According to Kobert (1992) because inventory is a huge asset on the balance sheet accounting for as much as 50% of current assets, inventory management plays a major role in a company's cost reduction strategy. It is also noted that a better control over inventory level results in improvements in such areas as purchasing, warehousing, distribution, labor utilization, equipment scheduling, data presentation, quality assurance, vendor relations, packaging, materials handling, and even personnel administration.
4 The need for this research first came out at a meeting with the Order Fulfillment Management of a Midwestern Manufacturing Company. Currently, the company establishes operating parameters using rules of thumb and experiential knowledge. This leads to inconsistencies and variations from planner to planner and factory to factory. It is believed that current practices are not leading to optimum business results. The company is on the journey to continuously improve operations execution and asset velocity. However, the company doesn't fully understand the mathematical relationship between operations execution parameters and the business outcome metrics. It is the administration's desire to discover and understand the relationships so that they may systematically establish the operating execution system parameters, to proactively drive future business results. More specifically, the company under study has asset reduction targets which will drive financial advantage to the company. However, there are no guiding principles or formulas for setting up optimal inventory levels. The company is doing business with more than six thousands suppliers from all over the world. Correlating optimal inventory levels to supplier lead times and supplier performances as well as factory execution performance will help the suppliers and order fulfillment activities get aligned in order to achieve asset reduction objectives. Research Questions Modeling and formulating an efficient inventory planning and control policy to guarantee the product availability at a certain level with the lowest cost is not an easy task. There are many uncertainties inherent to the process itself, such as delivery or replenishment lead time, inaccurate demand forecasting, and variations between delivery
5 and order quantities. These variations and uncertainties require the building up of safety stock. Although overstocking involves more inventory holding costs than necessary, being short of safety stocks may cause sales losses and higher rate of postponed orders than desirable, which at the end results in the deterioration of service levels and customer service standards. The current research addresses the following questions. The findings will be addressed in Chapter IV. 1. What are the most important variables that affect the quality and delivery performances of a supply chain? 2. What are the most important variables that affect the service rate or fill rate of a supply chain of a manufacturing company? 3. What levels of the selected variables could be used in order to minimize the inventory on hand? The research questions were evaluated in an experimental design that analyzes the effects of parameters at different levels. Also multiple regression analysis and analysis of variance methods were employed along with the design of experiments method.
6 Assumptions The following assumptions were made in pursuit of this study: 1. That the methods and the efficiencies of manufacturing, logistics, and supply management operations stayed the same during the data collection period at the suppliers' manufacturing facilities. 2. The data collected from the suppliers and from the company under study are considered to be valid and representative for simulation and statistical analysis purposes. 3. That the supply chain network and the inventory control operations can be simulated using ARENA® discrete-event simulation software. 4. That the parameters under consideration are measurable. Limitations This research study was conducted in view of the following limitations: 1. The simulation model will be developed in ARENA ® discrete event simulation program. The limitations of the program determine the model accuracy. 2. The detail and the representation quality of the simulation model depend on the needs and the system knowledge of the order fulfillment management team.
7 Definition of Terms To provide a clearer understanding of the terms used in this study, the following definitions are provided. 1. Discrete-event Simulation: "A discrete-event simulation is one in which the state of a model changes at only a discrete, but possibly random, set of simulated time points." (Schriber & Brunner, 1997) 2. Model: "A model is defined as a representation of a system for the purpose of studying the system. A model is not only a substitute for a system, it is a simplification of a system." (Mihram & Mihram, 1974) 3. Supply Chain: "A supply chain is a group of organizations (including product design, procurement, manufacturing, and distribution) that are working together to profitably provide the right product or service to the right customer at the right time" (Geunes & Pardalos, 2005) 4. Supply Chain Management: "All the management tasks necessary to obtain, move, transport, process, and deliver goods from vendors, through manufacturing, to the final customer." (Schniederjans & Cao, 2002) 5. Electronic Data Interchange (EDI): "A technology for electronic business that allows the computer to computer exchange between the organizations of standard transaction documents. EDI systems lower transaction costs because they automate transactions between information systems through a network. EDI systems can reduce the inventory costs by minimizing the amount of time that components are in the inventory." (Laudon & Laudon, 2004)
8 CHAPTER II LITERATURE REVIEW It is the goal of all manufacturing industries to produce high-quality products in the most economical and timely manner. In his study Altiok (1996) pointed out three parameters; quality, economics, and time as being the most important indicators of the customer-satisfaction. Thus, these parameters can also measure the manufacturing performance of a company. Companies invest into the information technologies such as computers, communication networks, sensors, actuators, and other equipment that give them an abundance of information about their materials and resources. In today's global competition, a manufacturing company's survival is becoming more dependent on how best this influx of information is utilized. Consequently, there evolves a great need for sophisticated tools of performance analysis that use this information to help decision makers in choosing the right course of action. These tools will have the capability of data analysis, modeling, computer simulation, and optimization for use in designing products and processes. According to Meyers and Stewart (2001), Frederic Taylor's "Scientific Management," which is a management approach for improving labor productivity, made the modern discipline of operations management possible. Not only did scientific management establish management as a discipline worthy of study, but also it placed a premium on quantitative precision that made mathematics a management tool for the first time. Meyers and Stewart (2001) claim that Taylor's primitive work formulas were the precursors to a host of mathematical models designed to assist decision making at all
9 levels of plant design and control. Later, these models became standard subjects in business and engineering curricula. Entire academic research disciplines sprang up around various operations management problem areas, including inventory control, scheduling, capacity planning, forecasting, quality control, and equipment maintenance. In this chapter the history of the mathematical modeling approach to inventory control, supply chain management, discrete-event system simulation, and simulation of inventory control and supply chains are reviewed. Inventory Control The Economic Order Quantity (EOQ) Model One of the earliest applications of mathematics to factory management was the work of Ford W. Harris (1913). In his pioneering study, Harris characterized the problem in a factory setting and dealt with the issue of setting manufacturing lot sizes. According to his problem design, he researched a factory producing various products. Depending on the orders, the production was switching between these products. However, these production changes were requiring costly setup changes. As an example, he described a metalworking shop that produced copper connectors. Each time the production changed from one type of connector to requiring another, the production and machines had to be stopped and adjusted for a different setup, clerical work to be done, and material might be wasted. Harris defined the sum of the labor and material cost to ready the shop to produce a product to be the setup cost.
10 Harris (1913) was consistent with the scientific management emphasis of his day on precise mathematical approaches to factory management. To derive a lot size formula, he made the following assumptions about the manufacturing system: 1. Production is immediate. There is no limit on the production capacity; the total number of orders can be produced instantly. 2. Delivery is instantaneous. There is no time interval between the production, shipment and delivery of the orders. 3. Certain demands. Time and the size of the order are known with certainty. 4. Constant demand size over time. If the minimum time interval is one day, the total yearly demand can be divided by the number of work days so that the daily demand can be calculated. 5. Setup cost is fixed. The size of the order or lot doesn't affect the setup cost. 6. Products can be analyzed individually. Either there is only a single product or there are no interactions between products. With these assumptions, the optimal production lot sizes can be computed for EOQ model. The notation will be as follows: D = annual demand c = cost of producing one unit in dollars without setup and inventory costs added A = setup related cost for the production of one lot in dollars h = the dollar cost of holding one unit per year. If the interest rate is the only factor considered in the calculation of holding cost, and if the interest = /, then h = ic. Q = the number of units in one lot; this is the variable we're trying to optimize
11 Harris (1913) treated time and product as continuous variables; this assumption was required for the modeling purposes. Because the demand size and time are known and fixed, we can order Q units of products when the inventory level drops down to zero. The result of this assumption is represented graphically in Figure 1. Quantity on hand Q Time Figure 1. EOQ Inventory model For every setup the cost is A, and the number of orders is D/Q per year. Thus, the setup cost per year is AD/Q. Since this cost of producing one unit is c, then for one year production, the production cost is cD. Thus, the total cost, which includes inventory, setup and production costs per year can be calculated as } 2 Q So, for the cost function above, the lot size that minimizes the Y(Q) can be expressed as 0 = 2AD
12 The formula above is the most basic form of economic order quantity (EOQ). This formula is also known as economic lot size. From this formula we can conclude that the optimal order quantity varies in direct relationship to the square root of the setup cost and the demand. However, optimal order quantity decreases with the square root of the holding cost. According to Harris, the most important implication of his study is that there is a tradeoff between lot size and inventory. In summary, when the lot size is increased, the average amount of inventory also increases; on the other hand the frequency of ordering is reduced. By inserting setup cost into the formula, Harris was able to panelize frequent orders and prove this relationship in economic terms. Dynamic Lot Sizing Although the EOQ model successfully proves the existence of a relationship between setup cost, holding cost and optimal order quantity, it is not precise enough to apply to real life situations. One of the main concerns about the EOQ model is in the unrealistic assumptions it makes. Among these unrealistic assumptions is that the constant demand assumption is relaxed by the Wagner-Whitin model (Wagner & Whitin, 1958). The Wagner-Whitin model was established on the same problem of determining production lot sizes. The model accepts all the EOQ assumptions as valid except the constant demand. Demand is considered to be varying overtime in the Wagner-Whitin model. The dynamic lot sizing model has the most important effect on the modern production control which is the origin of the materials requirement planning (MRP).
13 The dynamic lot sizing approach also has implications on the modeling of time. Because the demand occurs at specific times, the time must be divided into discrete periods like hours, days, weeks, or months. The length of the periods depends on the characteristics of the system. If the system has a very high volume production or if the demand is changing rapidly, short periods like days might be more appropriate. On the other hand if the production volume is low or the demand is changing slowly a larger time period such as monthly schedule might serve better. The News Vendor Model One of the earliest applications of statistical modeling in inventory control and production planning dates back to Wilson's work (1934). In order to analyze the problem, Wilson (1934) broke it into two parts: 1. The first part of the problem is to determine the order quantity, in other words, the quantity that will be purchased or produced for each order. 2. The second part consists of the determination of the reorder point. This is the level of inventory on hand at which the replenishment must be triggered. The news vendor model considers a single replenishment situation. Thus, the only problem is to find the appropriate quantity while the demand is uncertain. The model's name comes from the resemblance to the problem of a person who purchases newspapers in the morning without any prior information on demand. She sells a random amount of newspapers and discards the leftovers. In this situation, in order to find the appropriate production levels, two pieces of information are required. The first piece of information is the anticipated demand and the
14 second piece is the cost of producing more or less than the required amount. For this model Wilson's (1934) suggested assumptions can be summarized as follows: 1. Products are separable. Products can be considered one at a time since there are no interactions. 2. Planning is done for a single period. Future periods can be neglected since the effect of the current decision on them is negligible. 3. Demand is random. Demand can be characterized with a known probability distribution. 4. Deliveries are made in advance of demand. All stock ordered or produced is available to meet demand. 5. Costs of overage or underage are linear. The charge for having too much or too little inventory is proportional to the amount of the overage or underage. In order to develop the statistical model, the following notion is used with the assumptions above: X= demand (in units), a random variable G(x) = P(X 15 Q = Decision variable, which is the number of units to produce Using the notation above the expected cost function can be defined as follows: Y(Q) =c0^(Q- x)g(x)dx + cu £(x - Q)g(x)dx The value of Q to minimize expected overage plus underage cost is obtained by differentiating Y(Q). G(Q*) = C" Co + Cu If the demand is assumed to be normal, the above expression can be expressed as: G(Q*) = > Cu Co + Cu o- J where
15 Q = Decision variable, which is the number of units to produce Using the notation above the expected cost function can be defined as follows: Y(Q) =c0^(Q- x)g(x)dx + cu £(x - Q)g(x)dx The value of Q to minimize expected overage plus underage cost is obtained by differentiating Y(Q). G(Q*) = C" Co + Cu If the demand is assumed to be normal, the above expression can be expressed as: G(Q*) = > Cu Co + Cu o- J where