Eindhoven University of Technology
MASTER
Spare part management improvement with an application to a high-tech low-volume company
Matkovic, M.
Award date:
2019
Link to publication
Disclaimer
This document contains a student thesis (bachelor's or master's), as authored by a student at Eindhoven University of Technology. Student theses are made available in the TU/e repository upon obtaining the required degree. The grade received is not published on the document as presented in the repository. The required complexity or quality of research of student theses may vary by program, and the required minimum study period may vary in duration.
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.
Eindhoven, August 2019
SPARE PART MANAGEMENT IMPROVEMENT WITH AN APPLICATION TO A HIGH-TECH LOW-VOLUME COMPANY
Maja Matkovic
MSc Management Organisation and Business Economics Student identity number 1028781
In partial fulfilment of the requirements for the degree of Master of Science
in Operations Management and Logistics
Supervisors:
Company X :
O. Jorissen, Director Global Service Logistics Eindhoven University of Technology:
ir. dr. S.D.P. Flapper, Eindhoven University of Technology, OPAC
dr. S. Dabadghao, Eindhoven University of Technology, OPAC
TUE. School of Industrial Engineering
Series Master Thesis Operations Management and Logistics
Keywords: Spare Parts, Inventory Control, Capital Goods, Multi-echelon, Multi-item, Repair,
Linear Programming, Greedy Algorithm.
Abstract
This study investigates possibilities on how the supplier of a capital good can improve its spare
part management. It consists of two parts: analysing the reasons of unused and used for testing
returns and developing an improved inventory control model for the spare parts. The improved
inventory control model is the core of the research project. We consider a three-echelon spare
part inventory system consisting of one central warehouse and multiple warehouses at the second
and third echelon level with constant lead times for the replenishment of all warehouses and
fill rate as service measure. The central warehouse operates according to an (Q,R)-policy and
the other warehouses according to a continuous base-stock policy. The warehouses keep both
repairable and non-repairable parts in stock to serve their customers. Warehouses in the second
echelon level are resupplied by the central warehouse and every warehouse in the third echelon
level is resupplied by one warehouse in the second echelon level. All warehouses are subject
to Poisson demand. If a part is not in stock, it is backordered. Exact evaluation is used to
calculate the backorder and inventory levels. Near-optimal reorder points and base stock levels
are found through a greedy heuristic based on marginal analysis. The results show significant
improvement potentials if the proposed inventory control model is implemented.
All numbers and costs presented in this thesis are
fictional due to confidentiality reasons
Preface
This document contains the master thesis report, which is the final part of obtaining the Master of Science in Operations Management & Logistics at Eindhoven University of Technology.
The project has been executed within the Service Logistics department of a high-tech low- volume company. Working on this project proved to be very challenging for me, but I am very grateful for the opportunity, as I have gained a great amount of knowledge and experience.
This project could not have happened without the support of many people, both on personal and academic level, that I am very thankful for.
First and foremost I would like to thank S.D. Flapper, my first supervisor, for guiding me through the process, giving me constructive feedback and most of all for keeping me motivated.
Second I would like to thank S. Dabadghao for giving me feedback on the report.
I would like to thank my supervisor in the company, Olaf, for giving me the opportunity to execute the project within the company, referring me to the right people for my questions and helping and supporting me throughout the project. I would also like to thank Jaco, Mark and Rob for answering my endless questions. Many thanks to all the other colleagues for the great working environment and making my time in the company enjoyable.
I would like to express gratitude for my fellow students, Matthew and Yorick for the many extra lessons outside of lectures they have thought me on programming and mathematical models.
Your lessons were of great value during my thesis.
Last but not least, many thanks to my family, partner and friends for their never-ending support
during this project.
Management Summary
Company X is one of the suppliers of electron microscopes. This master thesis is executed within the Service Logisics department of company X, the department that is responsible for the spare part inventory management. Because the company wants to achieve a high service level, maintenance on the microscopes is performed by the repair-by-replacement method: rather than repairing the parts at site, the field service engineer replaces the old part with a new part and sends the old part back to the warehouse for repair.
The service logistics department is facing several problems with regards to the parts used for maintenance:
A relatively large percentage of spare parts is returning unused or used for test.
The return lead times for the parts from the customer to the local warehouse is highly variable and exceeds the desired 14 days.
Reasonably, this comes at extra (unnecessary) costs. Currently the company has no compre- hensive insight into the effect, nor the reasons for this.
The microscopes are divided in two groups: legacy tools and XYZ tools. The XYZ tools have been recently added to the portfolio and the legacy tools are the original microscopes. Up to now, the company has been able to achieve high fill rates for their legacy tools. This however does not seem to be the case for the XYZ parts. Reason for this is that the software tool that is currently used suggests too low base stock levels and reorder points. The low base stock levels and reorder points can be attributed to the low demand rates for the XYZ parts and the small installed base compared to the other tools.
Based on the problem statement, the following assignment is defined:
Analyse the current inventory model, thereby quantifying the effects of the unused returns, used for test returns and the long and variable return lead times.
Identify ways to improve the spare part supply chain and quantify the improvement po- tentials.
The analysis of the unused and used for testing returns is addressed by performing data-analysis and conducting interviews with several stakeholders. Analysing reasons for parts being returned unused or used after testing based on data turned out to be a complex task: no clear patterns were detected and the possible interaction between different factors make that endless combi- nations exist for reasons of return.
One important finding was that Japan accounts for relatively high percentage of unused and
used for testing returns and EU for high percentage of used for testing returns. The returns
for the other countries were consistent with the quantity ordered in that period. In addition,
many parts that are returned unused or used after testing are returned unused or used after
testing throughout the years, indicating that the returns for these specific parts are not random
incidents.
An interview with a field service engineer provided some important insights. For some parts it seems strange that the parts are returned, because diagnosis was easy or because the parts could be stocked at the customers site. For other parts correct diagnosis is difficult, because of the complexity of the microscopes. Looking at the number of times parts are returned per engineer, it seems that some engineers return parts unused/used for testing more often than others.
Additionally an effort was made to provide cost estimates of these returns. The total costs estimated to be incurred by the unused and used for testing returns in the period 01/01/2018- 31/10/2018 are $3,542,667.45. 70% of these cost are due to the repair and scrap costs of the used for testing returns.
The quantification of the effects of long return lead times is done in combination with the improved inventory control model for XYZ parts and hence is part of the second part of the thesis.
The second part of the thesis consisted of developing an improved inventory control model for XYZ parts. To this end, a model was developed that is suitable for a multi-echelon structure with constant lead times, Poisson demand, backorders and aggregate fill rate as service measure. The developed model is a multi-echelon model, in which the central warehouse operates according to a (Q,R)-policy and the local and regional warehouses to a base-stock policy. Based on exact evaluation we were able to derive backorder and inventory levels. Near-optimal base stock levels and reorder points were found with a greedy heuristic based on marginal analysis.
The model was applied to 494 XYZ parts and three warehouses, North America, Asia-Pacific Region and Europe. North America is seen as central warehouse and Asia-Pacific Region and Europe as warehouses in the second-echelon level (called inbetween hubs). Tables 1 and 2 show the results of implementing the proposed control model. Table 1 displays the fill rates and inventory values (expected inventory on hand expressed in purchase price( $)) if current stock levels are used as input for the model and table 2 displays the inventory values and fill rates if the proposed control model is implemented. Improvement potentials are significant, especially for the North American warehouse. Total inventory values decrease from $18,905,920.68 to
$8,365,030.10 and fill rates increase to 90%.
Warehouse Fill Rate Inventory Value( $)
Expected Inventory on Hand (units)
APR 83% 2,391,955.04 1034
NA 81% 12,256,857.92 3311
EU 78% 4,294,169.72 670
Grand Total 18,905,920.68 5015
Table 1: Performance current control model
Warehouse Fill Rate Inventory Value( $)
Expected Inventory on Hand (units)
APR 91% 3,215,325.10 346
NA 90% 2,918,997.56 1165
EU 91% 2,267,769.44 309
Grand Total 8,365,030.10 1820
Table 2: Performance proposed control model
Finally, the effect of the long return lead times on the inventory is analysed. Sensitivity analysis revealed that return lead times significantly impact inventory levels. The first important result is that if base stock levels and reorder points are kept as is and the actual return lead times differ from the return lead time as used in the model, the expected inventories will be lower than what the expected inventories should be. In the current practice the base stock levels are not constantly recalculated based on (increased) return lead times. This means that in practice, a backorder would occur that forces the company to procure a new part. According to figure 1 these costs add up to $30,000.00 per day that the parts are returned too late (figure 1).
Figure 1: Return leadtime increase, while keeping other variables constant
Moreover, effects of longer return lead times on the total fill rate were negligible. An increase of 12 days in return lead time for all repairable parts leads to an aggregate fill rate of 89% for the central warehouse and no changes for the inbetween hubs. This is due to the low percentage of repairable parts compared to the non-repairable parts. Effects on individual parts however are significant.
Figure 2: Return leadtime sensitivity
Additionally, if return lead times are decreased, inventory values would also decrease (see figure 2).
Although the model in this research was only applied to the XYZ product family, their current
software tool used to determine reorder points for the legacy tools, Servigistics, shows resem-
blance to the developed model for XYZ parts. With this research therefore, insight is gained
in how return lead times affect stock levels and that it is also important for the legacy tools to
have correct return lead times in the ERP system.
Recommendations
Conduct interviews with more Field Service Engineers, especially with engineers in Japan considering the high percentage of returns in that country.
Use the proposed control model to determine appropriate target fill rates and to re-evaluate the current stock levels. The output of the model indicated that some parts are stocked excessively and some parts insufficiently. In addition, too much stock is held in NA. This stock can be re-allocated to other warehouses.
As the values of the input parameters are important for the allocation of stock to parts, it is important that the company conducts research on input parameters and improves the accuracy of data in the ERP system. It was observed that the return lead times have not been updated for a long time. As shown in the sensitivity analysis, the return lead times are of great importance for the distribution of stock between the parts and warehouses.
This should not only be done for XYZ tools, but also for the legacy tools.
Reduce the return lead times. Reducing return lead times would lead to lower inventory
values. However, if the company decides not to change the return lead times, then at least
an effort should be made to make sure parts are returned in time.
Contents
Abstract III
Preface V
Management Summary VI
1 Introduction 1
1.1 Service Logistics . . . . 1
1.1.1 Global Planner . . . . 1
1.1.2 Order Desk . . . . 2
1.1.3 Purchasing . . . . 2
1.1.4 Part Return . . . . 2
1.1.5 Part Harvesting . . . . 2
1.2 Supply Chain Structure . . . . 3
1.3 Spare Parts Portfolio . . . . 4
2 Problem Definition & Research project 6 2.1 Problem Definition . . . . 6
2.1.1 Initial Problem Definition . . . . 6
2.1.2 Inventory Control . . . . 8
2.2 Research Project . . . . 10
2.2.1 Problem statement . . . . 10
2.2.2 Research Question . . . . 10
2.2.3 Project Approach . . . . 11
2.2.4 Deliverables . . . . 11
3 Analysis Returns 13 3.1 Data-analysis . . . . 13
3.2 Interviews . . . . 16
3.3 Costs . . . . 18
3.4 Conclusion . . . . 19
4 Spare parts inventory control model 22 4.1 Current spare part inventory control . . . . 22
4.2 Inventory control model . . . . 24
4.2.1 The model . . . . 24
4.2.2 Assumptions . . . . 26
4.3 Evaluation . . . . 27
4.3.1 Central Warehouse . . . . 27
4.3.2 Inbetween Hub . . . . 28
4.3.3 Local warehouse . . . . 29
4.4 Model formulation . . . . 30
4.5 Optimization . . . . 31
5 Case study 34 5.1 Implementation in Python . . . . 35
5.1.1 Evaluation . . . . 35
5.1.2 Optimization . . . . 36
5.2 Model Validation & Verification . . . . 37
5.3 Results . . . . 39
5.3.1 Values of input parameters . . . . 39
5.3.2 Results from case study . . . . 41
5.3.3 Sensitivity Analysis . . . . 43
6 Conclusion and Recommendations 49 6.1 Conclusion . . . . 49
6.2 Recommendations . . . . 50
References 52
Appendix A Implementation of Model 54
List of Abbreviations
APR Asia-Pacific Region
BH Inbetween Hub
CoGS Cost of Goods Sold
CW Central Warehouse
FRU Field Replaceable Unit FSE Field Service Engineer
HBO Hillsboro
LW Local Warehouse
SKU Stock Keeping Unit
List of Definitions
Aging Lead time from the moment of part shipment to moment that the part is returned to the local warehouse
CoGS Purchase price of a part
Consumables Parts for which the rate of depletion is directly related to a customer’s specific system application. Cartridges and oil are examples of consumables
Dead on Arrival(DOA)
A part has been delivered to the customer, but turned out to be defect after opening the box
Fill Rate % of order lines shipped out within 24hrs from the warehouse in which the order has been placed
FRU Parts which are considered to be stocked as part of the stocking plan
Legacy tools The portfolio of tools that the Material & Structural analysis Division offers, excluding EFA tools.
Net stock/in- ventory
The quantity on hand in inventory times the CoGS-value( $) minus the financial reserves.
Order line Order lines consist of the number of copies of a part ordered.
Parts Harvesting
Parts, mostly used for test, that are sent to Parts Harvesting instead of sending them to the central warehouse for repair. Parts are sent to Parts Havesting if they are parts from old microscopes and can reused
QC label broken
A label that is taped to the box to identify the part in the box was broken.
Repairable Parts
A part is repairable if the value of the part exceeds $2000, if it is technically feasible to repair the parts and if the repair costs are less than 30% of the original price.
Short Life A part has been delivered to the customer, installed in the microscope, but after a short period after installment the part broke down.
Spares Parts for which the rate of depletion is not directly related to a customer’s specific system application.
Unused Returns
Parts that are returned due to not being used during maintenance. The box should be unopened, otherwise the return is defined as ”used for test”
Used for Test Returns
Parts that are returned after being installed by the engineer, but did not fix
the problem. In case of an unused parts, but the box has been opened, this
part is also returned as ”Used for Test”.
1 Introduction
High-tech low-volume industries are characterized by long lead times, large bill of materials and complex resource requirements (de Kruijff, 2019). Downtime of the capital goods supplied by such industries can lead to substantial costs in case of downtime. As such, maintaining the capital goods requires the set-up of a service network in which maintenance and repair facilities are well controlled and locations are set up where enough spare parts are stocked.
This Master thesis project is executed within a high-tech low-volume company, hereafter referred to as company X, that supplies microscopes. In 2018 over 12000 microscopes of the company could be found in the field. Next to commercial sales the company is also active in the after- sales service. Company X offers 12 months warranty on products sold. After expiration of the warranty, customers have the opportunity to buy a service contract which covers corrective and preventive maintenance. Depending on the contract, the customer receives the corresponding service. Furthermore, customers can also buy Time & Material (T&M) if they do not have a contract. In this case they buy parts when needed and pay for the hours of maintenance.
1.1 Service Logistics
More specifically, this Master thesis project is executed within the Service Logistics department of the technology company. Service Logistics is responsible for the spare part inventory manage- ment. The activities of Service Logistics can be categorized by their function; Global Planner, Order Desk, Purchasing, Part Return and Part Harvesting. This section provides insight into the activities of these departments. The flow of goods and orders is controlled by an enterprise resource planning (ERP) system called QAD.
1.1.1 Global Planner
Global Planners are responsible for maintaining appropriate inventory levels of the spare parts portfolio in the global warehousing network and balancing the inventory targets with the service level commitments. In 2011 company X started using Servigistics, which is a software tool used for service part optimization. Servigistics offers a broad range of functionalities, including forecasting, demand planning and multi-echelon optimization.
The planners fill in a desired overall fill rate and the program calculates the fill rates per SKU and
the base-stock levels. Every day, the planners download an Excel sheet from Servigistics with
advice on what and how many parts to replenish in the central warehouse as well as an Excel
sheet that specifies the advice for local and regional warehouse replenishments. The planners go
through the lines one by one and use their own insight, with help of the program, to determine
whether to follow up on the advice or not. For this follow up, planners distinguish between
repairable and non-repairable parts. Advice for replenishments is always followed up, regardless
of the part category. In case of procurement advice however, planners use an additional program
created by the Business Intelligence department of the company. This program gives insight in the so called repair pool. Repair pool includes all inventory on hand, inventory in transit, inventory in repair and inventory in return.
1.1.2 Order Desk
When a customer requests repair, a call is created. A Field Service Engineer (FSE) or Global Technical Service (GTS) indicates what parts are deemed necessary for reparation. Order desk subsequently checks whether the parts are available in the local warehouse. If the part is in stock, a picklist is created and sent to the local warehouse. The local warehouse picks, packs and sends the part to the customer, usually within one day. If the part is not in stock in the local warehouse, a backorder is created and sent to the purchasers. There are two reasons for backorders to occur; either a part is not in stock at the local warehouse, but is at the world wide stock, or a part is nowhere in stock.
1.1.3 Purchasing
Purchasing is responsible for the inbound order flow/managing the open order books at the suppliers. If the planners determine that procurement or replenishment is necessary, purchas- ing receives a replenishment/procurement request and orders/replenishes the requested parts.
Moreover, if purchasing receives a backorder request from the order desk, they either transship the orders from one warehouse to the other or they buy a new part from the suppliers.
1.1.4 Part Return
The order and return process is presented in Figure 1.1 in a simplified way. Upon receiving a repair request, a field service engineer is sent to the customer. Maintenance is performed by the repair-by-replacement method: rather than repairing the parts at site, the FSE replaces the old part with a new part and sends the old part back to the warehouse for repair. Only in case of customized repairs, the part is sent for repair and returned to the customer. The engineer fills in the EQCR form, on which the reason for return needs to be specified, and sends the part back to the local/regional warehouse, where the decision is made to either repair the part, put it back in stock or let it be scrapped. A back-in-stock part, is put back in stock at the local/regional warehouse. If repair is needed, the part is sent to the central warehouse. A Return Material Authorization (RMA) request is sent to the supplier and finally the part that needs repair is sent to the supplier for repair. Used and non-repairable parts with a value lower than $2000 are scrapped by the FSE on site.
The $2000 threshold is determined by the financial manager and the logistics manager, based on their experience that the majority of the non-repairable parts were scrapped anyway and that most of the parts failed after the warranty period given by the supplier. Moreover, costs for returning parts are ± $1000-$1200.
1.1.5 Part Harvesting
Part harvesting buys back old parts & tools, dismantles them, tests & repairs the parts and
holds them in stock. In case a part is not in stock in the normal inventory, this part can be
ordered from the part harvesting inventory. This inventory is seen as a separate inventory.
Figure 1.1: High level return process
1.2 Supply Chain Structure
Company X has approximately 345 external suppliers and owns 3 production facilities that supply the company internally.
Spare parts used for maintenance are stored in 17 warehouses all over the world. Warehousing is outsourced to DHL. Hillsboro and Eindhoven serve as central hubs (and as regional warehouses) and supply parts to the regional and local warehouses. Immediately after delivery of parts in the hubs, the parts are redistributed between the local and regional warehouses. Additionally there are 10 local warehouses and 7 regional warehouses (including Eindhoven and Hillsboro).
Figure 1.2 shows a depiction of the service parts supply and the return flow. The external suppliers and internal suppliers deliver parts to the central hubs. The central hubs supply parts to the local and regional warehouses and the local and regional warehouses deliver parts to the customers. North America and Singapore also replenish three smaller local warehouses. Part return is explained in more detail in section 1.5.4, but the figure already shows the return flows.
All parts returned are firstly sent to the warehouse they were ordered from. The local/regional warehouse may send the items back to the central warehouse and finally the central warehouse returns the parts to the supplier.
Transport between the warehouses and to the customers is arranged with forwarders. Parts are
usually delivered to customers within 24 hours. Urgent orders can be delivered faster.
Figure 1.2: Service Part Supply
1.3 Spare Parts Portfolio
In order to obtain an optimal balance between costs and customer satisfaction Field Replace- able Units (FRUs) and Non Field Replaceable Units (Non-FRUs) are defined. New Product Development delivers a potential list of FRUs, after which Operations Engineering , NPI, in cooperation with Global Technical Support (Spare Parts Managers) decide on the final list.
GTS is later responsible to maintain it. Whether a part is FRU or non-FRU is dynamic and parts can change from being a FRU to becoming a non-FRU and vice versa. Criteria for FRUs are:
Part can and is expected to fail within lifetime of the system.
Part failure is fast to diagnose.
Part can be successfully replaced and system recovered.
Part Mean Time to Repair (MTTR) must meet design requirements.
A FRU consist out of a physical part and its related documentation including replacement and handling
Within FRUs another division is made: Consumables vs spares. Consumable parts are defined as those parts for which the rate of depletion is directly related to a customer’s specific system application. Whereas for spares the rate of depletion is not related to a customer’s specific system application. Cartridges and oil are examples of consumables and a printed card board is an example of a spare. Finally the consumables and spares can either be repairable or non- repairable. The Service Logistics department determines in which of these two categories the parts belong. A part is repairable if the value of the part exceeds $2000, if it is technically feasible to repair the parts and if the repair costs are less than 30% of the original price. The decision is made with help of the suppliers and technical staff. Figure 1.3 shows the subdivision of the FRUs. The figure shows the number of FRUs in the portfolio (e.g. out of 23871 FRU parts, 2214 parts are consumable) and the percentage of demand that is fulfilled by these items (e.g. the 2214 consumable parts account for 42% of the demand)
In order to obtain high service rate while minimizing costs, FRUs are kept in stock and non-
FRUs have to be ordered from suppliers when demanded.
FRU 23871 parts
Consumable
2214 parts
42% of usage Repairable
63 parts
4% of usage
Non-repairable
2151 parts
38% of usage
Spare
21657 parts
58% of usage Repairable
2382 parts
7% of usage
Non-repairable
19275 parts
51% of usage Figure 1.3: FRUs 2017
The breakdown of FRUs based on their purchase price can be seen in table 1.1. Most FRUs have a value between $200 and $4000.
COGS % of Total FRU portfolio
<$200 30%
$200- <$4000 50%
$4000- <$20000 14%
>$20000 6%
Table 1.1: CoGS of inventory 2017
The table can be read as follow 30% of the FRUs have a value of <$200
2 Problem Definition & Research project
After having introduced the company, this chapter will elaborate on the problems brought forth by the company (section 2.1). Gaining insight into the problem enables us to define the research project, which is discussed in section 2.2.
2.1 Problem Definition
The problem definition is divided in two parts: the first part will address the problems that the company initially wanted to be analysed (section 2.1.1). After some orientation, another possibility for improvement was detected. This will be discussed in the second part of the problem definition (section 2.1.2).
2.1.1 Initial Problem Definition
The service logistics department is facing two problems:
A relatively large percentage of spare parts is returning unused or used for test.
The return lead times from the customer to the local warehouse is highly variable and exceeds the desired 14 days.
Currently the company has no comprehensive insight into the effect, nor the reasons for this.
To get more insight into the problem, data-analysis is performed. Because the company was interested in the returns for 2018, return data for the period 01/01/2018 to 31/10/2018 is analysed. The dataset contains data for all the parts that are returned in that period and is extracted from QAD. This will reveal recent patterns in return flows.
Part Return
Service Logistics is responsible for the spare parts inventory management. As the department
wants to maintain a high service level, maintenance on the tools is done by the repair-by-
replacement principle. In 2018 28416 parts came back. Table 2.1 shows the number of parts
returned per return reason. Ideally, the table would also show the costs due to downtime of the
products and the decrease in value because of wrong ordering. Company X however does not
keep track of these costs.
Reason for Return Parts (pieces)
Customized Repair 84
Damaged in Shipment 198
DOA 2217
Installation Damage 15
New/Unused/Return to Stock 7611
Normal End of Life 11400
Part Damaged 348
Parts Harvesting 396
Parts Missing 45
QC Label Broken 1287
Recall 3
Short Life 2319
Used for Testing 2112
Vaccuum problem 3
Wrong Part in the Box 159
Wrong Part Ordered 219
Grand Total 28416
Table 2.1: Reasons for Return from 01/01/2018-31/10/2018
Some definitions require a bit more elaboration:
Dead on Arrival(DOA): A part has been delivered to the customer, but turned out to be defect after opening the box
Short Life: A part has been delivered to the customer, installed in the microscope, but after a short period after installment the part broke down.
QC label broken: A label that is taped to the box to identify the part in the box was broken. (note: not all parts have a QC label)
Used for Testing: a part has been used by the engineer, but didn’t turn out to solve the problem. Also in case the box was opened, despite possibly not using the part, the return has to be flagged as ”Used for Test”.
Parts Harvesting: Parts, mostly used for test, that are sent to Parts Harvesting instead of sending them to the central warehouse for repair. This is done if repair through the regular supply chain is not economically reasonable, but Parts Harvesting can attempt repair (even test if it is truly defective) and/or re-use components.
The largest percentage of parts returned is caused by orders that come back due to Short Life, DOA, Unused, Normal end of life, Used for testing and QC label broken. Together they account for 87% of the returns.
Parts that are returned unused and parts returned after testing are reasons that might be attributed to faulty diagnostics of the FSEs. These parts make up for 34% of the total returns.
Most of the unused parts go back to stock (table 2.2). Parts that are returned after testing
are mostly sent back to the central warehouse for repair. A significant value of parts that are
returned after testing is also scrapped. Sending parts to the central warehouse for repair and
scrapping parts unnecessarily is costly and should be avoided. This indicates the need for more
research on why parts come back unused and used for testing and how these can be prevented.
Action
after Return Unused Parts (pcs) Used for Test Parts (pcs)
Undefined in dataset 30 387
Back in Stock 7377 -
Direct Closed 18 -
Repair 45 777
Scrap 33 948
Grand Total 7611 2112
Table 2.2: Action after Return 2018
Aging
Another problem the Material & Structural Analysis department is facing is the aging of parts.
Parts should be returned within 14 days. If the part is not returned within 14 days, the part ends up in ”Aging”. As soon as the part is received by the local warehouse, the part is not in ”Aging” anymore. However, in practice the return lead times from 01/01/2018-31/10/2018 turned out to be 24 days on average. This is displayed in table 2.3. The standard deviation of the shipment lead times is in most cases larger than or close to the mean, indicating that the variance in return lead time is large. The largest return lead time is in China, and the smallest in Japan.
Warehouse Average
Aging
StdDev Aging
Min Aging
Max Aging
Australia warehouse 37 38 1 193
China warehouse 39 37 0 321
EU warehouse 30 38 1 324
HBO warehouse 21 33 0 786
Israel warehouse 21 26 2 184
Japan warehouse 15 16 0 190
South Korea warehouse 21 21 0 136
Singapore warehouse 28 33 0 274
Taiwan warehouse 16 19 0 179
Grand Total 24 32 0 786
Table 2.3: Aging 01/01/2018-31/10/2018
2.1.2 Inventory Control
As was mentioned at the start of this chapter, after orientation another possibility for improve- ment was detected. This problem will be discussed hereafter.
Three years ago company X added a product family to their portfolio after acquiring a company:
XYZ. In addition to many parts being requested only once per year, XYZ’s installed base
accounts for only 4% of the total installed base. Targets for fill rates are set for the complete
portfolio and not per product family. When using Servigistics to stock the parts for tools,
the company observed that the stocking strategy for the XYZ parts led to very low aggregate
fill rates for this tool. Intuitively this is because of the low demand rates combined with the
small installed base relative to the other tools. Due to this reason, the stocking strategy of the
XYZ parts is done separate from Servigistics with simple Excel sheets. Up to now however, the
performance for this product family is not as required. This is displayed in table 2.4. The service
measure used to evaluate company performance is fill rate. Fill rate is defined as percentage of order lines shipped out within 24 hours from the warehouse in which the order has been placed.
Moreover, order lines consist of the number of copies of a part ordered. The desired fill rate for the XYZ product family is 90%. EU especially is not able to meet the requirements, with an average realized fill rate of 56.4%
HBO APR EU
Month Fill Rate Nr or
Order Lines Fill Rate Nr or
Order Lines Fill Rate Nr or Order Lines
Januari 88.6% 44 93.1 29 42.9% 7
Februari 86.0% 43 94.9 38 53.8% 13
March 81.0% 42 92.5 53 66.7% 6
April 81.8% 33 90.9 44 77.8% 9
May 84.6% 39 91.7 48 52.9% 17
June 92.2% 51 94.4 18 33.3% 3
Grand Total 86.1% 252 92.9% 230 56.4% 55
Table 2.4: Realized fill rate per region for January-June 2019
2.2 Research Project
This section presents the research design and methodology. First the problem is defined. Second, the research questions are given and finally, the methodology and the deliverables of the project are presented.
2.2.1 Problem statement
The company believes that it can reduce the net stock via a more efficient usage of its parts and a decrease in the long and variable return lead times. The net stock is defined as the quantity on hand in inventory times the CoGS-value( $) minus the financial reserves. The target for the net stock is ≤$125 million and current net inventory equals $122 million. Currently the company has no comprehensive insight into the effects of these unused returns, used for test returns and the long and variable return lead times, nor how to deal with it. Additionally, the department is interested in other ways to determine the stocking strategy for their XYZ tools, such that higher fill rates can be achieved at lower stocking levels.
The research project is therefore aimed at analysing a multi-echelon supply chain and to develop ideas on how to improve the supply chain, by considering the long and variable return lead times, the unused and used for test returns and a different stocking strategy.
The following assignment is defined:
Analyse the current inventory model, thereby quantifying the effects of the unused returns, used for test returns and the long and variable return lead times.
Identify ways to improve the spare part supply chain and quantify the improvement po- tentials.
2.2.2 Research Question
The purpose of this thesis is to provide insight into the unused and used for test returns, the long return lead time on the inventory and on a new stocking strategy for XYZ parts. The contribution of the thesis is twofold. First, the effect of the unused returns and the return lead times will be quantified as to show the impact on company revenues so that a cost-benefit analysis can be made with regards to the improvement actions. Second, a software tool is developed for an improved stocking strategy for XYZ parts. With this tool, also the costs associated with the long return leadtimes can be evaluated by running different scenarios.
Hence, the following main research question has been defined:
How can company X improve their spare part inventory management such that costs are mini- mized, while not compromising on the service measures, i.e. fill rate?
In order to answer the research question the following subquestions are specified:
1. Why do parts come back unused or used after testing?
2. What are the costs associated with unused and used after testing returns?
3. What is an appropriate mathematical model for the inventory control of XYZ tools?
4. How does the modified inventory control model perform compared to the current inventory
control model?
5. What is the effect of long and variable return lead times on the inventory value (and fill rate)?
2.2.3 Project Approach
The project will be addressed using the Business Problem Solving approach by Van Strien (1997), illustrated in figure 2.1. The previous sections concerned the problem mess and problem definition. The rest of the project will focus on the analysis, diagnosis and the development of an action plan. The action plan consists of developing a design that is suitable for dealing with the problem statement. Intervention and evaluation are outside the scope of this project.
Figure 2.1: Regulative Cycle adapted from Van Strien(1997)
2.2.4 Deliverables
In accordance with the company the following deliverables are defined:
The results of the thesis will be input for a document to be delivered with data translated to management information.
A tool will be developed that will aid the department in reducing the net stock by cal-
culating the optimal inventory levels. In addition to determining a stocking strategy, the
tool will have other functionalities. The new decision support tool will also include options
for running different scenarios for the XYZ parts, e.g. increasing the return lead time of
repairable items, increasing the procurement lead time of non-repairable items etc.
PART I
3 Analysis Returns
In this chapter the first and second sub questions are answered;”Why do parts come back unused or used after testing?” and ”What are the costs associated with the unused and used after testing returns?”. First, data-analysis was performed to see whether certain patterns could be recognized and second, interviews were conducted with several stakeholders. It is important to keep in mind that since this was not the core of the thesis and time was limited, this part only serves to give the company a limited set of guidelines.
3.1 Data-analysis
Table 2.1 showed that used for testing returns and unused returns account for 34 percent of the total returns in pieces. Considering that these items can be attributed to internal factors within company X, it seems a good case to investigate the reasons for these returns and how they can be reduced. Data-analysis was performed to see whether certain patterns could be distinguished, such that correlations could be found between different factors and return reasons.
Returns per country
Warehouse 100*((Unused + Used for Test)
/(Total Returns)) Parts (pcs)
Australian warehouse 11% 21
China warehouse 29% 663
EU warehouse 31% 2229
HBO warehouse 38% 3135
Israel warehouse 42% 84
Japan warehouse 51% 2643
Singapore warehouse 11% 132
South Korea 5% 120
Taiwan warehouse 43% 696
Grand Total 37% 9723
Table 3.1: Unused/Used for test Returns per warehouse 01/01/2018-31/10/2018
Segmenting the returns per country shows that there are large differences between countries.
Japan, Taiwan and Israel have the highest percentage of orders unused or used for test, especially compared to Australia, Singapore and South Korea (table 3.1). This means that especially attention has to be paid to these regions.
In addition, the differences between countries are investigated. Table 3.2 shows the percentage of returns per warehouse relative to the total amount of returns, i.e. 35.55% of the unused returns are from HBO. The table also shows the percentage of quantity ordered per warehouse relative to the total quantity ordered, i.e. 36.52% of the total quantity ordered is from HBO.
One can see that most of the returns originate from EU, HBO and Japan. EU and HBO also
account for most of the quantity ordered. Since the largest percentage of the total installed base
is located in these regions, it isn’t surprising that these regions also show the largest percentage of parts returned. However, the used for testing returns in EU seem to be relatively high: 24%
of the quantity ordered is from Europe, but Europe accounts for 38% of the used for test returns.
Moreover, Japan accounts for a relatively small percentage of the total quantity ordered, but does account for a significant percentage of parts returned unused and used for testing.
Quantity ordered (pcs)
Quantity ordered (% of total)
Unused Returns(pcs)
Unused Returns (% of total)
Used for Test Returns (pcs)
Used for Test Returns (% of total)
Australia 2526 1.19% 3 0.04% 18 0.85%
China 18804 8.88% 606 7.96% 57 2.70%
Europe 51222 24.18% 1425 18.72% 804 38.07%
Hillsboro 77358 36.52% 2706 35.55% 429 20.31%
Israel 1923 0.91% 78 1.02% 6 0.28%
Japan 16953 8.00% 2130 27.99% 513 24.29%
South Korea 18231 8.61% 117 1.54% 3 0.14%
Singapore 11499 5.43% 39 0.51% 93 4.40%
Taiwan 13311 6.28% 507 6.66% 189 8.95%
Total 211827 100% 7611 100% 2112 100%
Table 3.2: Unused returns and Used for Testing returns per warehouse 01/01/2018-31/10/2018
Returns per part
A more detailed table on the returns is shown in table 3.3 and table 3.4. Visualized is the top 29 of parts returned unused and used for testing per year. The tables are constructed as follows:
first the parts are sorted by amount of times they were returned in the years 2016, 2017 and 2018, after which the top 40 most returned parts was selected. If the part belonged in the top 40 for all those years, the cell is highlighted in red. If the part belonged in the top 40 for two years, the cell is highlighted in yellow and finally if the part belonged in the top 40 for only one year, the cell is highlighted in green. The tables clearly show that some parts are returned in all three years, indicating that the returns for these parts are not random incidents.
Table 3.3:
Unused returns 01/01/2016- 31/10/2018Table 3.4:
Used for Test returns 01/01/2016- 31/10/2018Reasons for returning parts unused/used for testing
Moreover, data-analysis was performed to check whether certain correlations could be found between return reasons and other factors. There was however no clear cut conclusion to be drawn out of this analysis because of two reasons:
1. There were no clear outliers.
In 2018, 211827 parts were ordered and sent to customers. Out of these parts, 7611 came back unused and 2112 came back after being used for test. If some types of parts were returned very often, a possible correlation could be found in the data between the parts that came back and the reason they came back unused. However, in this case the part that was returned unused or used for testing most was part 23900, which was returned unused 96 times (table 3.3) and ordered 1950 times. It could have been possible to find a correlation between a certain factor and the amount of returns for this part, however no clear outliers were detected, i.e. if 90 out of 96 returns for this part come from one country, this would indicate a strong correlation between country and unused return.
No such correlation was found. Additionally, it was tested whether high-level (returns aggregated over all countries) reasons could be distinguished. This is discussed below.
It was checked whether certain markets have a high percentage of unused or used for test returns. This was not the case. Company X operates in two markets, namely Science- market (universities, research centers etc.) and Semi-market (semi-conductor market).
The percentage of returns relative to the total number of returns for the Semi-market is 44% and for the Science-market 49%. The total quantity ordered coming from the Semi- market account for 47% of the total number quantity ordered, and the total quantity ordered coming from the Science-market 52%. Note that the numbers add up to 93%
instead of 100%. This is because for some of the parts no market was indicated in the dataset. It does not seem to be the case that any of the two markets has a relatively high percentage of unused or used for test returns, as the percentage of returns for both markets is consistent with the percentage of quantity ordered for these markets.
In table 3.5 the percentage of returns for different product families and several warehouses is displayed. It shows per product family, i.e. APX, CMD, etc., the total number of parts that are returned unused or after being used for testing and the percentage of returns per product family per warehouse relative to the total quantity ordered for that product family per warehouse, i.e. 29% of the parts that were ordered for SEM in Japan are returned unused or after used for testing. The numbers are only shown for Europe, Japan and Hillsboro because these were the warehouses with the highest percentage of returns.
The only real outlier seems to be SEM in Japan, for which 29% of all ordered parts are returned. This is high compared to the returns of the other product families in Japan.
Warehouse APX CMD XYZ FPD LDB SDB SEM TEM XRS
Europe - - 8% 5% 86% 6% 6% 6% 0%
Hillsboro 6% 3% 7% 3% - 5% 6% 6% 14%
Japan 21% 19% 4% 20% - 17% 29% 18% -
Total Returns 132 51 330 624 54 3756 1593 2565 3
Table 3.5: Unused/Used for testing returns per product family
2. Possible interaction between many different factors, e.g:
The company operates in different markets, namely Science and Semi.
The company offers many different contracts. Ranging from complete service con- tracts to contracts that only cover labour or parts. Also, customers have the possi- bility to buy Time & Material, without having a contract.
There are many different microscopes. Even within product families, different prod- uct series exist.
Because of the possible interaction between many different factors, finding a correlation between number of returns and reason for return based on data-analysis will be complex.
Field Service Engineers
Additionally, analysis was done with the field service engineers that returned the parts. It might be possible that certain engineers return parts unused or used for testing more often than other engineers. A list with the number of parts returned unused per engineer was given to the company and showed that there are certainly engineers that return parts often compared to other colleagues. Due to confidentiality reasons, the list is not included in the thesis.
Summary
In conclusion, analysing reasons for parts being returned unused or used after testing based on data is a complex task. No clear outliers were detected and the possible interaction between different factors make that endless combinations exist for reasons of return. Because the parts are returned by field service engineers, interviews could provide more insight. Therefore the next section will present the outcomes of the interviews with several stakeholders.
3.2 Interviews
In order to find the cause of the unused and used for testing returns three interviews were conducted, two with managers within Global Technical Service and one with a field service engineer from the Netherlands. In table 3.6 the different reasons are displayed and the number of times the reasons were mentioned. It must be noted that a sample size of three is too small to give accurate results. However, as mentioned in the introduction, this section only serves to give a limited set of guidelines so that the company gains a little more insight.
Different root causes can cause the same failure (3)
Complex microscopes, difficult to diagnose cause of failure (3) New microscopes, less knowledge about repair (1)
While waiting for parts, engineer found other solution that fixed the failure (2) Pressure from the semi-customers.
Semi-customers most often have contracts with expensive down time (2) No KPI for return flows (no punishment) (2)
Weak communication between FSE and GTS (1)
No correct diagnosis because certain tools are not available (work instructions, training or troubleshoot methods) (2) Long travel time from engineer to customer (3) No awareness of costs (3)
Table 3.6: Return reasons based on interviews
The reasons are subsequently divided into 6 categories.
Machines: Because of the complexity of machines, it can be difficult to diagnose the root cause of the failure. Multiple parts can be ordered to assure that the right part is available at the time of repair.
Customer: The company operates among others in the semi-conductor market. Because downtime at these customers is very costly and expensive, engineers feel pressured by the customers to repair the failure quickly and thus order multiple parts.
Contract: Customers can buy a service contract or they buy materials and pay for the time of the engineers. If demand occurs from a customer with a contract engineers might order multiple parts, because downtime is expensive and if the failure is not fixed within a certain time frame the company must pay a high price. In case of failure at a customer that bought T&M, the customer must pay for the time of the engineer and therefore engineers might be more careful with ordering parts.
Travel Time: Because of long travel times to the customers engineers order multiple parts, so that they do not have to come back if a wrong part was ordered.
Processes: Lack of tools needed for correct diagnosis, e.g. work instructions, training or troubleshoot methods. Moreover, local service organizations are charged for ordered parts. These local service organizations are responsible for the budget. There is however no KPI for the return flows or a punishment. A lack of measures like such can lead to no ownership and thereby to excessive ordering.
People: Because of the many types of product series, some engineers might have less knowledge about certain product series than other engineers. These engineers might order multiple parts leading to unused returns. These engineers might also install the wrong parts, which can explain the parts being returned due to used for testing.
Because the field service engineers are responsible for returning the parts unused and used for testing, the section below will elaborate more on the interview with the field service engineer.
According to the FSE, the FSEs do not feel pressured by the customers. Time pressure however, can contribute to ordering more items: engineers want to fix the microscope as fast and efficient as possible. Therefore it happens that the engineer orders two parts, one part that can be replaced very quickly and one part that takes a lot of time to replace. Engineers first try to fix the failure with the quickly replaceable part and if that doesn’t solve the problem, the engineer tries the other part. In these instances, the engineer makes this decision in accordance with Global Technical Support.
Moreover, the engineer believes that there is enough training, although in very rare instances there is a lack of hands-on training. Improvement in the troubleshoot method is possible.
When an engineer receives a call from the customer, the engineer checks with a service CD what action can be taken to repair the failure. The CD shows what might be the cause of the failure. Cooperating with experienced field service engineers, by asking feedback, during the development of service CD’s can give a more comprehensive view of different failures and their causes. Another improvement with regard to the availability of tools needed for correct diagnosis is by giving the engineers a better insight into the most common failures in tools and how other engineers fixed these problems in earlier repairs
After showing the engineer the list of most returned parts (tables 3.3 and 3.4), several insights were gained:
Many of the parts returned are due to difficulty of correct diagnosis, because the failure
can be caused by several different failed parts, e.g. there are many Printed Card Boards
(PCB) in the microscope, the engineer does not exactly know which PCB caused the breakdown.
Several parts in the top 10 of unused and used for testing returns are parts that are ordered together because of the time it takes to replace certain parts, e.g. Field Emission Gun and HV-connector. The HV-connector is the cable that supplies power to the Field Emission Gun. Replacing the HV-connector takes 2 hours and replacing the Field Emission Gun takes 3 days. The engineer first tries to fix the problem by replacing the HV-connector.
If this doesn’t solve the problem, the engineer replaces the Field Emission Gun. Note that this reason is not equivalent to difficulty of correct diagnosis. In this instance, the engineer knows why the failure occurs.
For some of the parts the engineer was surprised that they were returned that often, because either the parts are very small and can be stored at the customer’s site or diagnosis for the repair is obvious and the part should not have been ordered.
3.3 Costs
In this section an overview of the costs associated with returning the parts unused and used for testing is given.
Quantifying the handling and transportation costs is not straightforward. In the total costs, no distinction is made between forward flows and return flows. Moreover, the company has contracts with forwarders and the company only receives digital invoices from one of them.
Checking the invoices one by one would be too time-consuming. For one of the forwarders a dataset was obtained with help of the transportation manager. This data set contains infor- mation about the freight costs per package for Europe in 2018. However, this data set too does not distinguish between forward and return flows. To overcome this, a selection was made for packages that were sent to the Netherlands from other countries. We should keep in mind that this selection also includes packages sent from suppliers to the Netherlands. The average freight costs per package based on this dataset for packages with the Netherlands as destination country is e33.58.
We can now calculate the total amount of money spent on transportation for unused and used for testing returns in 2018(until October 31th). First the costs should be converted to dollars instead of euros. e33.58 equals $38.10. Multiplying the average costs per return shipment by the total number of return shipments leads to total return transportation costs of $370,446.30 for the period of 01/01/2018-31/10/2018. Do keep in mind that this is only a rough estimation, since the data set only included returns in Europe and that this data comes from only one of the many forwarders. It was assumed that sending parts back to other countries costs the same as sending parts back to Europe. Logically, this is method is flawed. It was however the best that could be done at the moment.
Forward transportation costs should also be calculated to give a more complete calculation of the total transportation costs. If parts are sent unnecessarily to the customer, these costs are of course unnecessary costs. This is however much more difficult, since if multiple parts are ordered, the transportation costs from the warehouse to the customer are shared by these parts.
The costs incurred for transportation are not lost and this should not be added to the total costs
of returning parts unused and used for testing, because at least one of the parts was used for the
repair (assuming that the additional weight incurred by the extra parts does not lead to extra
costs). If however, only 1 part was ordered, the costs for transporting this part to the customer
should be added to the total costs of returning parts unused and used for testing. According to
the same data set that was used for the transportation costs for returning items, the average costs per package for sending items to customers in Europe is e36.96, which is equal to $41.72.
2766 parts that were returned, were ordered in singles(i.e. the order consisted of only that part and no other parts). Hence, additional costs for forward transportation due to unused and used for test returns is 2766 × 41.72 = $115,397.52.
Unused returns are put back in stock as soon as they arrive at the warehouse. Therefore, the main costs incurred by these parts are the transportation costs and the handling cost at the warehouse. Transportation costs have been treated above and handling costs per package are very low and are therefore left out of the analysis.
Used for testing returns have additional costs. As was shown in section 1.6.1 948 parts that returned after testing were scrapped and 777 parts were sent for repair. The value of the scrapped items was e608,345.31 or $688,617.09. Calculating the costs for repair of the items was not simple, due to lack of information. There exists a dataset that defines the average repair cost per part. However, since items are also flagged as used for testing if only the box was opened, it could be the case that the costs of repairing these items is much less. Nevertheless, if we use the average repair costs, repairing these items costs e2,092,145.82 or $2,368,206.54 In conclusion, the total costs incurred by unused and used for testing returns for the period 01/01/2018-31/10/2018 equals the sum of the transportation costs, the value of the scrapped items and the repair and scrap costs, which is $3,542,667.45.
3.4 Conclusion
In this chapter an effort was made to answer the question ”Why do parts come back unused or used after testing?” and ”What are the costs associated with the unused and used after testing returns?”.
The analysis revealed that Japan has the highest percentage of unused and used for testing returns. In addition, Europe accounts for a large percentage of used for testing returns. Atten- tion has to be paid to these regions. Moreover, many parts that are returned end up in the top 40 of parts returned year after year, indicating that the returns for these parts are not random incidents.
Furthermore, difficulty in correct diagnosis seems to be the main reasons for returning parts unused and used for testing. Looking at the number of times parts are returned per engineer, it seems that some engineers return parts unused/used for testing more often than others.
A major drawback of this analysis is the sample size of the interviews. Only 3 interviews were conducted, out of which 2 were with managers and not with field service engineers. The advice for the company is to conduct more interviews with field service engineers and follow up on the list with the top parts returned unused and used for testing. To gain more insight into the problem at least 4 engineers should be interviewed from Europe, 4 from Japan and 4 from HBO. Two interviews with an engineer that returns parts very often and two engineers that return parts not so often.
One should keep in mind however, that a trade-off needs to be made between ordering multiple
parts at once and the costs of sending engineers multiple times to the field to fix the machine. For
this trade-off to be made, it is important that more insight is gained into the costs of returning
the parts. From the analysis of the costs is was clear that there are many uncertainties with
regards to the costs, since many assumptions were made. Nevertheless, the costs calculated in
this research for the period 01/01/2019-31/10/2019 associated with the unused and used for
test returns were $3,542,667.45. 70% of these cost are due to the repair costs of the used for
testing returns. This indicates that more research is needed, especially for the used for testing
returns because of the unnecessary scrap and repair costs.
PART II
4 Spare parts inventory control model
This chapter introduces the second part of the thesis, which is the core of the thesis and is focused on the development of an improved inventory control model. First the current situation will be described (section 4.1), after which an improved inventory control model is developed (section 4.2) and verified & validated (section 4.3).
4.1 Current spare part inventory control
As was discussed in section 1.4, company X operates through central, local and regional ware- houses. The stocks in the warehouses are controlled by the Service Logistics department with the help of a software tool called Servigistics.
In determining target measures two approaches are typically followed: The item approach and the system approach. In the item approach a service measure is defined per item and in the system approach a target service measure is defined for the demand weighted average of the performance measure over all items(Topan, Bayındır, & Tan, 2017). Servigistics determines the service measures and targets based on the system approach.
Example Consider a system with three SKU’s. SKU 1 has a demand rate of 5, SKU 2 a demand rate of 6 and SKU 3 a demand rate of 4. The aggregate target fill rate is 0.95. In the item approach a fill rate would be defined per SKU, e.g. 0.95 per SKU. In the system approach an aggregate target fill rate per system is defined and calculated as follows: 0.95 =155 × fill rate of SKU 1 +156 × fill rate of SKU 2 +
4
15× fill rate of SKU 3. The difference between the approaches is that in the system approach there is no predefined target fill rate per SKU. The system approach can lead to higher fill rates for cheap parts and lower fill rates for expensive parts, e.g. a fill rate of 0.99 for SKU 1, 0.90 for SKU 2 and 0.98 for SKU 3 would also lead to an aggregate target fill rate of 0.95.
The module in Servigistics to calculate base stock levels is called Multi Echelon Optimization (MEO).
Calculations within this module are mainly based on the work of Sherbrooke(1968) and Muckstadt(2004).
Multi-echelon optimization uses probability distributions and target service levels to determine the stock- ing strategy. Optimization of the base stock levels is done by marginal analysis, meaning that the part and location giving the highest ”bang for the buck” alternative is chosen for allocation, i.e. is it better to stock part X in a local warehouse in Europe or to stock the part in the central warehouse in North America or is it better to stock part Y in a local warehouse in Europe or to stock the part in the central warehouse in North America. MEO makes a trade off between the increase of the aggregate fill rate if part X is added to the inventory and the cost of that part. To illustrate this, the example below is given.
Example Stocking part X in the local warehouse in Europe, increases the aggregate fill rate from 70%
to 75%. The price of X is$100,000.00. Stocking part Y in the local warehouse in Europe, increases the fill rate from 70% to 71%. The price of Y is$100.00. Despite the larger increase in the aggregate fill rate when X is stocked, it might be possible that Servigistics allocates part Y because of the significantly lower price of Y. How this trade-off is made is explained in more detail in section 4.5.
