Suggestions for further research - CONCLUSIONS AND RECOMMENDATIONS

9. CONCLUSIONS AND RECOMMENDATIONS

9.4 Suggestions for further research

Multiple assumptions are used in order to scope the problem. For this reason, a first suggestion for further research is to relax these assumptions and analyse the effect of the relaxation on the results.

The validation of the redesign involved four aspects that ask for further research. First, the redesign is only feasible for SPO projects. The SPO is part of Parcel and Postal projects. There are multiple other

Items Ratio

Lead-time ≤ 4 weeks 5052 85%

Lead-time > 4 weeks 857 15%

projects that contain purchase items used for the SPO. Consequently, it is suggested to first analyse those projects and extend the redesign over those items, and lateron analyse other projects.

Secondly, in Chapter 7 and 8, it is assumed that the lead-time to America is equal to the lead-time in Europe. Relaxing this assumption will have a big impact, as the lead-time to America can be a couple of weeks longer due to extra transportation time. The new sub-order allocation rule defined in sub-section 7.4.3 will make it possible to estimate the amount of demand that is required in America, however Vanderlande first has to gain information about all different characteristics before it can implement the true situation in America. Shipping material asks for batch sizing and consolidation of items, two aspects that are not considered in this master thesis project. On top of that, multiple political issues arise when shipping material to America instead of buying it there. Therefore a suggestion would be to first analyze the control structure of the factory and the differences between Europe and America in a logistical but also political manner.

Another limitation of the research is the lack of quantitative support for the redesign. It is suggested to analyse the quantitative effect of a redesign of the processes, control structure, and decision structure.

The fourth aspect for further research involves the IT capabilities. It is validated whether it is possible to adjust the IT system, but the detailed adjustments are not clarified yet.

The validation of the inventory control policies involved some more aspects that ask for further research.

Firstly, an assumption that is made in the inventory control policy is that the lead-time for the purchase item is five weeks. However, the dataset of the purchase items shows that some items have a lead-time of 6, 7 or 8 weeks. The effect of these lead-times on the inventory model have to be investigated in a next research project.

The demand for the first item was rather small, therefore it is hard to draw conclusions from this output.

The second item had a larger demand, resulting in more thrustworthy output. However, only two items were used as input. The operational validity can be extended by adding more items to the model.

Moreover, only 25 weeks with data were used as input for the demand of component (B) and (C). This can be extended in further research, making the research more reliable.

The SPO Future demand currently contains information about all characteristics but three. If SCCE is able to add information about the three characteristics, it can decrease uncertainty in purchase items demand and thus inventory. This should be investigated in cooperation with sales engineering.

Moreover, the planning of the SPO Future demand is considered as deterministic in this research. In real-life the planning can deviate, as the customer might ask for acceleration of deceleration. The possibility for deviations to the control structure and inventory control policy will increase the fit to the structure of Vanderlande. Furthermore, a what-if scenario on multiple second tier suppliers for a specific purchase item can be executed.

59 BIBLIOGRAPHY

Adan, I., Van eenige, M., & Resing, J. (1995). Fitting discrete distributions on the first two moments.

Bertrand, J., & Muntslag, D. (1993). Production control in engineer-to-order firms. Eindhoven, The Netherlands: Elsevier.

Bertrand, J., Wortmann, J., Wijngaard, J., Suh, N., Jansen, M., Fransoo, J., . . . de Jonge, T. (2015).

Design of Operations Planning and Control Systems. Eindhoven: School of Industrial Engineering, Eindhoven University of Technology.

Bijvank, M. (2013). Periodic review inventory systems with a service level criterion. Journal of the Operational Research Society 65, 1853-1863.

Chopra, S., & Meindl, P. (2013). Supply Chain Management. Edinburgh : Pearson Education Limited.

Christopher, M. (2005). Creating value-adding networks. Logistics and Supply Chain Management.

Durlinger, P. (2013). Demand Management: Voorraadbeheer. Durlinger Consultancy.

Gosling, J., Towill, D. R., Naim, M. M., & Dainty , A. R. (2009). Engineer-to-order supply chain management: A literature review and research agenda. International journal of Production Economics, 741-754.

Hobday, M. (2000). The project-based organisation: an ideal form for managing complex products and systems? Science and Technology Policy Research, 871-893.

Iida, T. (2015). Benefits of leadtime information and of its combination with demand forecast information. International Journal of Production Economics.

Ishikawa, K. (1990). Introduction to quality control. Tokyo, Japan.

Law, A. M. (2007). Simulation modeling and Analysis (Fourth edition). Tucson, Arizona, USA: McGraw-Hill.

Mitroff, I. I. (1974). A methodology for strategic problem solving. Management Science.

Ouyang, L.-Y., Wu, K.-S., & Ho, C.-H. (2004). Integrated vendor-buyer cooperative models with stochastic demand in controllable lead time. International journal of production economics.

Radke, A. M., & Tseng, M. M. (2012). A risk management-based approach for inventory planning of engineering-to-order production. CIRP Annals - Manufacturing Technology, 387-390.

Sargent, R. G. (2013). An introduction to verification and validation of simulation models. Syracuse, NY, USA: Syracuse University.

Shrivastava, P. (1987). Rigor and practical usefulness of research in strategic management. Strategic Management Journal, 77-92.

Topan, E. (2014). Supply Chain Operations Planning, Distribution systems. Eindhoven.

Van Aken, J. E. (2007). Problem solving in Organizations, A methodological handbook for business students. Cambridge: Cambridge University Press.

Vanderlande. (2015). Annual report. Retrieved from https://www.vanderlande.com/about-vanderlande/annual-report

61 APPENDIX

Appendix A. List of Abbreviations Abbreviation Description

CDF Cumulative Distribution Function

EDC European Distribution Centre

EOQ Economic order quantity

ETO Engineer-to-order

ERP Enterprise Resource Planning

FPS Field Problem Solving

JDE J.D. Edwards World; Enterprise Resource Planning software

KPI Key performance indicator

LR Literature Review

MOQ Minimum Order Quantity

MRP Material Resource Planning

ON Inventory Order

PDF Probability Distribution Function

PI Purchase item

PIO Purchase item order

PO Production order

OF Factory Order (Veghel)

RP Research Proposal

SCE Supply Chain centre Europe

SPO Posisorter

TU/e Eindhoven University of Technology

V Vanderlande Industries

VIA Vandende’s factory in America

VIM Vanderlande’s factory in Veghel

VIS Vanderlande’s factory in Spain

VP Vice President

WK Work Kit

WO Work Order

62 Appendix B. Figures, Tables and Graphs

Figure 1, Redesign of the control structure ... V Figure 2, ETO trade-off by Radke and Tseng (2012) ...1 Figure 3, Product hierarchy ...5 Figure 4, Material and information flow ...6 Figure 5, Lead-time from factory order release until finish date ...7 Figure 6, Ishikawa diagram of factory performance ... 11 Figure 7, Layout of the Posisorter ... 12 Figure 8, Regulative cycle by Van Strien (1975) ... 16 Figure 9, Situation 1: Local decision making and local stock points ... 18 Figure 10, Situation 2: Central decision making and local stock points ... 18 Figure 11, Situation 3: Central decision making and central stock ... 19 Figure 12, Sub-components and types of Component (B) ... 21 Figure 13, Sub-components and types of component (C) ... 22 Figure 14, Sub-component groups of Component (F) ... 22 Figure 15, Adjusted Regulative Cycle ... 26 Figure 16, Current Control Structure ... 27 Figure 17, Procurement activity ... 30 Figure 18, Goods Flow Control and production phases... 30 Figure 19, Aspects of GFC at Vanderlande ... 34 Figure 20, Production control redesign ... 34 Figure 21, Customer order acceptance function ... 34 Figure 22, Sub-order assignment and PU outsourcing decision ... 35 Figure 23, WO release function ... 36 Figure 24, Redesign control structure ... 36 Figure 25, Relationship between different BOM types (Bertrand and Muntslag, 1993) ... 37 Figure 26, Relation between theoretical model and simulation ... 42 Figure 27, Simplified version of the model of development process (Sargent, 2013) ... 48 Figure 28, Variants of component (B) ... 69 Figure 29, Variants of component (C) ... 69 Figure 30, Variants of component (F) ... 70

Table 1, Performance measures second tier suppliers ... 13 Table 2, Comparison of suppliers VIM ... 14 Table 3, Key Performance Indicators set for Inventory Control ... 17 Table 4, Example demand analysis ... 20 Table 5, Components and their measurements ... 20 Table 6, Improvement possibilities ... 28 Table 7, Production Units at Vanderlande ... 32 Table 8, Holding cost percentage according to Durlinger (2013) ... 47 Table 9, Input values, item 1 and item 2 ... 49 Table 10, Replenishment levels, item 1 ... 50 Table 11, Results of model 1 and 3, item 1 ... 50 Table 12, Results of model 2 and 4, item 1 ... 50 Table 13, Sensitivity analysis, item 1 ... 52 Table 14, Sensitivity analysis, item 2 ... 52 Table 15, Lead-time division of all purchase items ... 56 Table 16, Ratios of variants of component (B) ... 71 Table 17, Ratios of variants of component (C) ... 72

Table 18, Ratios of variants of component (F) ... 73 Table 19, Overview of PI analysis ... 74 Table 20, Variants and items of component (B) ... 78 Table 21, Variants and items of component (C) ... 79 Table 22, Results fitting procedure (Adan et al., 1995) ... 80 Table 23, Operational validation item 1 ... 82 Table 24, Operational validation item 2 ... 82 Table 25, Replenishment levels item 2 ... 83 Table 26, Results of model 1 and 3, item 2 ... 83 Table 27, Results of model 2 and 4, item 2 ... 83

Graph 1, Performance and costs of an inventory model with lead-time uncertainty ... VII Graph 2, Weekly performance VIM ... 10 Graph 3, Weekly performance VIS ... 10 Graph 4, SPO Future demand in meters ... 12 Graph 5, Delivery to request, second tier suppliers VIM... 13 Graph 6, Delivery to request, second tier suppliers VIS ... 14 Graph 7, Average performance and cost comparison, item 1... 51 Graph 8, Delivery to request, second tier suppliers SPO VIM ... 68 Graph 9, Delivery to request, second tier suppliers SPO VIS ... 68 Graph 10, Historic demand of component (C)... 77 Graph 11, Warm-up period... 80 Graph 12, Validation of Binomial distribution ... 82 Graph 13, Average performance and cost comparison, item 2 ... 83

64 Appendix C. Assumptions

All assumptions discussed throughout the research are described in order of occurrence.

Chapter 3: Performance measures analysed in section 3.1 and 3.4 are reliable

The performance of the factories is controlled by SCCE and the performance is measured similar for all first tier suppliers. The performance is based on a date set by SCCE. Therefore, it might be considered as unfair as the factories do not have much influence on the finish date. For the research, it was assumed that this finish date is acceptable and that the factories should be able to produce and finish the orders in time. Similarly, the performance of the second tier suppliers is based on the request date, even if the standard lead-time is longer than 4 weeks. Factories can only finish their production in time if the second tier suppliers deliver material in time. Therefore, this way of measuring is considered to be fair.

Chapter 5: Assumptions conceptual inventory model

A number of assumptions are made for the three conceptual models:

 A two-echelon distribution system with a service level constraint, that considers a single product at a single supplier and independency between the products is assumed;

 The model assumes periodic reviews;

 For each inventory model a base-stock model is assumed;

 It is assumed that demand is known for the first four weeks, and that there is stochastic demand for all other weeks;

 Lead-time between suppliers and factories is assumed to be deterministic;

 It is assumed that suppliers deliver the orders in full, and that there is a first-come-first-serve filling strategy at the supplier, as this is commonly used in literature (Topan, 2014);

 The holding costs for the factories and second tier suppliers are estimated based on a holding cost percentage found in literature;

 It is assumed that the target service level for each factory is 98%, assuming 100% reliability of the suppliers. This service level is equal to the critical ratio, being the probability that the order can be picked from inventory, and is used to compute the base-stock level of each stock point;

 The model does not assume batch discounts;

 Minimum order quantities and multiple order quantities are considered.

Moreover, the input parameters are defined in order to optimize the decision variables for each of the models:

 Expected demand;

 Lead-time;

 Holding costs of the factories;

 Holding costs of the second tier supplier;

 Review period;

 Service level.

Chapter 5: SPO Future demand quantities and characteristics are deterministic

To determine the expected demand for all purchase items, it was assumed that the quantities of SPO projects, exits and meters are deterministic. Moreover, the characteristics are decided upon and do not undergoe any changes in the next phases. This assumption is reliable, as the quantities and characteristics are determined in the sales engineering phase, and are agreed upon by the customer.

Chapter 7: Lead-times set by suppliers are deterministic and reliable in the redesign

The standard lead-times set by the second tier suppliers are assumed to be deterministic in the redesign.

Chapter 7: Lead-time to America is equal to that of Europe

In the inventory model, the standard lead-time within Europe is assumed for all factories, as the PI demand has to be allocated to three European factories and to one American factory and there is currently no information available about the part of the demand that should be allocated to VIA, nor about the exact extra lead-time required for the shipment.

Chapter 8: Suppliers are uncapacitated

When computing the base-stock levels for the SKU, it was assumed that the suppliers are uncapacitated and that there are no upper boundaries in the production of the second tier suppliers.

Chapter 8: Periodic review, base-stock (R,S) inventory model

A base-stock policy is assumed for the inventory policy. This is an appropriate assumption as the periodic reviews result from the SPO Future demand, which is updated on a weekly basis and used as input for the inventory model. Moreover, newsboy model uses a type 1 service level constaint to compute optimal order quantities. This optimal order quantity is used as input for the base-stock level. The service level is equal to the Delivery to request service computed for suppliers and therefore applicable at the business. The demand at the factories is known in the first four weeks and unknown for the weeks later on. In the inventory model, the standard time is 5 weeks. This can deviate from reality, as lead-times can be up to 8 weeks. However, one of the items considered in the model does have a standard lead-time of 5 weeks. Holding costs are computed over on hand inventory. This is different from reality, as Vanderlande computes holding costs over all material received and not just over its on hand inventory.

Material is not perishable, and can be held across multiple periods and unfilled demand is in reality back-ordered. Therefore, these are also realistic assumptions.

Chapter 8: Stochastic lead-time in the inventory model

Two of the four models considered in Chapter 8 consider stochastic lead-time, as the standard lead-time given by a supplier can deviate from the actual lead-time. The cause for the deviation is unknown.

Appendix D. Dash explanation of an SPO sub-component

The sub-component is part of a specific sub-component group. This can, for example, be 0L8569. This number shows on a higher level a number a characteristics of the group. The sub-component can be specified with a second number: the dash-number. The dash consists of 5 numbers that consists information about other characteristics. The dash explanation of 0L8569 is illustrated in the figure below.

67 Appendix E. Factory Processes

Appendix F. Pie diagram with causes

30%

35% 0%

13%

14%

Average causes per week

Excessive workload Errors due to IS

Defect materials/Quality Defect machines Purchase parts overdue Human mistake Unknown

68 Appendix G. Performance SPO suppliers

Graph 8, Delivery to request, second tier suppliers SPO VIM

Delivery to request, SPO second tier suppliers VIS

Orderlines Delivery to request

Delivery to request, SPO second tier suppliers VIS

Orderlines Delivery to request Graph 9, Delivery to request, second tier suppliers SPO VIS

69 Appendix H. Components and characteristics

Figure 28, Variants of component (B)

Figure 29, Variants of component (C)

Figure 30, Variants of component (F)

71 Appendix I. Components and ratios

Historic ratios

The historic data involve all produced SPO sub-components from December 2015 until November 2016.

The following computation was used for each component:

𝑎_𝑦^𝑥: 𝑜𝑟𝑑𝑒𝑟 𝑞𝑢𝑎𝑛𝑡𝑖𝑡𝑦 𝑖𝑛 2015 − 2016 𝑝𝑒𝑟 𝑣𝑎𝑟𝑖𝑎𝑛𝑡𝑠 𝑦 𝑜𝑓 𝑠𝑢𝑏𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡 𝑥

The analysed data of the SPO future demand contain projects from December 2016 until June 2017. The computation is done differently as not all characteristics are defined, resulting in groups of variants. The computations for each component:

This results in the ratios displayed in Table 16, Table 17, and Table 18.

Table 16, Ratios of variants of component (B)

Component B Ratio

Table 17, Ratios of variants of component (C)

Component C Ratio

Variances Forecast Historic

Switchframes dual 20 missing pin detection solid merge block 2% 1%

molded merge block 0%

no missing pin detection solid merge block 0%

molded merge block 1%

30 missing pin detection solid merge block 40% 1%

molded merge block 4%

no missing pin detection solid merge block 4%

molded merge block 30%

single 20 left missing pin detection solid merge block 52% 3%

molded merge block 5%

no missing pin detection solid merge block 11%

molded merge block 8%

right missing pin detection solid merge block 2%

molded merge block 5%

no missing pin detection solid merge block 10%

molded merge block 8%

30 left missing pin detection solid merge block 6% 0%

molded merge block 1%

no missing pin detection solid merge block 0%

molded merge block 2%

right missing pin detection solid merge block 0%

molded merge block 1%

Table 18, Ratios of variants of component (F)

Component F Ratio

Variances Forecast Historic

Carriers plastic 20 900 mm 0% 1%

1000 mm 0% 6%

1100 mm 19% 0%

1200 mm 0% 1%

1300 mm 0% 0%

1400 mm 27% 20%

30 900 mm 0% 0%

1000 mm 0% 2%

1100 mm 2% 3%

1200 mm 0% 2%

1300 mm 1% 6%

1400 mm 0% 0%

steel 20 900 mm 0% 0%

1000 mm 0% 2%

1100 mm 0% 0%

1200 mm 0% 0%

1300 mm 1% 0%

1400 mm 40% 40%

30 900 mm 0% 0%

1000 mm 1% 2%

1100 mm 4% 9%

1200 mm 0% 0%

1300 mm 1% 2%

1400 mm 4% 4%

74 Appendix J. Conclusion purchase item analysis

Table 19, Overview of PI analysis

Component Lead-time Unknown characteristics Item Number B C F Production Total

0P4380-01300 x 7 8

0P4380-01400 x 7 8

0L2352-21105 x 3 4

006002-17013 x 1 2

006002-17011 x 2 3

77 Appendix K. Historic demand at sub-component level

Graph 10, Historic demand of component (C) 0

10 20 30 40 50 60 70 80

49 50 51 52 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48

Quantity

Week

Historic demand 2015-2016 at sub-component level, component (C)

0L8561 0L8562 0L8563 0L8564 0L8565 0L8566 0L8567 0L8568 0L8569 0L8570 0L8571 0L8572

78 Appendix L. Occurrence of purchase items in the variants

Table 20, Variants and items of component (B)

Variants, Component B

Pre-sort Shoe merge

900 mm 1000 mm 1100 mm 1200 mm 1300 mm 1400 mm left right

Item left right left right left right left right left right left right molded solid molded solid Lead-time

1 0 1 0 1 0 1 0 1 0 1 0 1 1 0 0 0 5

2 0 1 0 1 0 1 0 1 0 1 0 1 0 0 0 0 4

3 0 1 0 1 0 1 0 1 0 1 0 1 0 0 0 0 4

4 0 1 0 1 0 1 0 1 0 1 0 1 0 0 0 0 4

5 1 0 1 0 1 0 1 0 1 0 1 0 1 0 0 0 4

6 1 0 1 0 1 0 1 0 1 0 1 0 0 0 0 0 4

7 1 0 1 0 1 0 1 0 1 0 1 0 0 0 0 0 4

8 1 0 1 0 1 0 1 0 1 0 1 0 0 0 0 0 4

Ratio 0% 0% 4% 3% 0% 6% 2% 2% 5% 7% 2% 17% 14% 12% 25% 1% 100%

Table 21, Variants and items of component (C)

Dual Single

Appendix M. Fitting procedure and validation of the simulation 1. Fitting procedure

The probability distribution function of the demand in components is approximated by the fitting procedure of Adan et al. (1995), as there is only limited data availalble. First, the mean (𝜇) and standard devaition (𝜎) of the demand over the available weeks is computed, next the mean and standard deviation are used fo fit the theoretical discrete probability function with:

𝑎 = 𝜎²

𝜇 − 1 𝜇

Adan et al. (1995) developed a fitting procedure that considers four discrete probability functions: the geometric, negative binomial, Poisson, and binomial distribution.

 The binomial distribution is chosen when −1 < a < 0;

 The Poisson distribution is chosen when a = 0;

 The negative binomial distribution is chosen when 0 < a < 1;

 The geometric distribution is chosen when a ≥ 1.

The results of the fitting procedure are illustrated in Table 22. As both avalues are smaller than 0.1 or -0.1, it is chosen to use the Poisson distribution, with mean 𝜆. A characteristic of the Poisson distribution is that events occur independently of the time since the last event. This is true for Vanderlande due to the project environment.

Table 22, Results fitting procedure (Adan et al., 1995)

𝝁 𝝈 𝒂

Projects 2,70 1,71 -0,086

Exits 64,43 43,88 0,052

2. Validation of the warm-up period of model 1, item 1

Law (2007) reported values should be the averages of at least 10 replications and each replication should have a warming up phase of at least 500 weeks. These 500 weeks are followed up by 700 weeks of which the statistics can be recorded. Graph 10 shows that after 500 weeks of warm-up period the performance level remains nearly the same.

Graph 11, Warm-up period

3. Confidence interval bij Law (2007)

Law (2007) determined a method to obtain a confidence interval for simulation outputs, given the number of iterations in the simulation. In each model, 10 repetitions of each 1200 iterations are chosen,

Performance up to week 500 (warm-up period)

resulting in 12,000 iterations per model. By theory of Law (2007) it is checked whether this number of iterations is high enough to obtain an acceptable range when applying a 99.5 % confidence interval.

A point estimate and confidence interval for the mean 𝜇 = 𝐸(𝑋), where X is a random variable, can be

In document Eindhoven University of Technology MASTER Redesign of the control structure and inventory control at Vanderlande Gubbels, P.M. (pagina 73-0)