Eindhoven, September 2011

Improving the Vehicle Fill Rate for Procter & Gamble by Tugce Tali

BSc Industrial Engineering — Bilkent University 2009 Student identity number 0730345

in partial fulfilment of the requirements for the degree of Master of Science in Operations Management and Logistics

Supervisors: Prof. Dr. T. Van Woensel, TU/e, OPAC Dr. Ir. H. Reijers, TU/e, IS D. Jammes, P&G, SNIC

TUE. School of Industrial Engineering. Series Master Theses Operations Management and Logistics

Subject headings: vehicle fill rate, vehicle capacity utilization, outbound transportation, retailer, logistics costs, service, inventory

2

Abstract This thesis aims to improve the Vehicle Fill Rate (VFR) of P&G outbound deliveries in order to achieve a win/win/win solution for the manufacturer/retailer/consumer. For this purpose, processes of P&G outbound deliveries were analyzed and the impacts of increased VFR on service level and logistics costs were assessed. It was found that increasing VFR without any other changes in the supplier-retailer chain has negative impacts in most cases, specifically on the inventory level. Afterwards, it was shown that reducing Minimum Order Quantities (MOQs) while increasing the VFR changes the direction of the impact to positive. This provides an opportunity to increase the VFR while keeping the balance with service level and logistics costs; and even improving them in most cases.

3

Preface This is the master thesis of Tugce Tali, presented on 16 September 2011 at the Eindhoven University of Technology in partial fulfilment of the requirements for the degree of Master of Science in Operations Management and Logistics. I would like to take this opportunity to express my gratitude to all the people who have supported me throughout my thesis project. Firstly, I would like to thank Prof. Tom van Woensel, my first supervisor at TU/e, for his continuous support and patience. His extensive knowledge as well as insightful and humorous approach were very valuable for me. Secondly, I would like to thank Hajo Reijers, my second supervisor at TU/e, for his constructive comments and opinions. Furthermore, I would like to thank David Jammes, my supervisor at P&G. His experience, guidance and professional support helped me a lot throughout my project. I also would like to thank all my colleagues at SNIC for their contributions to my project and for creating such a pleasant work environment. I would like to thank all my friends for their continuous support and encouragement. You made my last two years in the Netherlands and Belgium an adventure. Finally, I would like to express my deepest gratitude to my family for their unconditional love and support. Anneciğim,babacığım ve Buse’ciğim, koşulsuz sevginiz ve desteğiniz için teşekkür ederim. Tugce Tali Eindhoven, the Netherlands, 2011

4

Executive Summary This project was performed in the context of an internship in the Supply Network Innovation Center (SNIC) of Procter & Gamble (P&G) in Brussels, Belgium. Market competition requires firms to respond to customer needs in a quick, cost effective and sustainable manner maintaining acceptable service levels. Hence, companies continuously search ways to improve their operations. Pibernik (2006) states that in recent years companies have been looking for ways to achieve increased supply chain efficiency with the help of higher resource utilization. The vehicle fill rate (VFR) is defined as the ratio of the actual capacity used to the total capacity available in terms of weight and volume (McKinnon, 2010). It is a measure of how efficiently the freight sector is transporting goods with its vehicles. If it can be improved, then the same goods can be carried with fewer vehicle movements thereby leading to reduced congestion, emissions, accidents and other environmental impacts of freight transport (European Environment Agency, 2010). As the European Energy Agency (2006, 2008) reports that the average VFR of trucks in Europe is below 50%, there is an obvious improvement opportunity. The challenge in increasing VFR is to use the full vehicle capacity while keeping the balance with service and cost through the end-to-end supply chain. The opportunity to deliver more products with each vehicle has a positive effect in terms of transportation cost and external costs (i.e. congestion, emissions, accidents and other environmental impacts). On the other hand, this changes the dynamics of the supplier-retailer chain, raising the concern about the potential impact on customer service and/or inventory levels. Considering the abovementioned facts, the research assignment was set as follows: Assess the impacts of increased VFR on the other performance measures of the system; and then, to come up with potential decisions to improve the VFR in outbound transportation in order to achieve a win/win/win solution for the manufacturer/retailer/consumer. The scope of the study is the P&G outbound deliveries to the customers in Western Europe. It contains the P&G DC, the Customer DC and the transportation of goods from the P&G DC to the Customer DC. In the study, firstly, the drivers that lead to low VFR were categorized as product characteristics, vehicle characteristics, supply chain characteristics and health and safety regulations. Among these, supply chain characteristics were determined to be investigated further. P&G supply chain characteristics were analyzed for outbound deliveries in order to observe the behavior of the system while the VFR is increased, identify the risks and negative impacts of increased VFR; and develop insights to mitigate the negative impacts and improve the

5

system. For this purpose, an overall mapping of the outbound delivery elements was provided with their interactions. The analysis revealed the following risks and opportunities: Higher VFR leads to delivery of more goods per shipment. For the total volume, this induces fewer shipments which will cut part of the transportation cost, administration cost and CO2 emissions. On the other hand, a negative impact is anticipated on customer’s inventory level as more goods will be pushed to Customer DC. Meanwhile, reducing the Minimum Order Quantities (MOQs) will be an opportunity to mitigate the negative impacts of increased VFR on customer’s inventory. If the MOQs are reduced, higher number of SKUs can be delivered within the same vehicle each with lower volumes. As a result, a decline can be expected in the total inventory. The target service level (i.e. the product availability at the customer’s site to the downstream orders) was assumed to be constant throughout the study, as service level is thought to be critical for the success of business. The study continued with the redesign. The VFR is increased and MOQs were reduced. Two comparative analyses were performed to quantify the extent of the impacts: 1. Current truck load (CTL) vs. Full truck load (FTL) 2. Current truck load (CTL) vs. Full truck load with lower minimum order quantities (FTL with lower MOQs) Firstly, the VFR was increased; CTL is modified to FTL. The impacts of increased VFR on service level and logistics costs were assessed for several business scenarios (i.e. various shipment frequencies/business volumes on a lane, various numbers of SKUs on a lane, various forecast accuracy levels). Calculations were performed using the ‘Retailer Inventory Model’ (RIM) of P&G - an Excel based tool developed in 2008. Verification and validation were carried out to confirm that the model and the calculations were reliable and represented the actual system. The results revealed that the extent of the impacts on performance measures differed mainly according to the volume/frequency of the lane and the number of SKUs on the lane. It was found that increasing VFR without any other changes in the supplier-retailer chain has negative impacts, specifically on the inventory level. Improving the outbound deliveries from CTL to FTL can only be reasonable for high frequency lanes. Afterwards, MOQs were reduced; CTL is modified to FTL with lower MOQs. It was found that reducing MOQs while increasing the VFR changes the direction of the impact on the inventory level from negative to positive. Besides, improving the outbound deliveries from CTL to FTL with lower MOQs was reasonable for most of the scenarios. The performance measures were improved especially for medium and low frequency lanes. Clearly, increasing VFR meanwhile lowering MOQs is an opportunity to improve the VFR while keeping the balance with service level and logistics costs; and even improving them in most cases. This study contributes to the relevant research area by providing an example of increasing the VFR in outbound transportation. Firstly, it provides a categorization of the outbound delivery 6

elements as well as an overall mapping of them with their interactions. These can be used as a check-list in related studies; and/or can facilitate detecting the indirect dynamics. Afterwards, the study presents the possible outcomes of increasing the VFR. This helps to identify the related risks and opportunities. Furthermore, the study suggests an approach which will mitigate the possible negative impacts of increased VFR and improve the logistics costs further. The study contributes to the company in several ways. Firstly, the categorization of the outbound delivery elements as well as the overall mapping of them with their interactions can provide guidance for people who are not involved in supply chain operations in understanding the part of supply chain dynamics. Moreover, the company can use this study in determining the type of businesses that exhibit the VFR improvement potential. Similarly, when the VFR of a specific lane is considered to be increased, the company can consult the results of the relevant scenarios of this study. Furthermore, displayed mutual benefits of improving CTL to FTL with lower MOQs will motivate both the suppliers and retailers to increase the VFR. It should be noted that this was not a financial study. The focus was on the goods flow. The specific technical solutions, possible investments and the absolute changes in terms of cost were not analyzed.

7

Contents Abstract………………………………………………………………………………………..3 Preface…………………………………………………………………………………………4 Executive Summary…………………………………………………………………………...5 Contents………………………………………………………………………………………..8 List of Figures………………………………………………………………………………..10 List of Tables………………………………………………………………………………....11 List of Abbreviations…………………………………………………………………………12 Chapter 1……………………………………………………………………………………..13 Introduction..........................................................................................................................13 1.1 Research Motivation ......................................................................................................13 1.2 Problem Description and Research Assignment............................................................14 1.3 Company Description ....................................................................................................14 1.3.1 P&G ........................................................................................................................14 1.3.2 SNIC .......................................................................................................................15 1.3.3 Supply Chain Operations ........................................................................................15 1.4 Project Scope and Approach ..........................................................................................16 1.5 Outline............................................................................................................................17 Chapter 2……………………………………………………………………………………..18 Description of the Vehicle Fill Rate ....................................................................................18 2.1 VFR Definition and Measures .......................................................................................18 2.2 VFR in P&G ..................................................................................................................18 2.3 Drivers of low VFR .......................................................................................................19 Chapter 3……………………………………………………………………………………..22 Analysis of the P&G Supply Chain Characteristics ............................................................22 3.1 Outbound Delivery Processes and Overall Mapping.....................................................22 3.2 Structural Mechanisms in Outbound Deliveries ............................................................24 3.3 Performance Measures...................................................................................................29 3.4 Summary of the findings................................................................................................30 Chapter 4……………………………………………………………………………………..32 Redesign: Increasing the Vehicle Fill Rate and Reducing the Minimum Order Quantities 32 4.1 Typology ........................................................................................................................32 4.2 Assumptions...................................................................................................................33 4.3 Methodology ..................................................................................................................34 4.3.1 Sample Data Building .............................................................................................34 4.3.2 Scenario Building....................................................................................................36 8

4.3.3 Retailer Inventory Model (RIM) and Calculations .................................................39 4.3.4 Verification and Validation.....................................................................................42 4.4 Results............................................................................................................................43 CTL vs. FTL ........................................................................................................................44 CTL vs. FTL with lower MOQs ..........................................................................................46 Chapter 5……………………………………………………………………………………..49 Conclusion and Recommendations......................................................................................49 5.1 Conclusions....................................................................................................................49 5.2 Recommendations..........................................................................................................49 References……………………………………………………………………………………51 Appendices…………………………………………………………………………………...53 Appendix 1: An overview of P&G data...............................................................................53 Appendix 2: Outbound delivery elements ...........................................................................56 Appendix 3: Retailer Inventory Model ................................................................................60 Appendix 4: Results.............................................................................................................61

9

List of Figures Figure 1: SNIC in P&G............................................................................................................15 Figure 2: Supplier-retailer chain of P&G.................................................................................16 Figure 3: Current truck load.....................................................................................................19 Figure 4: Drivers of low VFR ..................................................................................................21 Figure 5: Supplier-retailer chain processes..............................................................................22 Figure 6: # SKUs vs Daily Shipment Frequency of investigated lanes ...................................23 Figure 7: Full truck load ..........................................................................................................24 Figure 8: Shipment frequency & time interval in between shipments.....................................25 Figure 9: Dynamics of the outbound delivery system .............................................................31 Figure 10: Visualization of the typology .................................................................................33 Figure 11: Example for sample data building..........................................................................35 Figure 12: CO2 emission levels vs load weight .......................................................................41 Figure 13: Impacts on the customer inventory cost as a result of improving CTL to FTL .....45 Figure 14: Impacts on the customer inventory cost as a result of improving CTL to FTL with lower MOQs.............................................................................................................................47 Figure 15: The lane selected for analysis.................................................................................55 Figure 16: Mapping of the elements in outbound deliveries ...................................................56 Figure 17: RIM user interface..................................................................................................60

10

List of Tables Table 1: VFR measures............................................................................................................18 Table 2: H/M/L shipment frequency........................................................................................24 Table 3: Relationship matrix of structural mechanisms ..........................................................28 Table 4: Selection criteria of structural mechanisms ...............................................................28 Table 5: Performance measures of outbound deliveries ..........................................................29 Table 6: Assumptions of the quantitative calculations ............................................................34 Table 7: Scenarios considered in quantitative calculations .....................................................34 Table 8: Necessary data for quantitative calculations..............................................................35 Table 9: Categorization of the standard deviation of the forecast error ..................................35 Table 10: values according to forecast accuracy scenarios ........................................38 Table 11: Inputs of RIM ..........................................................................................................39 Table 12: Outputs of RIM........................................................................................................40 Table 13: Verification - number of trucks needed to deliver the total volume when CTL was improved to FTL/FTL with lower MOQs................................................................................42 Table 14: Verification - number of trucks needed to deliver the total volume according to different scenarios in CTL case ...............................................................................................43 Table 15: Behaviors of performance measures when CTL is improved to FTL .....................44 Table 16: Number of trucks needed to deliver the total volume according to different scenarios (CTL vs FTL)...........................................................................................................44 Table 17: Behaviors of performance measures when CTL is improved to FTL with lower MOQs.......................................................................................................................................46 Table 18: Number of trucks needed to deliver the total volume according to different scenarios (CTL vs FTL with lower MOQs).............................................................................47 Table 19: VFR in P&G ............................................................................................................53 Table 20: Max allowed legal weight........................................................................................53 Table 21: P&G plant locations in Western Europe (as of 30 June 2010) ................................54 Table 22: VFR (Inbound deliveries vs Outbound deliveries in Western Europe) ..................55 Table 23: Categorization of the elements in outbound deliveries (P&G Operations) .............57 Table 24: Categorization of the elements in outbound deliveries (Customer Operations)......58 Table 25: Categorization of the elements in outbound deliveries (System characteristics) ....59 Table 26: Increase in inventory level for all scenarios when the CTL was improved to FTL 61 Table 27: Example for the dynamics between the cycle stock and safety stock .....................62 Table 28: Increase in inventory level for all scenarios when the CTL was improved to FTL with lower MOQs ....................................................................................................................63

11

List of Abbreviations BPS CF CTL Customer DC D FMCG FTL GBS GBU H L M MDO MOQ OPAC P&G P&G DC R&D RIM SNIC SNO SKU TU/e VFR

12

Business problem solving Corporate function Current truck load Customer distribution center Distributed Fast moving consumer goods Full truck load Global business service Global business unit High Low Medium Market development organization Minimum order quantity Operations planning, accounting, and control Procter & Gamble Procter & Gamble distribution center Research & development Retailer inventory model Supply network innovation center Supply network organization Stock keeping unit Eindhoven University of Technology Vehicle fill rate

Chapter 1 Introduction This report finalizes the MSc Operations Management and Logistics program of the Eindhoven University of Technology (TU/e), in the sub-department of Operations Planning, Accounting and Control (OPAC). The project was performed in the context of an internship in the Supply Network Innovation Center (SNIC) of Procter & Gamble (P&G) in Brussels, Belgium. This chapter explains the motivation of the research, introduces the problem description and the research assignment, provides brief information of the company, and then presents the project scope and approach as well as the report outline.

1.1 Research Motivation Market competition requires firms to respond to customer needs in a quick, cost effective and sustainable manner maintaining acceptable service levels. Hence, companies continuously search ways to improve their operations. Pibernik (2006) states that in recent years companies have been looking for ways to achieve increased supply chain efficiency with the help of higher resource utilization. Transportation is a significant factor in the supply chain economic system as in the EU it accounts for about 5% of the GDP (European Comission, 2011). It links the producer and consumer as well as all the other actors in the chain and provides goods flow. Contrary to common belief, due to inefficiencies transportation costs dominate the logistics costs in a supplier-retailer chain with its 30% share (Van Der Vlist, 2007). In addition, transportation activities result in significant other external costs such as accidents, noise, air pollution, traffic congestion and climate change (INFRAS, 2004). The transportation sector is responsible for the 26% of CO2 emissions (Chapman, 2007). Within the context of rising oil prices and increased concerns about sustainability by consumers, the abovementioned facts indicate a challenge for both the profitability and the public image of a company. Vehicle utilization is a measure of how efficiently the freight sector is transporting goods with its vehicles. If the vehicle utilization can be improved, then the same goods can be carried with fewer vehicle movements. This helps to reduce total freight vehicle traffic, measured as vehicle-km, thereby leading to reduced congestion, emissions, accidents and other environmental impacts of freight transport (European Environment Agency, 2010). As the European Energy Agency (2006, 2008) reports that the average weight utilization of trucks in Europe is below 50%, there is an obvious opportunity to improve the vehicle capacity utilization level (or the vehicle fill rate (VFR)). The challenge in increasing VFR is to use the full vehicle capacity while keeping the balance with service and cost through the end-to-end supply chain. This research focuses on the 13

outbound lanes of P&G - Retailer (i.e. Customer) chains where it is significant to explore the interactions between the supply chain drivers for both P&G and the Customers while the VFR is modified (increased).

1.2 Problem Description and Research Assignment VFR is defined as the ratio of the actual capacity used to the total capacity available in terms of weight and volume (McKinnon, 2010). In P&G, the loading of a vehicle is optimized considering the use of floor area. However, in many cases, there is still space in terms of volume available within the legal weight constraints (Table 19 and Table 20 in Appendix 1). The opportunity to deliver more products with each vehicle has a positive effect in terms of transportation cost and external costs (i.e. congestion, emissions, accidents and other environmental impacts). On the other hand, this changes the dynamics of the supplier-retailer chain, raising the concern about the potential impact on customer service and/or inventory levels. The impacts can be very different for each supplier-retailer chain depending upon the features of it such as the volume of the business in between. These indicate the significance of exploring the influences of increased VFR on the other elements of the supplier-retailer system; in order to achieve a profitable balance with reasonable service levels. Meanwhile, the characteristics of the business depending on the customer being served should also be considered. Regarding the problem description, research assignment is set as follows: Assess the impacts of increased VFR on the other performance measures of the system; and then, to come up with potential decisions to improve the VFR in outbound transportation in order to achieve a win/win/win solution for the manufacturer/retailer/consumer.

1.3 Company Description This thesis is a result of the research carried out at P&G in Brussels, as the industrial partner in this project. This section introduces P&G, including information about the SNIC and the supply chain operations. 1.3.1 P&G P&G, founded in 1837, is one of the leading companies in the fast moving consumer goods (FMCG) industry, with net sales of $78,938 million and net earnings of $12,736 million in 2010 (P&G Annual Report, 2010). The company serves consumers in more than 180 countries with 127,000 employees and more than 300 brands in Health & Well Being, Beauty & Grooming, and Household Care categories. P&G is structured under four main divisions: The Global Business Unit (GBU) operates on product categories, and is responsible for the innovation pipeline, profitability and shareholder returns of the businesses. Market Development Organization (MDO) is the 14

business unit that acts locally. It is in charge of knowing consumers and retailers in each market where P&G competes. Besides, it integrates the innovations flowing from the GBUs into business plans for each market. Global Business Service (GBS) is responsible for providing business support services to the other business units; and finally, Corporate Function (CF) has a support role in every specific department that ensures ongoing functional innovation and capability improvement. 1.3.2 SNIC SNIC is a multi-skilled team of P&G, based in Brussels, organized under the Research and Development (R&D) section of the GBU and Supply Network Operations (SNO) section of the MDO (Figure 1).

Global Business Services and Lean Corporate Functions GBUs - Product oriented R&D

MDOs - Customer oriented

SNIC

SNO

Figure 1: SNIC in P&G

SNIC leverages research in supply network operations in order to build a comprehensive vision of the future; as well as to explore robust improvement opportunities and innovative solutions for P&G based on knowledge and experience. The research areas cover shopper and customer solutions, transportation, warehousing, customization and end-to-end supply chains. 1.3.3 Supply Chain Operations Characteristics of the supply chain in P&G show differences depending on the customer type being served. Deliveries to large retailers comprise diverse product categories; on the other hand, deliveries to smaller customers (i.e. pharmacies, perfume shops etc.) consist of particular products. Hence, the latter requires a more customized supply chain. In this project, supplier-retailer chains were examined, as they constitute the biggest share among all P&G business. The main characteristics of the supplier-retailer chain are explained below (Figure 2). Products are produced in P&G plants, located at the regional (e.g. Belgium) or continental level (e.g. Europe) (Table 21in Appendix 1 provides a list of P&G plants in Western Europe). Production is performed according to forecasts. Then, the products are shipped to P&G Distribution Centers (P&G DCs) with a lead time of 1-5 days by trains and/or trucks; and products are stored there until the customer order arrival.

15

Upon customer order, pallets are prepared and delivered to the Customer Distribution Center (Customer DC) with a lead time of 1-3 days by trucks; and stored there until the store order arrival. A Customer DC is generally served by a single P&G DC, rarely two or three P&G DCs due to category (for ex: personal beauty care and fabric care) difference reasons. P&G is the owner of the goods until the delivery is received by the Customer DC. Although transportation is outsourced, planning is being carried out by P&G.

Figure 2: Supplier-retailer chain of P&G

1.4 Project Scope and Approach The scope of the study is the outbound deliveries to the customers in Western Europe. The inbound deliveries are left out of scope. The VFR of the outbound deliveries is significantly lower than the inbound deliveries (Table 22 in Appendix 1); and hence it is more important to elaborate the VFR in outbound deliveries primarily. The scope contains the P&G DC, the Customer DC and the transportation of goods from the P&G DC to the Customer DC. Each supplier-retailer chain can be considered as a single echelon multi item system; since numerous shipments with multiple stock keeping units (SKUs) are being performed on an outbound trade lane between P&G DC and Customer DC. This project was conducted in collaboration with a company. Thus, it is a Business Problem Solving (BPS) project. A BPS project, as Van Aken et al. (2007) indicates, has as purpose of designing a sound solution along with the realization of performance improvement through planned changes. In BPS settings a specific problem situation within the company provides the starting point of the research. From a relevance perspective the project is client-centered, performance-focused and design oriented. On the other hand, rigor of the project is strengthened by taking the theory as a basis and by justification (Van Aken, 2007). Specifically, the BPS project considers all service and performance measures in understanding the current system, coming up with solution alternatives and identifying improvements of proposed solutions.

16

BPS projects follow classic problem-solving cycle (elaborated as regulative cycle in Van Strien (1997)) considering the setup (Van Aken, 2007). The steps of this cycle are given as: -

Problem definition Analysis and diagnosis Plan of action Intervention Evaluation

This project covered all steps of the regulative cycle except intervention. In order to come up with a clear problem definition, the literature was reviewed and several interviews were performed with the company supervisor. The observations and insights obtained were converted to a problem definition and research assignment. In the analysis and diagnosis step, the processes were investigated in detail to map the relationships that exist between the elements of the system. Then, these were used to design the action plan. Intervention (or implementation) step was left out of scope due to timing reasons. Although a physical intervention was not possible, still, the evaluation step was covered by a quantitative comparison of the situation before and after the redesign.

1.5 Outline The rest of the report is structured as follows: Chapter 2 provides a description of the VFR. Chapter 3 explains the analysis of the P&G supply chain characteristics. Chapter 4 demonstrates the redesign steps of the study and the results. Finally, in Chapter 5 conclusions are presented; limitations of the study and the possible directions for future work are discussed.

17

Chapter 2 Description of the Vehicle Fill Rate Before starting the analysis, it is necessary to understand the VFR concept. For this purpose, this chapter provides a clear definition of the VFR and its measures. Moreover, it presents the VFR in P&G; and explains the drivers of low VFR.

2.1 VFR Definition and Measures The VFR refers to the extent to which a vehicle is loaded compared to its maximum carrying capacity. A 0% fill rate means the vehicle is carrying no loads; and 100% fill rate means the vehicle is travelling with loads bringing the vehicle to its maximum carrying capacity. There exist five measures of VFR (Table 1) in the literature (McKinnon, 2010). Table 1: VFR measures

Deck-area coverage (i.e. Floor fill) Weight-based loading factor (i.e. Weight fill) Volumetric loading factor (i.e. Cube fill) Level of empty running Tonne-km loading factor

Net floor area covered by load/ Floor area per vehicle type (In P&G: Pallets loaded to a truck/ Total available pallet spots) Net weight of load (excluding pallets/gaps/fillers)/ Max legal weight per specific lane & vehicle type Net volume of load (excluding pallets/gaps/fillers)/ Volume per vehicle type The proportion of truck-kms run empty Net weight of load (excluding pallets/gaps/fillers)/ Max legal weight per specific lane & vehicle type (This measure allows weight fill to vary during the journey.)

Level of empty running and tonne-km loading factor were not assessed further. The reason was that empty running and varying loads are not a part of direct outbound shipments from the supplier to the customer.

2.2 VFR in P&G P&G keeps track of the floor fill, the weight fill and the cube fill in its outbound shipments (Table 19 in Appendix 1). The weight fill and the cube fill are significantly lower than the floor fill. The reason is, a full pallet spot on the floor is only explained by the occupation of the concerned spot by a pallet; and not necessarily that the spot is occupied till the ceiling and/or it carries its full weight capacity. This study assumes a vehicle type of a standard semitrailer truck, commonly used by P&G. The vehicle is 2.4 m high, 2.45 m width and 13.6 m length; and it has 33 pallet spots on its floor (Figure 3). 18

In line with the current practice, the Current Truck Load (CTL) in this study is 33 full pallet spots in a vehicle each with 1.8 m high pallets (Figure 3). Although 1.8 m high pallets in a 2.4 m high vehicle theoretically means 75% cube fill, in practice actual cube fill is around 50% since 1.8 m is the maximum height of a pallet and not all the pallets are filled up to 1.8 m.

Figure 3: Current truck load

2.3 Drivers of low VFR Although the objective is to exploit the full vehicle capacity, filling the cube and so increasing the VFR is not always possible. The efficiency of an outbound delivery is severely constrained by the requirements of stakeholders. As a result of interviews in P&G the drivers that lead to low VFR were gathered in four categories (Figure 4): Product characteristics, vehicle characteristics, supply chain characteristics and health and safety regulations. The categorization is supported by the findings of literature study wherever it is appropriate. Product characteristics as a driver of low VFR -

P&G has a wide product assortment with differing weights and volumes. Vehicles will reach the maximum allowed legal weight with heavy loads before the cube is filled (i.e. weight-out); or, vehicles will be filled with voluminous loads before reaching the maximum allowed legal weight (i.e. cube-out) (Department for Transport, UK, 2007). Besides, it is inevitable to end up with empty spaces in between stacks when various types of P&G products are loaded to a vehicle.

-

Packaging protects the products from damage in transit (McKinnon, 1999); poor packaging strength may prevent stacking and so using the full vehicle space. Besides, in FMCG industry, packaging is mostly used to draw shopper’s attention; and thus, essential for value creation. However, in many cases, it consumes a significant portion of the vehicle space because of the empty space within the packages and the resulting spaces in between stacks.

Vehicle characteristics as a driver of low VFR -

If the vehicle design does not ensure the stability of the payload or it requires specific equipments in loading/unloading which are not available within the warehouse, up layering or multiple stacking of the goods within the vehicle may not be possible (Department for Transport, UK, 2007). 19

-

The design of the vehicle, specifically the number of the axles it has, plays a decisive role in the legal maximum payload weight.

Supply chain characteristics as a driver of low VFR -

Building new facilities or adding capacity to the current ones requires huge investments; therefore, the capacity of a warehouse remains the same for a long time. Limited storage and loading/unloading area at the customer’s site might lead to inefficient loads and lower VFR causing a physical limitation on the delivery amount.

-

Lead time can be strictly short for some customers; thus, not letting to consolidate the loads in time in order to fill the vehicle.

-

Customers release orders on an ‘as required’ basis for the smooth flow of materials through a traditional supply chain; which is likely to result in less than full truck load (FTL) consignments (Disney et al., 2003; McKinnon, 1999). In case of high uncertainty of demand, the vehicle utilization deteriorates further.

-

Economies of scale reasons force an agreed minimum order quantity for each of the SKUs between the supplier and the customer. This may influence VFR negatively in two ways: Firstly, each resulting order quantity possesses the unfit risk to a partly loaded vehicle, which might cause lower VFR eventually; and secondly, partial orders that will fill the empty space in the vehicle cannot be released.

Health and safety regulations as a driver of low VFR -

Equipment-related regulations may limit the payload due to stability reasons during loading/unloading and/or transportation (Department for Transport, UK, 2007).

-

Traffic regulations allow a defined maximum payload weight limit according to the characteristics of the vehicle.

Although there are several drivers of low VFR, it was not possible to analyze them all within the scope of this project. Product-related drivers of low VFR are under the control of product design and packaging design departments. Vehicle-related drivers of low VFR can be handled by assessing the advantages and disadvantages of various vehicle types; and then by investing further in the vehicles if it is worthwhile. Health and safety-related limitations require the intervention of legal authorities or extra investments on the loading/unloading equipment. If supply chainrelated drivers of low VFR are considered, customers are in charge of their storage capacities; and the lead time is a result of the supply chain process design.

20

Figure 4: Drivers of low VFR

Customer ordering behavior and the minimum order quantities (MOQs) are linked to each other. Any changes applied to both do not require any further investment on the equipment, the intervention of the other parties and a supply chain process redesign. Thus, they were contemplated to be a reasonable starting point to the analysis and determined to be investigated further.

21

Chapter 3 Analysis of the P&G Supply Chain Characteristics This chapter analyzes the P&G supply chain characteristics for outbound deliveries in order to observe the behavior of the system while the VFR is increased, identify the risks and negative impacts of increased VFR; and detect the ‘customer ordering behavior and MOQs’related redesign options to mitigate the negative impacts and improve the system. This chapter provides four elements: A short overview of the end-to-end supply chain processes followed by an overall mapping of the outbound delivery elements with their interactions; a list of supply chain mechanisms which might appear while the VFR driver is modified; the performance measures of the outbound deliveries identified in the analysis; and finally the summary of the findings which also explains the direction of the following steps.

3.1 Outbound Delivery Processes and Overall Mapping P&G

Plant

Warehousing

Inbound transportation

P&G DC Warehousing

Outbound transportation

Warehousing

Pallet loading

Vehicle unloading

Vehicle unloading

Vehicle loading

Stocking

Stocking

Order accept/deliver

Order accept/deliver

Order picking

Order picking

Pallet loading

Pallet loading

Vehicle loading

Vehicle loading

Customer

Consumer

Cust. DC

Store

Ordering

Transportation

Vehicle unloading

Stocking

Figure 5: Supplier-retailer chain processes

The end-to-end supplier-retailer chain has several processes and stakeholders. The goods move along the production plant, P&G DC, Customer DC and store before it reaches to the shopper (Figure 2). Figure 5 provides a list of the end-to-end supplier-retailer chain processes 22

and highlights the ones that are directly related to the outbound deliveries in line with the scope of this study. A sample outbound trade lane was considered in identifying the outbound delivery processes. To select that, data of 13 weeks of deliveries from P&G DCs to the Customer DCs in France were investigated (i.e. 29 lanes). The data revealed that the shipment frequencies and the number of SKUs of the lanes cover a wide range (Figure 6). One of these lanes was selected for further analysis (Figure 15 in Appendix 1). The selected lane is a good representative of an outbound trade lane: It has a regular delivery cycle; it does not possess any cross-docking operations; and the number of SKUs of the lane is at the average (i.e. 164). The shipment frequency of the lane (i.e. 1.11 shipments/day) is above the average; and this was an opportunity to derive different shipment frequency scenarios in the following phases of the study.

Figure 6: # SKUs vs Daily Shipment Frequency of investigated lanes

Deliveries on the selected lane are order driven. Demand occurs to the Customer DC and it is supplied from stock. The customer checks its inventory levels on a daily basis and releases orders if necessary. At the P&G DC orders are prepared, loaded to pallets and vehicles (i.e. trucks); and then shipped to the Customer DC. The lead time of this process is 3 days on the lane considered. Normally it depends on the specific lane; and it is mainly between 1-3 days. Orders are unloaded at the Customer DC; and then stocked until a store order arrives. The lane has a given volume of business (shipments) which defines a given frequency of deliveries. Even if the profile of deliveries is not flat, we would consider here for simplicity that there is a standard delivery frequency. For that very same trade lane, a change in the VFR will have consequences on the profile of deliveries, and potentially on the frequency. Figure 16 in Appendix 2 is a mapping of the elements involved in the outbound deliveries. Table 23, Table 24 and Table 25 in Appendix 2 categorizes decision, dependent and independent variables as well as performance measures of the outbound delivery elements

23

(i.e. mapped in Figure 16 in Appendix 2) under the related processes; and provide a definition of the elements where necessary.

3.2 Structural Mechanisms in Outbound Deliveries This section clarifies the complex structural supply chain mechanisms in outbound deliveries which will happen when the VFR is increased. To analyze the increased VFR, a full truck load (FTL) was assumed for convenience. The FTL has 33 full pallet spots in a vehicle each with 2.4 m high pallets (Figure 7). The logic behind is that the vehicle is filled basically by up layering the pallets in order not to waste the empty space on top of them without any other changes. Similar to the CTL case, although 2.4 m high pallets in a 2.4 m high vehicle theoretically means 100% cube fill, in practice cube fill is expected to be between 60-80% (Source: interviews within the company) since 2.4 m is the maximum level and not all the pallets could be filled up to 2.4 m.

Figure 7: Full truck load

Increasing the VFR leads to several structural changes in the system. These structural mechanisms as a result of increased VFR are separated and listed to be able to understand when and how they will happen and impact the output measures. Business volume & increased VFR Increased VFR leads to delivery of more goods per shipment. For the total volume, this induces fewer shipments (trucks). If the theoretical impact of improving the VFR from 50% to 100% is considered, it would reduce the number of trucks needed to deliver the same volume by two. However, the impact on the delivery frequency is not linear. A Customer DC delivered by two trucks per day would have only one delivery per day after the VFR improvement; thus, still a frequency of daily delivery. On the other hand, a Customer DC delivered only once a week would then be delivered every two weeks. Therefore, to categorize the impact of the business volume on the delivery frequency a typology of high/medium/low (H/M/L) frequency was proposed (Table 2). Table 2: H/M/L shipment frequency

Shipment frequency H frequency lane M frequency lane L frequency lane 24

Definition From very frequent shipments until around 2 shipments/day From around 2 shipments/day until around 2 shipments/week From around 2 shipments/week until very infrequent cases

Note also that, as another consequence, the time interval in between shipments increases for M/L frequency lanes (Figure 8).

Figure 8: Shipment frequency & time interval in between shipments

Increased delivery time interval The increased time interval in between shipments might prompt several impacts on the supplier-retailer chain especially for M/L frequency lanes: -

Fewer replenishments each with more goods and longer cycle times drive higher order quantities and as a consequence higher cycle stock levels. Average cycle stock level = Q / 2 (Silver et al., 1998)

(Eq 1)

Q: Order quantity in units -

Customers fill the orders of stores from their stock in between replenishments; and hence they try to foresee and stock their needs in advance. As the likelihood of unexpected events is higher in a longer period of time, the forecast accuracy is mostly decreasing when covering a longer forecast horizon to build the orders (Makridakis and Hibon, 2000; Van Der Vorst and Beulens, 2002; Van Der Vorst et al., 1998) Reacting to lower forecast accuracy requires to buffer against uncertainties and drives higher safety stock levels (Eq 2). Failing to react to lower forecast accuracy may cause problems in product availability. Safety stock level = k *

(Silver et al., 1998)

(Eq 2)

k: Safety factor : Standard deviation of errors of forecasts over a replenishment lead time in units

25

-

In a longer time period the likelihood of date constrained events such as promotions, which might force deliveries in between the regular delivery frequency, are also higher.

Urgent deliveries Urgent deliveries are the quickest expediency to handle with unexpected incidents particularly for M/L frequency lanes; while they are more expensive and have low efficiency in terms of VFR. Therefore, considering the advantages they bring in terms of service and the disadvantages in terms of cost, it is a challenge to allow them. If urgent deliveries are considered to be allowed, the consequence of the above mechanism (i.e. increased delivery time interval) will drive to more urgent (or partial) deliveries to mitigate the risk of product unavailability issues at the Customer DC. Delivery timeliness Delivery timeliness is essential for the smooth operation of the system. Poor delivery timeliness might prevent having the right product at the right time and place; and create product availability concerns. While the number of trucks delivered to the Customer DC will be reduced, the volume shipped remains the same (assumption of neutrality on business). From a statistic point of view, the delivery timeliness remains the same per truck: While one truck –which can be on time or late- will deliver more products, there will be proportionally less trucks delivering the same total amount of product. So the delivery timeliness at the SKU level will remain the same. Fewer trucks will positively impact the traffic density (considering at the industry level). Besides, fewer trucks will moderate the traffic and lighten the reception workload at the Customer DC by reducing the administrative work needed; and this might enable to discuss additional opportunities with customer such as more flexible delivery time windows. Taking these into account, increased VFR might provide opportunities to improve the delivery timeliness. MOQs Customer orders are released based on MOQs per SKU. MOQ can be defined as the minimum order amount per SKU in terms of pallets/layers/cases. For instance, if the MOQ of product A is agreed to be a full pallet, the customer cannot order less than a full pallet for product A. Increased VFR leads to delivering more goods per shipment. The additional load in a vehicle as a result of increased VFR can contain either the same SKUs the vehicle already contains or different ones. In order not to raise the inventory per SKU, it is preferable to load different SKUs. Moreover, if the MOQs of each SKU are diminished, we can expect further mitigation on the impacts. Then, the customer will have the opportunity to order and replenish less per 26

SKU. This mechanism would be limited in case of M/L frequency lanes due to the increase in the time interval in between shipments. As cycle time increases and forecast accuracy declines, the customer will have a difficulty in determining the right ordering levels accurately and will not leverage the capability of a lower MOQ. Customer inventory Customer inventory can be examined in two parts: Cycle stock which results from economies of transport; and safety stock which provides a buffer against demand uncertainties. Several mechanisms will play positively and negatively on the inventory level as mentioned above. Fewer replenishments each with more goods and longer cycle times drive higher cycle stock levels (Eq 1). On the other hand, lower MOQ might lead to lower order quantities and so lower cycle stock levels. Safety stock level depends on targeted service level and the uncertainties in the supply chain (Eq 2); and an increase in forecast horizon and resulting forecast accuracy issues will drive higher safety stock levels (Makridakis and Hibon, 2000; Van Der Vorst and Beulens, 2002; Van Der Vorst et al., 1998). The actual inventory level will need to be analyzed. It appears clearly that the frequency of the lane (H/M/L) is a key driver here. Handling Handling costs are mainly measured by the time spent to perform the handling processes. Therefore, they constitute higher proportion in markets (or countries) where wages per hour are substantially high. Long distances or infrequent deliveries help to mitigate the impacts of handling costs; hence, it might be profitable to reduce handling particularly in H frequency lanes and where the customer is nearby. Fewer shipments and so less number of vehicles may have a reducing impact on handling both at the P&G DC and the Customer DC unless loading/unloading practice is altered in a way to require more time and effort (for ex: double decking). Nevertheless, loading more goods and wider assortment to a vehicle may have an increasing impact on handling costs per shipment. Furthermore, if the contents of pallets get more complex with smaller delivery quantities per SKU, the handling performed in order picking and pallet loading at the P&G DC as well as pallet splitting at the Customer DC will rise. The separation and listing of structural supply chain mechanisms was significant as the following steps of the project aimed at analyzing and understanding some of them quantitatively. Still, interactions were detected between the several mechanisms as presented in Table 3 (A ‘+’ sign indicates a direct interaction and a ‘-’ sign indicates no direct interaction between the involved mechanisms.); and there might be joint effects that bring opportunities or risks depending upon each specific outbound lane.

27

Table 3: Relationship matrix of structural mechanisms

Structural mechanisms

Business volume & increased VFR Increased delivery time interval Urgent deliveries Delivery timeliness MOQs Customer inventory Handling

Business volume & increased VFR

+ + + + + +

Increased delivery time interval

Urgent deliveries

Delivery timeliness

+ + + + -

+ + -

+ + -

MOQs

+ + + +

Customer inventory

Handling

+ + + + -

+ + +

When all the structural mechanisms were considered, ‘Business volume & increased VFR’, ‘Increased delivery time interval’, ‘MOQs’ and ‘Customer inventory’ were decided to study further. Table 4 demonstrates the criteria used in selection. Table 4: Selection criteria of structural mechanisms

Structural mechanisms Business volume & increased VFR Increased delivery time interval Urgent deliveries Delivery timeliness MOQs Customer inventory Handling

Data availability + + + + +

Independence from technical solution/investment + + + + + + -

‘Urgent deliveries’ and ‘Delivery timeliness’ were left out of the scope of this study because of data unavailability. There was no differentiation between the regular and urgent deliveries for the analyzed outbound trade lane; and thus no ‘Urgent deliveries’ data was available. Indepth analysis of ‘Delivery timeliness’ needs traffic-related data which is not available within the company. ‘Handling’ was left out of scope, because comprehensive analysis of ‘Handling’ needs picking out a specific technical solution (for ex: double stacking and new handling equipment) by analyzing the handling processes, equipments and investment opportunities. All the other mechanisms are based on goods flow and independent from technical solution/investment. ‘Urgent deliveries’, ‘Delivery timeliness’ and ‘Handling’ also found to possess less interaction with the rest of the structural mechanisms (Table 3). ‘Business volume & increased VFR’, ‘Increased delivery time interval’, ‘MOQs’ and ‘Customer inventory’ are linked to each other with interactions (Table 3); they are 28

considerably in line with what was determined to be investigated further in Section 2.3 (i.e. customer ordering behavior and MOQs); and they provide sufficient evidence of the significant performance measures of outbound deliveries (Section 3.3).

3.3 Performance Measures Logistics performance measures, ideally, should capture all stakeholders, measure and compare the current and future performance, include all related activities along the process, recognize and allow for trade-offs between the different dimensions of performance, be understandable as well as provide a guide for the action to be taken (Caplice and Sheffi, 1995). Performance measures of this project were identified based on this statement and the careful consideration of outbound delivery elements. Table 5 provides a list of all the performance measures considered and highlights the ones that were assessed in this project. Table 5: Performance measures of outbound deliveries

Category Service

Cost

External cost

Performance measure P&G on time delivery performance P&G product availability Customer product availability P&G transportation cost P&G administration cost P&G inventory cost P&G handling cost Customer administration cost Customer inventory cost Customer handling cost CO2 emissions Noise Pollution Traffic density

Three categories were defined for the performance measures: Service, cost and external cost. For both P&G and customer it is significant to keep the balance with service and cost in their operations as well as to keep the impact of the operations to their environment at minimum. Service as a performance measure Improvement is anticipated in P&G on time delivery performance upon increasing the VFR as explained in the structural mechanism ‘Delivery timeliness’; however, in-depth analysis requires more traffic data which is not available within the company. Therefore, it was not assessed in this project. P&G product availability (i.e. fill rate) measures the fraction of customer orders satisfied from inventory. To produce efficiently manufacturers have to produce in batches; and stock at manufacturer’s warehouse is mainly the result of producing goods in certain batches (Van 29

Der Vlist, 2007). P&G product availability is influenced mostly by the upstream operations and not by the modifications in VFR. Thus, it was not assessed in this project. Customer product availability (i.e. fill rate) measures the fraction of store orders satisfied from inventory. It is not merely providing a product where and when it is wanted; but meeting the customer’s needs in a manner which makes the customer want to do business again and again in preference to any other company (Livingstone, 1992). P&G aims to deliver superior quality service to its customers and the consumers as a part of its purpose (www.pg.com). Hence, the target customer product availability was assumed to be constant (i.e. 99%) throughout the study. Cost as a performance measure Daganzo and Newell (1993) categorized logistics costs as follows: transportation (freight rates), inventory (the opportunity cost and loss of value associated with items in possession), storage (the cost of the physical facilities needed to hold stationary items) and handling (the cost of loading and unloading, storing and retrieving, and otherwise manipulating the items in question). Administration cost of outbound delivery processes was also included to the list after the analysis of the processes in the company. P&G transportation, P&G administration, customer administration and customer inventory costs were assessed in this project. Storage costs were included in inventory costs. P&G inventory cost was not considered since that was found to be not directly related to the VFR modification. P&G and customer handling costs were not considered because the impacts of the VFR modification on these measures depend on a specific technical solution (for ex: double stacking and new handling equipment) to be selected. External cost as a performance measure Transportation activities result in significant other external costs such as accidents, noise, air pollution, traffic congestion and climate change (INFRAS, 2004). CO2 emissions were assessed in this project as they are quantifiable. To evaluate noise, pollution and traffic density more data is required which is not available within the company; still improvement is anticipated in all these measures upon increasing the VFR.

3.4 Summary of the findings The previous sections provided an understanding of the P&G outbound delivery system such that the VFR is increased. Higher VFR leads to delivery of more goods per shipment. For the total volume, this induces fewer shipments which will cut part of the transportation cost, administration cost and CO2 emissions. On the other hand, a negative impact is anticipated on customer’s inventory level as more goods will be pushed to Customer DC. Reducing the MOQs will be an opportunity to mitigate the negative impacts of increased VFR on customer’s inventory. If the MOQs are reduced, higher number of SKUs can be delivered within the same vehicle each with lower volumes. As a result, a decline can be expected in the total inventory. 30

The target service level (i.e. the product availability at the customer’s site to the downstream orders) was assumed to be constant throughout the study, as service level is thought to be critical for the success of business. Figure 9 summarizes the explained dynamics.

Figure 9: Dynamics of the outbound delivery system

In the light of the identified dynamics, the following steps of the project aimed and redesigning the P&G outbound deliveries. Firstly, the VFR was increased; current truck load is modified to full truck load. Afterwards, MOQs were reduced; full truck load is modified to full truck load with lower MOQs.

31

Chapter 4 Redesign: Increasing the Vehicle Fill Rate and Reducing the Minimum Order Quantities This chapter presents two comparative analyses: 3. Current truck load (CTL) vs. Full truck load (FTL) 4. Current truck load (CTL) vs. Full truck load with lower minimum order quantities (FTL with lower MOQs) Firstly, the VFR was increased; current truck load is modified to full truck load. The comparison CTL vs FTL provides a quantitative understanding of system behavior and reveals the changes in performance measures. Afterwards, MOQs were reduced; current truck load is modified to full truck load with lower MOQs. The comparison CTL vs FTL with lower MOQs also provides a quantitative understanding of system behavior and reveals the changes in performance measures. Besides, it enables to explain the extent of improvement in P&G outbound deliveries when the VFR is increased and MOQs are reduced. The typology, the assumptions and the methodology of the analyses are clarified below. Finally, results are presented.

4.1 Typology CTL The CTL in this study is 33 full pallet spots in a vehicle each with 1.8 m high pallets (Figure 3 and Figure 10; further detail in Section 2.2). FTL The FTL has 33 full pallet spots in a vehicle each with 2.4 m high pallets (Figure 7 and Figure 10; further detail in Section 3.2). FTL with lower MOQs FTL with lower MOQs has 33 full pallet spots in a vehicle each with two pallets of 1.2 m high. As a consequence, the vehicle contains 66 pallets which are double-stacked (i.e. for the convenience of the calculations) and which are smaller than the current practice (Figure 10). This approach leads to lower MOQs for the SKUs those with an MOQ of one full pallet. The main idea of this approach is to have lower MOQ levels instead of double stacking. Hence, double stacking should not be interpreted as a solution proposal; instead it should be regarded as a way to visualize the lower MOQs.

32

Figure 10: Visualization of the typology

4.2 Assumptions A sample outbound trade lane (Figure 15) was considered in calculations. A vehicle type of a standard semitrailer truck, commonly used by P&G was assumed. The vehicle is 2.4 m high, 2.45 m width and 13.6 m length; and it has 33 pallet spots on its floor. VFR was targeted to be modified from CTL (i.e. around 50%) to FTL (i.e. around 60-80%). The analysis was based on goods flow and independent from any technical solution (for ex: double stacking and new handling equipment) and investments. Retailer inventory model (RIM) of P&G was used as a tool in calculations (Further explanation in Section 0). Table 6 lists the assumptions used in quantitative calculations. 13 weeks of data was available and that whole period was considered in calculations. The lead time and review period were taken as 3 and 1 day(s) respectively, in line with the practice. The target service level (i.e. the product availability at the customer’s site to the downstream orders) was assumed to be constant and 99% throughout the study.

33

Table 6: Assumptions of the quantitative calculations

A month: A week: Total analysis period: Lead time of the shipments: Review period: Target service level:

4 weeks 7 days 13 weeks 3 days 1 day 99%

In order to develop insights on different kind of businesses several scenarios were examined. Table 7 lists the scenarios considered in quantitative calculations (Further explanation in Section 0). Table 7: Scenarios considered in quantitative calculations

Number of the SKUs: Shipment frequency: Forecast accuracy:

25, 53, 159, 265, 371, 477, 583 3/day (i.e. 3 shipments/ day), 2/day, 1/day, 4/week, 3/week, 2/week, 1/week, 3/month, 2/month, 1/month distributed, high, medium, low

4.3 Methodology This section explains how the sample data that were used in calculations and the examined scenarios were built; and then introduces the calculations and the model used in the calculations as well as the verification and validation processes. 4.3.1 Sample Data Building P&G has all the necessary data for quantitative analysis except the demand information (Table 8). The outbound shipments of P&G are order driven, which means shipments are performed according to the orders of customers. Customers order according to the demand forecast and their stock levels; and apart from a few cases (for ex: vendor managed inventory), P&G does not have access to the base demand and forecast information of customers. Therefore, shipments are the only way to presume demand information. In depth examination of selected lane data revealed a very high variation of historical shipments (i.e. σ – can be interpreted as standard deviation of forecast error, Eq 3) of some SKUs, which might mean that the data is polluted with promotions, product introductions, product endings etc. and the results obtained by using this data would be distorted. Hence, cleaning or correction was necessary to remove the unusual events and disturbances from the actual data. =

34

(

(

)

)

(Eq 3)

Table 8: Necessary data for quantitative calculations

General Outbound lane level

Total analysis period Lead time of the shipments Review period Target service level Number of SKUs on the lane Shipment frequency on the lane Demand Number of cases per pallet Number of cases per layer

SKU level

Cleaning and correction of data P&G categorizes the standard deviation of the forecast error (i.e. σ ) as in Table 9. Considering this categorization 36 of 164 SKUs on the selected lane which have σ smaller than 1 (i.e. regular products) were selected for the sample data set. SKU 81143837 in Figure 11 is an instance of this selection. Afterwards, 17 more SKUs were detected which have σ smaller than 1 for either the first 6 weeks or the last 6 weeks (i.e. product endings/introductions). For these SKUs the first/last 6 weeks of data were replicated for the last/first 6 weeks; and they were also selected for the sample data set. SKU 83715007 in Figure 11 is an example. The remaining SKUs were cleaned (SKU 81108433 in Figure 11). As a consequence 53 SKUs were obtained for the sample data set. Table 9: Categorization of the standard deviation of the forecast error

20% Best in class 30% Automatic system based ordering 50% Good forecast 70% Default value 100-120% Heavily promotional

Figure 11: Example for sample data building

35

4.3.2 Scenario Building The investigation of 13 weeks of deliveries from P&G DC’s to Customer DC’s in France revealed a wide range of shipment frequencies and number of SKUs for the lanes (Figure 6). This illustrates that the type of business on each lane can have distinct characteristics. Furthermore, the information flow in between the supplier and the retailer will not have the same efficiency level in each of the lanes; and thus the accuracy of the forecast can be different on each lane. Considering these facts, several scenarios were built and examined in order to develop insights on the different kinds of businesses. The impacts of increased VFR were analyzed for: -

Various numbers of SKUs on a lane Various shipment frequencies on a lane Various forecast accuracy levels

Number of SKUs The product assortment on P&G outbound trade lanes vary from little to ample. The base sample data set (i.e. 53 SKUs) was replicated to obtain larger data sets. (i.e. 159, 265, 371, 477 and 583 SKUs) Meanwhile, the total volume of the business on the lane was kept constant by a variable (For ex: A lane which had 53 SKUs and a frequency of 2 shipments/week still had the same shipment frequency when the number of SKUs was increased to 159). Therefore, a higher number of SKUs should not be interpreted necessarily as higher volumes. Additionally, in order to be able to examine very low number of SKUs case, 25 SKUs were selected randomly from the base 53 SKUs. As a result 7 different scenarios were obtained to analyze various numbers of SKUs on a lane. In fact, the main idea was to observe the behavior of the system at high/medium/low (H/M/L) number of SKUs. Differentiation points for the H/M/L number of SKUs were not defined; since it was thought to be more reasonable to interpret the impacts along the lowhigh continuum of number of SKUs. Hence, the numbers presented here should not be interpreted in their absolute values. The reader is suggested to capture the general understanding. Shipment frequency The range of delivery frequency in P&G outbound trade lanes is wide; P&G delivers some of its Customers’ DCs with several trucks per day, while some others receive one truck every two weeks or less. To simulate and observe the situation as we go through this range 10 different scenarios were developed. (i.e. 3/day, 2/day, 1/day, 4/week, 3/week, 2/week, 1/week, 3/month, 2/month, 1/month) Similar to the ‘Number of SKUs’, the main idea was to observe the behavior of the system at H/M/L lane frequencies (Table 2); hence, the frequencies presented here should not be interpreted in their exact values. The reader is suggested to capture the general understanding. 36

Forecast accuracy Even though the disturbances such as promotions and product introductions/endings were attempted to be cleaned from the 53 SKUs of sample data set, the resulting inventory levels with these data were very much above the actual inventory levels of P&G. The reason was found to be the σ (Eq 3) values.

Increased VFR leads to less number of shipments (trucks) and an increased time interval in between shipments. This prompts significant impacts. The forecast accuracy decreases when covering a longer horizon (Makridakis and Hibon, 2000; Van Der Vorst and Beulens, 2002; Van Der Vorst et al., 1998); since the likelihood of unexpected events is higher in a longer period of time. Acknowledging this fact, σ values needed to be arranged according to different scenarios. Distributed (D) forecast accuracy High frequency deliveries - several shipments per day- require a close relationship and efficient information flow between the supplier and retailer. Considering the fact that 0.2 is categorized as best in class in Table 9, it was assigned to σ when the frequency of the lane is very high, specifically when the average time interval in between shipments is less than or equal to one day. The rest of the σ values -for different average time interval in between shipments- were developed by using Eq 4. σ

,

+σ

,

= σ

(Eq 4)

,

Derivation of Eq 4:

Var (X+Y) = Var(X) + Var(Y) + 2 * Cov(X,Y) (Ross, 2000).

(Eq 5)

X: First replenishment time interval Y: Second replenishment time interval X+Y: Time interval which is equal to the sum of first and second replenishment time intervals Var (X): σ Var (Y): σ

Var (X+Y): σ

,

,

,

X and Y are independent random variables; hence Cov(X,Y) = 0 σ

,

+σ

,

= σ

,

(Eq 4)

37

Table 10 lists the σ values. The resulting inventory levels with these σ values were in line with the actual inventory levels of P&G. Thus, this setting can be interpreted as valid for simulating the current forecast accuracy levels. Still, in addition to this, H/M/L forecast accuracy level cases were also derived to examine the other situations. Table 10:

values according to forecast accuracy scenarios

Avg time interval in btw shipments 1 day 2 day 3 day 4 day 5 day 6 day 7 day 8 day 9 day 10 day 11 day 12 day 13 day 14 day 15 day 16 day 17 day 18 day 19 day 20 day 21 day 22 day 23 day 24 day 25 day 26 day 27 day 28 day 29 day 30 day

σ +σ =σ Distributed 0.200 0.283 0.346 0.400 0.447 0.490 0.529 0.566 0.600 0.632 0.663 0.693 0.721 0.748 0.775 0.800 0.825 0.849 0.872 0.894 0.917 0.938 0.959 0.980 1.000 1.020 1.039 1.058 1.077 1.095

Linear increase High Medium. Low 0.200 0.500 0.800 0.210 0.510 0.810 0.221 0.521 0.821 0.231 0.531 0.831 0.241 0.541 0.841 0.252 0.552 0.852 0.262 0.562 0.862 0.272 0.572 0.872 0.283 0.583 0.883 0.293 0.593 0.893 0.303 0.603 0.903 0.314 0.614 0.914 0.324 0.624 0.924 0.334 0.634 0.934 0.345 0.645 0.945 0.355 0.655 0.955 0.366 0.666 0.966 0.376 0.676 0.976 0.386 0.686 0.986 0.397 0.697 0.997 0.407 0.707 1.007 0.417 0.717 1.017 0.428 0.728 1.028 0.438 0.738 1.038 0.448 0.748 1.048 0.459 0.759 1.059 0.469 0.769 1.069 0.479 0.779 1.079 0.490 0.790 1.090 0.500 0.800 1.100

High forecast accuracy To simulate high forecast accuracy within a supplier-retailer chain, values between 0.2 and 0.5 were assigned to σ in a way to increase linearly as the average time interval in between shipments raises (Table 10). Eq 4 could not be used in this scenario in order to be able to keep the σ values strictly low. Medium forecast accuracy

Similar to ‘High forecast accuracy’, to simulate medium forecast accuracy within a supplierretailer chain, values between 0.5 and 0.8 were assigned to σ in a way to increase linearly 38

as the average time interval in between shipments raises (Table 10). Eq 4 could not be used in this scenario in order to be able to keep the σ values in the determined interval. Low forecast accuracy

Similarly, to simulate low forecast accuracy within a supplier-retailer chain, values between 0.8 and 1.1 were assigned to σ in a way to increase linearly as the average time interval in between shipments raises (Table 10). Eq 4 could not be used in this scenario in order to be able to keep the σ values strictly high.

In the end 280 different scenarios were examined for CTL, FTL and FTL with lower MOQs: 10 different shipment frequency scenarios were tested for 7 different scenarios of number of SKUs on a lane; and each of them were simulated for 4 different types of forecast accuracy level. 4.3.3 Retailer Inventory Model (RIM) and Calculations RIM After the sample data set and the different scenarios were developed, calculations were performed using the ‘Retailer Inventory Model’ (RIM) of P&G which is an Excel based tool developed in 2008. RIM considers a single supplier-retailer lane; and it provides outputs according to the descriptive characteristics of the lane as well the attributes of the SKUs on that lane. The tool calculates the stock levels of the customer and its orders. Orders are based on the business need to avoid a stock-out in daily reviews and on additional quantities that are ordered to load the truck at its maximum floor fill capacity. This means that the service level is maintained and the loads are consolidated forward and not backward. Table 11 and Table 12 present the inputs and outputs of the tool respectively. The specific additions/adaptations performed on RIM for this study are presented below. Figure 17 in Appendix 3 presents the user interface of the tool. Table 11: Inputs of RIM

General Outbound lane level

SKU level

Input Variable Total analysis period Number of pallets per truck Lead time of the shipments Review period Target service level Number of layers per pallet Number of cases per layer MSU/period

σ

Unit and Explanation weeks pl/truck days (order to delivery cycle time) days (between DC orders) percentage ly/pl cs/ly cs/week (can be calculated by P&G data; it is a measure of sales used to obtain demand information) (Eq 3)

39

Table 12: Outputs of RIM

Output Variable Total Number of SKUs on the lane Total demand in pallets per period Total demand in pallets per day Number or trucks needed per period to deliver the total volume Average replenishment interval Cycle stock level Safety stock level Total stock level

Unit and Explanation units pl/week pl/day trucks/period days days days days

Adaptations of RIM RIM assesses the truck load by the number of full pallet spots in a vehicle. Hence, it computes the number of trucks needed by assuming a 33 pallets load (each with 1.8 m high) for each vehicle. This structure provides the results for CTL. To be able to simulate FTL and FTL with lower MOQs, the number of layers per pallet for each of the SKUs were rendered as configurable by a variable; so that the pallet height could be adapted according to the examined case: -

For FTL: number of layers per pallet * 2.4 / 1.8; rounded to the closest integer For FTL with lower MOQs: number of layers per pallet * 1.2 / 1.8; rounded to the closest integer

For each of the 10 shipment frequency scenarios, 7 different number of SKUs scenarios were examined. To do so, it was required to keep the total shipped volume constant, while the number of SKUs was changing. Therefore, MSU/period values for each of the SKUs were also rendered as configurable by a variable, to arrange the total volume. Calculations Calculations were performed with the help of RIM to assess the impacts of increased VFR on the performance measures of P&G outbound deliveries. The details of calculations for each of the performance measures are explained below. P&G transportation cost Unit P&G transportation cost is measured in €/kms. Its actual value depends on many factors such as the location of the DCs, oil prices, employee wages etc. Therefore, it is specific for each different lane. The decline in P&G transportation cost can be calculated by multiplying this unit cost by the eliminated number of shipments through increasing VFR. RIM provides the number of trucks needed per period to deliver the total volume; and, eliminated number of shipments per period is the difference of this value between CTL case and FTL/FTL with lower MOQs case.

40

P&G administration cost Increasing VFR eliminates significant amount of paperwork as well as related costs. P&G administration cost is measured in €/order/invoice. The standard industry average is €30 /order/invoice; which means each eliminated shipment leads a €60 of reduction in this measure. Customer administration cost Similar to ‘P&G administration cost’, increasing VFR eliminates significant amount of paperwork as well as related costs of the customer. The Customer DC was assumed to incur analogous administration costs; which means each eliminated shipment leads a €60 of reduction in this measure. Customer inventory cost RIM provides both resulting cycle and safety stock levels. This enables to calculate the total inventory cost and the amount of capital tied to inventory. In addition, allocation of the cycle and safety stock levels gives an opportunity to comprehend the drivers of the present inventory. Unit inventory holding cost can be taken as €10 /spot/month. Capital tied to inventory can be taken as 7-10% per year. CO2 emissions Figure 12 (Bilan Carbone, 2007) presents the CO2 emission levels as a result of energy consumption for heavy duty trucks: The higher the load factor the higher the emission per vehicle (i.e. blue line); however, the higher the load factor, the lower the emission per tonnekm (i.e. red line). Therefore, although a slight increase in CO2 emissions will be observed per shipment due to a heavier payload, the increased VFR and so the reduction of the number of trucks used will definitely reduce the total emissions.

Figure 12: CO2 emission levels vs load weight

41

In P&G, CO2 emissions of an average loaded truck is assumed to be 0.9 kg/km. The improvement in CO2 emissions can be calculated multiplying this value by the eliminated number of shipments. The increase in CO2 emissions due to a heavier payload is assumed to be negligible. The calculations can be performed for any scenario as long as all the input variables are known. In this project, all developed scenarios were simulated using RIM; however the absolute changes in total cost were not calculated. The reason was that some cost measures depend on the specific lane to be selected. Although the absolute changes in cost could not be calculated, the results (presented in Section 4.4) obtained by simulating the scenarios were enough to develop insights. 4.3.4 Verification and Validation This study involves simulating scenarios (Section 0) on an Excel based tool (Section 0) with numerous assumptions (Section 4.2) in order to compare the CTL with alternative situations. This section explains the steps of verification and validation processes that are performed to check whether the analyses provide reliable results. Verification Verification is the process of ensuring that the model behaves in the way it was intended according to the modeling assumptions made (Law and Kelton, 1991). In this study, it was necessary to determine whether the developed scenarios were correctly represented in the model. Higher VFR leads to delivery of more goods per shipment and for the total volume, this induces fewer shipments. The model was expected to provide smaller values for the number of trucks needed to deliver the same volume when CTL was improved to FTL/FTL with lower MOQs. Table 13 shows that it is reliable to verify that higher VFR leads to fewer shipments in the analyses. Table 13: Verification - number of trucks needed to deliver the total volume when CTL was improved to FTL/FTL with lower MOQs

# of trucks needed FTL w. CTL FTL MOQ

#SKUs

Ship. Fr

265

3/day

280

201

188

265

3/week

39

28

26

265

1/month

3

3

2

One of the assumptions in building the scenarios was to keep the total shipped volume constant while the number of SKUs was changing. A variable was added to the model in order to arrange the volume in the described way. Table 14 presents the resulting number of trucks needed to deliver the total volume according to the different scenarios in CTL case. Although there are minor differences in some values, it is reliable to affirm that the volume was kept constant whenever needed in the analyses. 42

Table 14: Verification - number of trucks needed to deliver the total volume according to different scenarios in CTL case

Ship. Fr.

Number SKUs 25

53

159

265

371

477

583

3/day

277

275

278

280

279

278

277

3/week

39

39

39

39

39

39

39

1/month

3

3

3

3

3

3

3

Validation Validation is the process of determining whether a built model is an accurate representation of the actual system. Thus, it requires confirming the outputs of the model with real life knowledge. Although statistical tests can be carried out in validation, a good dose of common sense is also respected (Kelton et. al, 2003). To validate the analyses, firstly, the outputs of the model were compared with the actual statistics. In this study numerous scenarios were analyzed; however, it was not possible to compare the results of the each scenario with the real data. Still, apart from some extreme cases (i.e. low frequency and low forecast accuracy level scenarios) it was observed that resulting inventory levels in CTL case stay within the limits of the actual inventory levels that statistics indicate. The deviated outcomes of the extreme cases were also assumed as valid; because it was predictable to have very high inventory levels in CTL case when the shipments were infrequent and the time interval between replenishments were high as well as when the forecast accuracy level was low. Afterwards, the model’s behavior was tested to validate whether it was parallel to real life expectations (Law and Kelton, 1991). For this purpose, business volume, σ , review period, lead time and target service level were altered respectively. Higher VFR led to fewer shipments and longer replenishment time intervals; as a consequence cycle stock levels were also higher as anticipated. When σ values were altered, the number of shipments stayed constant; because there was no change in business volume. On the other hand, lower σ values kept safety stock levels lower, while higher σ values led to higher safety stock levels in line with the expectations. Increased review period as well as increased lead time caused an increase in safety stock levels; because increasing any of them reduces the flexibility of the goods flow. Lastly, alterations in target service level also induced presumed variations in safety stock level. To sum up, it can be concluded from above explanations that the model and the calculations were reliable and represented the actual system.

4.4 Results This section presents the results of two comparative analyses: 1. CTL vs. FTL 2. CTL vs. FTL with lower MOQs 43

The results for each of the performance measures are explained separately. CTL vs. FTL This part explains the behavior of the performance measures when VFR was increased and the CTL was improved to FTL (Table 15). Table 15: Behaviors of performance measures when CTL is improved to FTL Shipment frequency (C TL)

3/day

SAVING IN TOTAL VOLUME Customer inventory

Cash

P&G admin Customer inventory Customer admin

External Cost

1/day

4/week

3/week

2/week

1/week

CO2 emissions

3/month

MIDDLE FREQUENCY

2/month

1/month

LOW FREQUENCY

no impact variable/other factors medium positive major positive impact impact variable/other factors no impact medium positive major positive impact impact

P&G transport Cost

2/day

HIGH FREQ

major negative impact minor positive impact major negative impact minor positive impact

P&G transportation cost Eliminating the number of trucks required to deliver the total volume led to a positive impact in ‘P&G transportation cost’ in all of the scenarios (Table 15); the cost decreased and savings realized. The extent of the savings depended mainly on the frequency of the lane (i.e. the volume of the business) as well as the distance travelled. Table 16: Number of trucks needed to deliver the total volume according to different scenarios (CTL vs FTL) Number SKUs 25

53

159

265

371

477

583

Ship. Fr.

CTL

FTL

CTL

FTL

CTL

FTL

CTL

FTL

CTL

FTL

CTL

FTL

CTL

FTL

3/day

277

198

275

197

278

200

280

201

279

200

278

200

277

198

2/day

181

129

182

130

181

130

182

130

182

130

181

130

183

131

1/day

91

65

91

65

91

65

91

65

90

64

91

65

91

65

4/week

52

37

52

38

52

37

51

36

52

37

52

37

52

38

3/week

39

28

39

28

39

28

39

28

39

28

39

28

39

28

2/week

26

19

26

18

26

19

26

18

26

19

26

18

26

19

1/week

13

9

13

10

13

10

13

9

13

9

13

9

13

10

3/month

10

7

10

7

10

7

10

7

10

7

10

7

10

7

2/month

7

5

7

5

7

5

7

5

7

5

7

5

7

5

1/month

3

3

3

3

3

3

3

3

3

3

3

3

3

3

Table 16 presents the number of trucks needed to deliver the total volume according to different shipment frequency and number of SKUs scenarios. The values are the same for all different forecast accuracy level scenarios; hence, they are not presented here separately. It can be seen that the higher the volume and frequency of the lane the higher the difference in number of trucks needed to deliver the total volume between CTL and FTL and thus the savings. 44

Moreover, the results show that the number of SKUs on a lane or the forecast accuracy levels in the system did not show a significant impact on the ‘P&G transportation cost’. P&G administration cost / Customer administration cost Similar to ‘P&G transportation cost’, eliminating the number of trucks required to deliver the total volume led to a positive impact in the administration costs in all of the scenarios (Table 15); the cost decreased and savings realized. The extent of the savings depended on the frequency of the lane (i.e. the volume of the business); the higher the volume and frequency of the lane the higher the savings. Customer inventory cost When the CTL is improved to FTL, resulting total inventory level showed differences depending on the analyzed scenario. For instance, the impact on the inventory level and so on the ‘Customer inventory cost’ was very minor positive when the shipment frequency of the lane was high; the number of SKUs on the lane was low; and the forecast accuracy was low. Figure 13 summarizes the observed impacts.

Figure 13: Impacts on the customer inventory cost as a result of improving CTL to FTL

When the frequency of the lane was low, inventory level increased critically (Table 26 in Appendix 4), because the longer time interval in between replenishments prompted higher cycle stock. The impact was mainly major negative (Figure 13, Table 15); causing an increase in ‘Customer inventory cost’. The impact on ‘Customer inventory cost’ was still negative at medium shipment frequencies (Figure 13, Table 15). The extent of this negative impact depended on the other drives (i.e. the number of SKUs and forecast accuracy level). Almost no impact (i.e. negligible very minor positive) was observed in high frequency lanes (Figure 13, Table 15) as the modified VFR did not change the time interval in between replenishments -and thus cycle stock- in such lanes significantly. The higher the number of SKUs on a lane, the more the increase in inventory level through VFR modification (Table 26 in Appendix 4); because as more goods were pushed to the Customer DC per SKU, the total piled amount proliferated with numerous SKUs. Therefore, negative impact on ‘Customer inventory cost’ turn into major negative impact gradually as the number of SKUs rose (Figure 13). 45

Lower forecast accuracy level means a higher probability of unexpected events; and consequently higher safety stock level. The inventory level was already high when the forecast accuracy was low in CTL case. Therefore, the increase in cycle stock level when the system was improved to FTL was not obvious. Contrarily, a negligible minor positive impact was observed when the forecast accuracy was low, the shipment frequency of the lane was high and the number of SKUs on the lane was low (Figure 13, Table 26 in Appendix 4). This is a consequence of the dynamics between the cycle stock and safety stock. As the cycle stock and so its daily demand coverage increases, it leads a decline in safety stock. In this case, the decline in safety stock surpassed the increase in cycle stock (Table 26 and Table 27 in Appendix 4). It can be concluded that the forecast accuracy level was not one of the main drivers of the changes observed in inventory level and related costs. CO2 emissions Similar to ‘P&G transportation cost’, ‘P&G administration cost’ and ‘Customer administration cost’ eliminating the number of trucks required to deliver the total volume led to a positive impact in ‘CO2 emissions’ in all cases (Table 15); the emissions reduced. The extent of the savings depended mainly on the frequency of the lane (i.e. the volume of the business); as well as the distance travelled. The higher the volume and frequency of the lane and the longer the distance to be travelled the higher the savings. CTL vs. FTL with lower MOQs This part explains the behavior of the performance measures when VFR was increased and MOQs were reduces; specifically when the CTL was improved to FTL with lower MOQs (Table 17). Table 17: Behaviors of performance measures when CTL is improved to FTL with lower MOQs Shipment frequency (C TL) SAVING IN TOTAL VOLUME Cash

Customer inventory P&G transport

Cost

P&G admin Customer inventory Customer admin

External Cost

CO2 emissions

3/day

2/day

1/day

HIGH FREQ

4/week

3/week

2/week

MIDDLE FREQUENCY

variable/other factors no impact major positive medium positive impact impact variable/other factors no impact major positive impact

medium positive impact

1/week

3/month

2/month

LOW FREQUENCY

major positive impact minor positive impact major positive impact minor positive impact

P&G transportation cost / P&G administration cost /Customer administration cost / CO2 emissions Eliminating the number of trucks required to deliver the total volume led to a positive impact in these costs (Table 17); they decline and savings realized. Table 18 presents the number of trucks needed to deliver the total volume according to different shipment frequency and number of SKUs scenarios. The values are the same for all different forecast accuracy level scenarios; hence, they are not presented here separately. Again, the extent of the savings depended on the frequency of the lane (i.e. the volume of the business). It can be seen that the 46

1/month

higher the volume and frequency of the lane the higher the difference in number of trucks needed to deliver the total volume between CTL and FTL with lower MOQs and thus the savings. Improving to system from CTL to FTL with lower MOQs instead of FTL led further savings (Table 16 and Table 18). Table 18: Number of trucks needed to deliver the total volume according to different scenarios (CTL vs FTL with lower MOQs) Number SKUs 25

53

CTL

FTL w. MOQ

3/day

277

2/day

Ship. Fr.

159

CTL

FTL w. MOQ

185

275

181

121

265

CTL

FTL w. MOQ

184

278

182

122

371

CTL

FTL w. MOQ

187

280

181

122

477

CTL

FTL w. MOQ

188

279

182

122

583

CTL

FTL w. MOQ

CTL

FTL w. MOQ

187

278

187

277

186

182

122

181

122

183

123

1/day

91

61

91

61

91

61

91

61

90

60

91

61

91

61

4/week

52

35

52

35

52

35

51

34

52

35

52

35

52

35

3/week

39

26

39

26

39

26

39

26

39

26

39

26

39

26

2/week

26

18

26

17

26

18

26

17

26

18

26

17

26

18

1/week

13

9

13

9

13

9

13

9

13

9

13

9

13

9

3/month

10

7

10

7

10

7

10

7

10

7

10

7

10

7

2/month

7

5

7

5

7

5

7

5

7

5

7

5

7

5

1/month

3

2

3

2

3

2

3

2

3

2

3

3

3

2

Customer inventory cost When the CTL is improved to FTL with lower MOQs, resulting total inventory level showed differences depending on the analyzed scenario. The impacts were almost the opposite of improving the system from CTL to FTL. For instance, the impact on the inventory level and so on the ‘Customer inventory cost’ was very minor negative when the shipment frequency of the lane was high and the number of SKUs on the lane was low; and the forecast accuracy was high. Figure 14 summarizes the observed impacts.

Figure 14: Impacts on the customer inventory cost as a result of improving CTL to FTL with lower MOQs

When the frequency of the lane was low, inventory level decreased (Table 28 in Appendix 4), because the redundant part of the cycle stock could be eliminated through lowering the MOQ. 47

The impact was mainly major positive (Figure 14, Table 17); causing a decline in ‘Customer inventory cost’. The impact on ‘Customer inventory cost’ was still positive at medium shipment frequencies (Figure 14, Table 17). The extent of the effect depended on the other drives (i.e. number of SKUs and forecast accuracy level). Almost no impact (i.e. negligible very minor positive) was observed in high frequency lanes (Figure 14, Table 17) as the modified VFR and reduced MOQs did not change the time interval in between replenishments in such lanes significantly. The higher the number of SKUs on a lane, the more the decrease in inventory level through VFR modification (Table 28 in Appendix 4). More stock could be eliminated from the total amount with numerous SKUs as part of the redundant cycle stock could be cut per SKU. Therefore, positive impact on ‘Customer inventory cost’ turn into major negative impact gradually as the number of SKUs rose (Figure 14). The inventory level was already high when the forecast accuracy was low in CTL case. Therefore, the decrease in cycle stock level when the system was improved to FTL with lower MOQs was not obvious. Contrarily, a negligible minor negative impact was observed when the forecast accuracy was low and the shipment frequency of the lane was high (Figure 14). This is a consequence of the dynamics between the cycle stock and safety stock. As the cycle stock and so its daily demand coverage decreases, it leads an increase in safety stock. In this case, the increase in safety stock surpassed the decrement in cycle stock (Table 27 and Table 28 in Appendix 4). It can be concluded that the forecast accuracy level was not one of the main drivers of the changes observed in inventory level and related costs. The results indicated negative impacts, specifically on the inventory level, when the VFR was increased (i.e. CTL was improved to FTL) without any other changes in the supplier-retailer chain. Subsequently, it was shown that the direction of the negative impact changes to positive when the MOQs were reduced while increasing the VFR (i.e. CTL was improved to FTL with lower MOQs). Therefore, reducing the MOQs provides an opportunity to improve the VFR while keeping the balance with service level and logistics costs; and even improving them in most cases.

48

Chapter 5 Conclusion and Recommendations This chapter summarizes the main findings of the project. Afterwards, it provides recommendations.

5.1 Conclusions This project was held in cooperation with SNIC, P&G. The research assignment was set as follows: Assess the impacts of increased VFR on the other performance measures of the system; and then, to come up with potential decisions to improve the VFR in outbound transportation in order to achieve a win/win/win solution for the manufacturer/retailer/consumer. Firstly, the VFR was increased; CTL is modified to FTL. The impacts of increased VFR on service level and logistics costs were assessed. The analyses revealed that the extent of the impacts on performance measures differed mainly according to the volume/frequency of the lane and the number of SKUs on the lane. It was found that increasing VFR without any other changes in the supplier-retailer chain has negative impacts, specifically on the inventory level. Improving the outbound deliveries from CTL to FTL can only be reasonable for high frequency lanes. In practice, this is tricky; because high frequency deliveries usually performed between the supplier and retailers where the DCs are nearby. Hence, it should be checked if the expected improvements in total cost actually satisfy the modifications in VFR. Afterwards, MOQs were reduced; CTL is modified to FTL with lower MOQs. It was found that reducing MOQs while increasing the VFR changes the direction of the impact on the inventory level from negative to positive. Besides, improving the outbound deliveries from CTL to FTL with lower MOQs was reasonable for most of the scenarios. The performance measures were improved especially for medium and low frequency lanes. Clearly, increasing VFR meanwhile lowering MOQs is an opportunity to improve the VFR while keeping the balance with service level and logistics costs; and even improving them in most cases.

5.2 Recommendations This study contributes to the relevant research area by providing an example of increasing the VFR in outbound transportation. Firstly, it provides a categorization of the outbound delivery elements (Table 23, Table 24and Table 25) as well as an overall mapping of them with their interactions (Figure 14). These can be used as a check-list in related studies; and/or can facilitate detecting the indirect dynamics. Afterwards, the study presents the possible 49

outcomes of increasing the VFR. This helps to identify the related risks and opportunities. Furthermore, the study suggests an approach which will mitigate the possible negative impacts of increased VFR and improve the logistics costs further. The study contributes to the company in several ways. Firstly, the categorization of the outbound delivery elements (Table 23, Table 24and Table 25) as well as the overall mapping of them with their interactions (Figure 14) can provide guidance for people who are not involved in supply chain operations in understanding the part of supply chain dynamics. Moreover, the company can use this study in determining the type of businesses that exhibit the VFR improvement potential. Similarly, when the VFR of a specific lane is considered to be increased, the company can consult the results of the relevant scenarios of this study. Furthermore, displayed mutual benefits of improving CTL to FTL with lower MOQs will motivate both the suppliers and retailers to increase the VFR. It is important to note that the conclusions should be interpreted considering the assumptions made during the analysis. Furthermore, it should be noted that this was not a financial study. The focus was on the goods flow. The specific technical solutions, possible investments and the absolute changes in terms of cost were not analyzed. As a future research direction, handling costs can be incorporated into the study. Due to time limitations, specific technical solutions (for ex: double stacking and new handling equipment) could not be assessed and handling costs could not be covered. Consideration of investments on the technical solution as well as resulting handling costs will provide a more realistic overall idea on the results. This study mainly focuses on what will happen and what should be done when the VFR is increased instead of how VFR can (optimally) be increased. Another potential research can focus on the density of the payload for instance by mixing the heavy and light SKUs in order to leverage the increased VFR and/or not to face with obstacles in practice. During the study, it was recognized that the literature lacks a whole quantifiable model of the VFR in which all the aspects of the outbound deliveries are included. Taking the overall mapping of the outbound delivery elements as a basis (Figure 14), the relations of the VFR with related elements can be quantified. Being aware of the high complexity in the supply chain, still, it will be worthwhile to consider modeling for the setting. Lastly, while this project is focusing on the outbound trade lane scope, the connections with the wider end-to-end perspective could be the scope of further studies.

50

References Bilan Carbone. “Emission factor guide”. France; 2007. Caplice, C, and Sheffi, Y. “A review and evaluation of logistics performance measurement systems”. The International Journal of Logistics Management 1995; 6, 1; p. 61-74. Chapman, L. “Transport and climate change: a review”. Journal of Transport Geography 2007;.15; p. 354–367. Daganzo, C.F, and Newell, G.F. “Handling operations and the lot size trade-off”. Transportation Research Part B: Methodological 1993; 27, 3; p. 167-183. Department for Transport, UK. “Key performance indicators for food and drink supply chains”. Freight Best Practice Publications. UK; 2007. Disney, S, Potter, A, and Gardner, B. “The impact of vendor managed inventory on transport operations”. Transportation Research Part E: Logistics and Transportation Review 2003; 39; p. 363–380. European Commission. “White Paper: Roadmap to a single European Transport Area – Towards a competitive and resource efficient transport system”. Brussels; 2011. European Energy Agency. “Climate for a transportation change, TERM 2007, indicators tracking transport and environment in the European Union”. Copenhagen; 2008. European Energy Agency. “Transport and environment: Facing a dilemma, TERM2005, indicators tracking transport and environment in the European Union”. Copenhagen; 2006. European Environment Agency. “Load factors for freight transport, TERM 030”. 2010. http://www.eea.europa.eu/data-and-maps/indicators/load-factors-for-freight-transport/ load-factors-for-freight-transport-1. Last retrieved in 22.08.2011. INFRAS. “Final report: External costs of transport”. Zurich/Karlsruhe; 2004. Kelton, W.D, Sadowski, R.P , and Sturrock, D. T. Simulation with Arena. New York: McGraw-Hill; 2003. Law, A.M, and Kelton, W.D. Simulation modeling and analysis. New York: McGraw-Hill; 1991. Livingstone, G. “Measuring customer service in distribution”. International Journal of Physical Distribution & Logistics Management 1992; 22, 6; p. 4-6. Makridakis, S, and Hibon, M. “The M3-Competition: results, conclusions and implications”. International Journal of Forecasting 2000; 16; p. 451-476.

51

McKinnon, A. “A logistical perspective on the fuel efficiency of road freight transport”. Workshop “Improving fuel efficiency in road freight: The role of information technologies”. Paris; 1999. McKinnon, A. “Report prepared for the 15th ACEA scientific advisory group meeting: European freight transport statistics: limitations, misinterpretations and aspirations”. Brussels; 2010. Pibernik, R. “Managing stock-outs effectively with order fulfillment systems”. Journal of Manufacturing Technology Management 2006; 17, 6; p. 721-736. Procter&Gamble. “Annual Report”. 2010. Procter&Gamble. “Sustainability Report”. 2010. Procter&Gamble website, www.pg.com. Last retrieved in 20.08.2011 Ross, S.M. Introduction to probability and statistics for engineers and scientists. USA: Harcourt Academic Press; 2000. Silver, E.A, Pyke, D.F, and Peterson, R. Inventory management and production planning and scheduling. USA: John Willy & Sons; 1998. Van Aken, J.E, Berends, H, and Van der Bij, H. Problem-solving in organizations: Methodological handbook for business students. Cambridge: University press; 2007. Van Der Vlist, P. “Synchronizing the Retail Supply Chain”. Erasmus University, Rotterdam School of Management; ERIM Ph.D. Series Research in Management no. 110, http://hdl.handle.net/1765/1, ISBN 90-5892-142-0. Rotterdam; 2007. Van Der Vorst, J.G.A.J, Beulens, A.J.M, De Wit, W, and Van Beek, P. “Supply chain management in food chains: Improving performance by reducing uncertainty”. International Transactions in Operational Research 1998; 5, 6; p. 487-499. Van Der Vorst, J.G.A.J, and Beulens, A.J.M. “Identifying sources of uncertainty to generate supply chain redesign strategies”. International Journal of Physical Distribution & Logistics Management 2002; 32. 6; p. 409-430. Van Strien, P.J. “Towards a methodology of psychological practice: the regulative cycle”. Theory Psychology 1997; 7; p. 683 – 700.

52

Appendices Appendix 1: An overview of P&G data

Table 19: VFR in P&G

CONFIDENTIAL

Table 20: Max allowed legal weight

Max allowed legal weight of a vehicle

24 tons of payload

The 24 tons of maximum allowed legal payload weight presented in Table 20 is the value for a standard semi-trailer truck which was assumed in this project.

53

Table 21: P&G plant locations in Western Europe (as of 30 June 2010)

Country Belgium Belgium France France France France Germany Germany Germany Germany Germany Germany Germany Germany Germany Germany Germany Ireland Ireland Ireland Italy Italy Italy Italy Netherlands Portugal Spain Spain Spain Spain United Kingdom United Kingdom United Kingdom United Kingdom

City Aarschot Mechelen Amiens Blois Poissy Sarreguemines Berlin Cologne Crailsheim Euskirchen Gross-Gerau Hünfeld Kronberg Marktheidenfeld Rothenkirchen Walldürn Worms Carlow Nenagh Newbridge Campochiaro Gattatico Pescara Rome Coevorden Guifões Jijona Mataró Mequinenza Montornès del Vallès London Manchester Reading Whitley Bay

(Source: P&G Sustainability Report, 2010)

54

Plant Name Aarschot Mechelen Amiens Blois Poissy Sarreguemines Berlin Cologne Crailsheim Euskirchen Gross-Gerau Huenfeld Kronberg Marktheidenfeld Rothenkirchen Wallduern Worms Carlow Nenagh Newbridge Campochiaro Gattatico Pescara Pomezia Coevorden Porto Jijona Mataró Mequinenza Montornès London Manchester Reading Seaton Delaval

Category Owner Duracell Fabric Care Fabric Care Hair Care Personal Beauty Care Hair Care Blades & Razors Personal Beauty Care Feminine Care Baby Care Oral Care Hair Care Braun Oral Care Hair Care Braun Fabric Care Braun Personal Beauty Care Oral Care Fabric Care Home Care Feminine Care Fabric Care Pet Care Fabric Care Baby Care Fabric Care Baby Care Feminine Care Home Care Baby Care Personal Beauty Care Personal Beauty Care

Table 22: VFR (Inbound deliveries vs Outbound deliveries in Western Europe)

CONFIDENTIAL

Figure 15: The lane selected for analysis

CONFIDENTIAL

55

Appendix 2: Outbound delivery elements Urg ent s hipment cos t parameters Total lane bus ines s (cs )

External cos t (Euro's ) Time interval in between s hipments (hrs -days -wks )

VFR (%)

# of s hipments (s hpmnts /wk)

PG trans port cos t (Euro's )

Traffic dens ity Urg ent s hipment allowed? (binary) Flexibility of time windows impos ed by C us tomer

Trade terms

D is tance travelled (kms )

Variable vehicle us e cos t

Probability of an urg ent s hipment (%)

# of urg ent s hipments (s hpmnts /wk)

Probability of being late per s hipment (%) Lead time from PG D C to C us tomer D C (hrs -days -wks )

# decks (units )

C us tomer vehicle unloading cos t (Euro's )

Max # of pallets (pl) Pallet heig ht (m)

PG handling cos t (Euro's )

PG on time delivery rate (%)

# of pallets (pl)

PG order picking & pallet loading cos t (Euro's )

C us tomer's forecas t horizon (hrs -days -wks ) D elivery quantity for each SKU per s hipment (cs -ly-pl/SKU)

C us tomer order s plitting cos t (Euro's ) Max leg al weig ht (kg s )

C us tomer handling cos t(Euro's ) C us tomer unit vehicle unloading cos t (Euro's /pl /deck /s hpmnt)

Pallet leng th (m) Pallet width (m) Vehicle width (m) Vehicle heig ht (m) Vehicle leng th (m)

P& G unit order picking & pallet loading cos t (Euro's /pl)

# of SKUs on the lane (units )

C us tomer unit order s plitting cos t (Euro's / pl)

SKU dimens ions (cs /pl/SKU)

PG cas e fill rate (%)

C us tomer cycle s tock per SKU (cs /SKU)

SKU weig ht (kg /cs /SKU) Promotion O ne time introduction Phas e-in Phas e-out Seas onality Smoothnes s Targ et s ervice level in terms of product availability (%)

P& G unit vehicle loading cos t (Euro's /pl /deck /s hpmnt)

Fixed vehicle us e cos t (Euro's /s hpmnt) PG vehicle loading cos t (Euro's )

R eview period (hrs -days -wks ) MO I per SKU (cs -ly-pl/SKU)

External cos t parameters

Shopper demand variability per SKU (cs /SKU)

C us tomer's forecas t accuracy per SKU (%)

C us tomer's forecas t per SKU (cs /SKU)

C us tomer's order per SKU (cs -ly-pl/SKU)

C us tomer product availability (%) C us tomer pipeline s tock per SKU (cs /SKU)

C us tomer s afety s tock per SKU (cs /SKU)

Shopper demand per SKU (H/M/L) (cs )

C us tomer total inventory per SKU (cs /SKU)

C us tomer total inventory holding cos t (Euro's )

C us tomer s torag e capacity

C us tomer unit inventory holding cos t (Euro's /pl s pot/month)

Figure 16: Mapping of the elements in outbound deliveries

In Figure 16, the element at the back of an arrow has a direct influence on the element at the head of the very same arrow. For red elements measurement unit could not be provided; more analysis is required. Table 23, Table 24 and Table 25 categorizes decision, dependent and independent variables as well as performance measures of the outbound delivery elements in Figure 16 under the related processes and provide a definition of the elements where necessary. 56

Table 23: Categorization of the elements in outbound deliveries (P&G Operations) P&G Operations Variable (Definition) (Unit)

Order Accepting

Order Picking & Pallet Loading

Outbound transportation

Vehicle Loading

Regular Transportation

Urgent Delivery

Target VFR (Vehicle fill rate in terms of weight and volume) (%)

Decision variables

Process Steps

Warehousing

# of pallets (Also has the information of what a pallet contains) (pl)

# of decks (units)

Independent variables

P&G unit vehicle loading cost (Euro's /pl /deck /shpmnt)

P&G unit order picking & pallet loading cost (According to what pallet contains) (Euro's /pl)

Pallet height (m) Pallet length (m) Pallet width (m) Vehicle height (m) Vehicle length (m) Vehicle width (m) Max legal weight (kg)

Fixed vehicle use cost (Euro's /shpmnt) Variable vehicle use cost parameters Distance travelled (km) External cost parameters Time windows imposed by customer (The time interval which the delivery can be performed; the narrower the time window the less the probability to be on time)

Urgent shipment allowed (Y/N)? (If the system allows urgent deliveries, it is "Yes"; otherwise it is "No") (1/0) Urgent shipment cost parameters

Performance measures

Dependent variables

# of shipments (shpmnt/wk)

Delivery quantity for each SKU per shipment (Delivery quantity per SKU a shipment contains in terms of cases, layers and pallets) (cs-ly-pl/ SKU) P&G case fill rate (The percentage of customer orders that are filled from inventory on hand) (%)

Time interval between 2 shipments (hrs-days-wks)

Max # of pallets (pl)

P&G order picking & pallet loading cost (handling performed in order picking & pallet loading) (Euro's) P&G vehicle loading cost (handling performed in vehicle loading) (Euro's) P&G handling cost (Euro's)

Lead time (Time in between an order release and delivery of goods) (hrs-days-wks)

Probability of an urgent shipment (%)

Probability of being late per shipment (%)

# of urgent shipments (shpmnts/wk)

P&G transport cost (Euro's) P&G on time delivery rate (%) External cost (including CO2 and traffic related costs) (Euro's)

For red elements in Table 23 measurement unit could not be provided; more analysis is required.

57

Table 24: Categorization of the elements in outbound deliveries (Customer Operations) Customer operations Variable (Definition) (Unit)

Vehicle Unloading

Order Splitting

Stocking

Ordering

Forecasting & Ordering

Independent variables

Decision variables

Process Steps

Warehousing

Customer unit vehicle unloading cost (Euro's /pl /deck /shpmnt)

Customer unit inventory holding cost (cost & cash) (Euro's/pl spot/month) Customer unit order splitting cost (According to what pallet contains) (Euro's / pl)

Customer storage capacity (units or volume) Review period (hrs/days/wks)

Target service level in terms of product availability (%)

Dependent variables

Customer's forecast horizon (hrsdays-wks) Customer's forecast accuracy per SKU (That is due to forecast horizon and demand variability per SKU) (%) Customer cycle stock per SKU (cs/SKU) Customer safety stock per SKU (cs/SKU)

Performance measures

Customer pipeline stock per SKU (cs/SKU) Customer vehicle unloading cost (handling performed in vehicle unloading) (Euro's) Customer order splitting cost (handling performed in order splitting) (Euro's) Customer handling cost (Euro's)

Customer's forecast per SKU (cs/SKU) Customer's order per SKU (cs-lypl/SKU)

Customer total inventory level per SKU (cs/SKU) Customer total inventory holding cost (Euro's) Customer product availability (The percentage of store/shopper demand that are filled from inventory on hand) (%)

For red elements in Table 24 measurement unit could not be provided; more analysis is required.

58

Table 25: Categorization of the elements in outbound deliveries (System characteristics) System characteristics Variable (Definition) (Unit) Product

Other

Decision variables

Process Steps

Lane

MOI per SKU (MOI: Minimum order increment, the minimum order amount in terms of cs/ly/pl for an SKU) (cs-ly-pl/ SKU)

Independent variables

Trade terms Shopper demand per SKU (Demand profile per SKU, the distribution with its mean and variance, as well as the level of rotation, H/M/L) (cs)

# of SKUs on the lane (units)

SKU dimensions (To assess how many cases will fit to a pallet) (cs per pl per SKU) SKU weight (kg/cs/SKU)

Traffic density (It has a special status since it is both dependent on and independent from system variables)

Performance measures

Dependent variables

Total lane business (Amount of the delivery that has been done within a specific time period for a lane) (cs)

Shopper demand variability per SKU (Demand variability per SKU based on characteristics of the product such as smoothness, seasonality, promotions, phase-ins, phase-outs, one time introduction) (cs/SKU)

For red elements in Table 25 measurement unit could not be provided; more analysis is required.

59

Appendix 3: Retailer Inventory Model

Figure 17: RIM user interface

Yellow markings in Figure 17 indicate the inputs of the tool; green markings indicate the outputs of the tool. Blue markings indicate the additions/adaptations performed on RIM for this study.

60

Appendix 4: Results Table 26: Increase in inventory level for all scenarios when the CTL was improved to FTL

Distributed forecast accuracy level 3/day

2/day

1/day

4/week

3/week

2/week

1/week

3/month

2/month

1/month

25

0,00

0,00

1,52

1,29

1,22

1,23

3,67

3,18

5,13

1,57

53

0,01

0,02

1,53

1,34

1,31

2,30

3,12

3,60

5,97

4,39

159

0,09

0,16

1,69

1,76

1,94

2,46

5,26

6,52

9,94

16,69

265

0,17

0,34

1,94

2,34

2,68

4,32

8,14

10,14

15,24

27,36

371

0,29

0,50

1,09

2,88

3,47

4,84

10,92

13,41

19,24

39,82

477

0,42

0,70

2,53

3,50

4,34

6,80

13,58

17,61

26,40

51,63

583

0,52

0,86

2,85

4,08

5,19

7,55

15,45

21,23

32,29

63,80

# SKUs

Frequency

High forecast accuracy level 3/day

2/day

1/day

4/week

3/week

2/week

1/week

3/month

2/month

1/month

25

0,00

0,00

0,23

0,42

0,53

0,71

2,06

2,48

3,45

1,94

53

0,01

0,02

0,28

0,48

0,65

1,22

1,95

2,91

4,69

4,99

159

0,09

0,16

0,58

1,03

1,38

2,10

4,45

6,07

9,04

17,58

265

0,17

0,34

0,92

1,69

2,20

3,51

7,16

9,83

14,56

27,81

371

0,29

0,50

1,22

2,27

3,02

4,58

10,02

13,14

18,62

40,90

477

0,42

0,70

1,64

2,95

3,95

6,17

12,85

17,44

25,98

52,61

583

0,52

0,86

2,02

3,57

4,85

7,36

15,05

21,09

31,98

64,67

# SKUs

Frequency

Medium forecast accuracy level 3/day

2/day

1/day

4/week

3/week

2/week

1/week

3/month

2/month

1/month

25

-0,07

-0,03

-0,15

0,02

0,10

0,35

1,81

2,11

3,29

1,80

53

-0,03

-0,09

-0,07

0,16

0,26

0,83

1,75

2,57

4,39

4,73

159

-0,13

-0,13

0,23

0,66

0,99

1,70

4,14

5,73

8,81

17,20

265

0,03

-0,13

0,56

1,33

1,80

3,19

6,90

9,48

14,33

27,62

371

-0,08

0,30

0,53

1,90

2,62

4,18

9,77

12,82

18,41

40,44

477

0,08

0,24

1,28

2,58

3,55

5,84

12,59

17,12

25,75

52,20

583

0,20

0,61

1,65

3,21

4,45

6,97

14,75

20,79

31,75

64,31

# SKUs

Frequency

Low forecast accuracy level 3/day

2/day

1/day

4/week

3/week

2/week

1/week

3/month

2/month

1/month

25

-0,19

-0,09

-0,86

-0,69

-0,76

-0,26

1,28

1,45

2,93

1,58

53

-0,10

-0,25

-0,71

-0,41

-0,42

0,13

1,34

1,99

3,95

4,36

159

-0,49

-0,59

-0,40

0,06

0,33

1,00

3,60

5,15

8,33

16,68

265

-0,20

-0,89

-0,08

0,68

1,15

2,61

6,41

8,92

13,87

27,35

371

-0,69

-0,02

-0,59

1,25

1,97

3,52

9,28

12,26

17,96

39,80

477

-0,47

-0,51

0,60

1,93

2,90

5,26

12,09

16,59

25,32

51,62

583

-0,34

0,21

1,01

2,55

3,76

6,33

14,23

20,28

31,33

63,78

# SKUs

Frequency

61

Table 27: Example for the dynamics between the cycle stock and safety stock

For 159 SKUs

3 2 1 4 3 2 1 3 2 1

3 2 1 4 3 2 1 3 2 1

3 2 1 4 3 2 1 3 2 1

Distributed for. acc.

High for. acc.

/day /day /day /week /week /week /week /month /month /month

Cycle stock 0,82 1,01 1,75 3,07 4,14 6,21 12,42 17,07 24,83 54,63

Safety stock 3,31 3,24 3 3,82 4,49 4,79 5,47 6,22 6,31 7,82

Cycle stock 0,82 1,01 1,75 3,07 4,14 6,21 12,42 17,07 24,83 54,63

/day /day /day /week /week /week /week /month /month /month

Cycle stock 1 1,29 2,39 4,19 5,65 8,47 16,94 23,3 33,88 72,55

Safety stock 3,22 3,13 4,05 4,47 4,93 4,99 6,21 6,51 7,19 6,59

Cycle stock 1 1,29 2,39 4,19 5,65 8,47 16,94 23,3 33,88 72,55

/day /day /day /week /week /week /week /month /month /month

Cycle stock 0,73 0,88 1,6 2,8 3,77 5,52 11,04 14,85 21,6 49,78

Safety stock 3,35 3,28 4,29 4,79 5,32 6,28 7,64 7,73 8,76 7,14

Cycle stock 0,73 0,88 1,6 2,8 3,77 5,52 11,04 14,85 21,6 49,78

62

Safety stock 3,31 3,24 3 2,78 2,71 2,52 2,2 2,12 1,9 1,8

Medium for. acc. CTL

Cycle stock 0,82 1,01 1,75 3,07 4,14 6,21 12,42 17,07 24,83 54,63

FTL Safety Cycle stock stock 3,22 1 3,13 1,29 2,94 2,39 2,7 4,19 2,59 5,65 2,35 8,47 2,13 16,94 1,97 23,3 1,88 33,88 1,47 72,55 FTL with lower MOQs Safety Cycle stock stock 3,35 0,73 3,28 0,88 3,11 1,6 2,89 2,8 2,79 3,77 2,74 5,52 2,58 11,04 2,43 14,85 2,43 21,6 1,6 49,78

Low for. acc.

Safety stock 10,21 9,99 9,16 8,24 7,84 7,13 6 5,62 4,97 4,3

Cycle stock 0,82 1,01 1,75 3,07 4,14 6,21 12,42 17,07 24,83 54,63

Safety stock 21,4 20,93 19,15 17,03 16,04 14,48 11,96 11,05 9,69 7,87

Safety stock 9,9 9,59 8,75 7,79 7,33 6,56 5,61 5,13 4,73 3,58

Cycle stock 1 1,29 2,39 4,19 5,65 8,47 16,94 23,3 33,88 72,55

Safety stock 20,73 20,06 18,11 15,98 14,87 13,22 11,03 9,98 8,98 6,63

Safety stock 10,35 10,12 9,28 8,34 7,9 7,47 6,59 6,13 5,83 3,89

Cycle stock 0,73 0,88 1,6 2,8 3,77 5,52 11,04 14,85 21,6 49,78

Safety stock 21,71 21,2 19,22 17,11 16,03 15,03 12,75 11,76 10,88 7,18

Table 28: Increase in inventory level for all scenarios when the CTL was improved to FTL with lower MOQs

Distributed forecast accuracy level 3/day

2/day

1/day

4/week

3/week

2/week

1/week

3/month

2/month

1/month

25

0,00

0,00

1,50

1,22

1,14

1,97

3,14

2,39

4,26

4,09

53

0,00

-0,01

1,43

1,13

1,02

1,82

2,80

1,90

3,15

2,31

159

-0,05

-0,09

1,13

0,69

0,46

0,80

0,80

-0,71

-0,78

-5,54

265

-0,11

-0,19

0,80

0,11

-0,25

-0,31

-1,51

-3,62

-5,25

-16,10

371

-0,16

-0,32

-0,69

-0,30

-0,80

-1,11

-3,35

-5,71

-7,63

-21,55

477

-0,24

-0,42

0,21

-0,79

-1,46

-2,20

-5,55

-8,66

-13,06

-32,00

583

-0,33

-0,52

-0,07

-1,20

-2,03

-2,99

-6,85

-10,91

-17,79

-36,86

# SKUs

Frequency

High forecast accuracy level 3/day

2/day

1/day

4/week

3/week

2/week

1/week

3/month

2/month

1/month

25

0,00

0,00

0,19

0,31

0,41

0,69

1,26

1,41

2,37

5,63

# SKUs

Frequency 53

0,00

-0,01

0,16

0,23

0,27

0,61

0,99

0,87

1,28

3,54

159

-0,05

-0,09

-0,05

-0,17

-0,29

-0,47

-1,00

-1,91

-2,70

-5,05

265

-0,11

-0,19

-0,32

-0,74

-1,02

-1,55

-3,24

-4,75

-6,98

-16,03

371

-0,16

-0,32

-0,69

-1,07

-1,48

-2,20

-4,82

-6,58

-9,01

-21,20

477

-0,24

-0,42

-0,80

-1,55

-2,15

-3,28

-6,98

-9,58

-14,41

-32,69

583

-0,33

-0,52

-1,02

-1,91

-2,67

-4,00

-8,19

-11,76

-19,43

-36,82

Medium forecast accuracy level 3/day

2/day

1/day

4/week

3/week

2/week

1/week

3/month

2/month

1/month

25

-0,01

-0,01

-0,09

0,02

0,05

0,50

1,22

1,23

2,31

4,98

53

0,04

0,09

-0,04

0,03

0,05

0,39

0,98

0,77

1,30

3,02

159

0,05

0,00

-0,04

-0,17

-0,31

-0,34

-0,79

-1,72

-2,37

-5,27

265

0,06

-0,07

-0,12

-0,53

-0,84

-1,30

-2,89

-4,51

-6,63

-16,02

371

-0,06

0,04

-0,70

-0,97

-1,42

-2,06

-4,62

-6,48

-8,81

-21,35

477

-0,04

-0,21

-0,58

-1,34

-1,96

-3,02

-6,68

-9,37

-14,14

-32,40

583

-0,10

-0,31

-0,81

-1,70

-2,48

-3,76

-7,93

-11,56

-18,94

-36,84

# SKUs

Frequency

Low forecast accuracy level 3/day

2/day

1/day

4/week

3/week

2/week

1/week

3/month

2/month

1/month

25

-0,02

-0,01

-0,62

-0,51

-0,71

0,07

1,01

0,86

2,11

4,08

53

0,10

0,26

-0,43

-0,37

-0,38

-0,06

0,81

0,55

1,32

2,30

159

0,22

0,14

-0,08

-0,19

-0,39

-0,14

-0,59

-1,51

-2,05

-5,54

265

0,33

0,13

0,13

-0,29

-0,61

-0,99

-2,51

-4,22

-6,25

-16,00

371

0,10

0,61

-0,73

-0,91

-1,42

-1,94

-4,45

-6,52

-8,67

-21,56

477

0,28

0,12

-0,33

-1,07

-1,69

-2,71

-6,34

-9,11

-13,83

-31,99

583

0,27

0,00

-0,53

-1,45

-2,29

-3,46

-7,60

-11,33

-18,33

-36,86

# SKUs

Frequency

63