Department of Technology Management and Economics Division of Logistics and Transportation CHALMERS UNIVERSITY OF TECHNOLOGY Göteborg, Sweden 2016 Report No. E2016:002 Parts feeding of low-volume parts to assembly lines in the automotive industry Master of Science Thesis in the Supply Chain Management Programme ANDREAS KARLSSON MARKUS SVANSTRÖM MASTER’S THESIS E2016:002 Parts feeding of low-volume parts to assembly lines in the automotive industry ANDREAS KARLSSON MARKUS SVANSTRÖM Tutor, Chalmers: Robin Hanson Department of Technology Management and Economics Division of Logistics and Transportation CHALMERS UNIVERSITY OF TECHNOLOGY Göteborg, Sweden 2016 Parts feeding of low-volume parts to assembly lines in the automotive industry ANDREAS N. R. KARLSSON MARKUS E. V. SVANSTRÖM © ANDREAS N. R. KARLSSON & MARKUS E. V. SVANSTRÖM, 2016. Master’s Thesis E2016:002 Department of Technology Management and Economics Division of Logistics and Transportation Chalmers University of Technology SE-412 96 Göteborg, Sweden Telephone: + 46 (0)31-772 1000 Chalmers Reproservice Göteborg, Sweden 2016 i Parts feeding of low-volume parts to assembly lines in the automotive industry ANDREAS KARLSSON MARKUS SVANSTRÖM Department of Technology Management and Economics Division of Logistics and Transportation Chalmers University of Technology ABSTRACT Parts feeding to mixed-model assembly lines in the automotive industry is a large challenge, since diverse customer demands have increased the amount of parts handled within the production facilities; hence a large amount of parts can be categorized as low-volume parts. Existing theory states that the design of the parts feeding system impacts the performance of the production system, however, there is a gap in existing literature regarding research focusing on parts feeding policies appropriate for low-volume parts. The aim of this study is therefore to develop guidelines for when different parts feeding policies are suitable to apply for LVPs and highlight the effects of design options related to the parts feeding system. The aim has been broken down into three research questions which the study should answer: 1. How can low-volume parts be defined for parts feeding in the automotive industry? 2. Which parts feeding policies are suitable to use for low-volume parts, and for what part characteristics does each policy fit best? 3. How should the parts feeding system related to each parts feeding policy for low-volume parts be designed? A multiple case study of four companies within the automotive industry has been performed in order to fulfill the aim. The study has been qualitative and data has been collected through direct observations, semi-structured interviews and internal documents. It was identified that companies within the automotive industry can benefit from categorizing their parts based on consumption volume, where low-volume parts could be grouped into a separate segment. This allows for reduced complexity in the assignment of parts feeding policies. This study has concluded that it is less beneficial to use continuous supply for low- volume parts compared with other parts feeding policies. The assignment of parts feeding policies for low-volume parts can be performed based on part characteristics, where the part size and amount of part variants within the part family have been identified as most relevant to consider. Findings related to the design of the parts feeding system include that space limitations near the assembly line has a large influence on design choices. Furthermore, it is beneficial to consolidate parts when transporting them to the lineside presentation. In addition, the picking accuracy has been identified as more important for low-volume parts than picking efficiency, and picking information such as pick-by-voice can be helpful to enable increased accuracy of picking operations. Keywords: parts feeding policy, parts feeding system, low-volume parts, automotive industry, continuous supply, kanban-based continuous supply, kitting, sequencing ii iii ACKNOWLEDGEMENTS This master’s thesis was conducted within the Supply Chain Management programme at Chalmers University of Technology between September 2015 and January 2016. The study was performed within a research project at Chalmers University of Technology. There are several people that have been involved in this thesis to whom we want to express our gratitude. We would like to thank our supervisor, Robin Hanson at Chalmers University of Technology, for his guidance and feedback during the process. In addition, we would like to thank the four participating companies for collaborating in this research project, and especially the employees with whom we have held interviews. Without your guidance and collaboration, this thesis would not have been possible. Göteborg, 2016 Andreas Karlsson and Markus Svanström iv v TABLE OF CONTENTS 1 Introduction .................................................................................................................................... 1 1.1 Problem background ................................................................................................................ 1 1.2 Aim .......................................................................................................................................... 3 1.3 Research questions .................................................................................................................. 3 1.4 Scope and limitations .............................................................................................................. 3 1.5 Thesis outline .......................................................................................................................... 4 2 Theoretical framework .................................................................................................................... 5 2.1 Classification of LVPs ............................................................................................................. 5 2.2 The parts feeding system and its constituents ......................................................................... 5 2.2.1 Parts feeding policies ....................................................................................................... 6 2.2.2 Packaging of parts ......................................................................................................... 10 2.2.3 Storage of parts .............................................................................................................. 11 2.2.4 Transportation of parts .................................................................................................. 12 2.2.5 Picking of parts .............................................................................................................. 13 2.3 Contextual factors impacting the design of the parts feeding system ................................... 14 2.4 How to assess the performance of the parts feeding system for LVPs .................................. 15 3 Methodology ................................................................................................................................. 17 3.1 Research approach ................................................................................................................. 17 3.2 Research strategy ................................................................................................................... 18 3.3 Research design ..................................................................................................................... 19 3.4 Work procedure ..................................................................................................................... 19 3.4.1 Literature review ........................................................................................................... 20 3.4.2 Data collection ............................................................................................................... 20 3.4.3 Data analysis .................................................................................................................. 22 3.5 Research quality .................................................................................................................... 22 4 Empirical findings ........................................................................................................................ 25 4.1 Company A............................................................................................................................ 25 4.1.1 LVPs at Company A ...................................................................................................... 25 4.1.2 Factors influencing the choice of parts feeding policies at Company A ....................... 25 4.1.3 Description of the parts feeding system at Company A ................................................ 26 4.1.4 Future design of the production system ......................................................................... 28 4.1.5 Summary of Company A ............................................................................................... 29 4.2 Company B ............................................................................................................................ 29 4.2.1 LVPs at Company B ...................................................................................................... 29 4.2.2 Factors influencing the choice of parts feeding policies at Company B ....................... 29 4.2.3 Description of the parts feeding system at Company B ................................................ 30 vi 4.2.4 Summary of Company B ............................................................................................... 32 4.3 Company C ............................................................................................................................ 32 4.3.1 LVPs at Company C ...................................................................................................... 33 4.3.2 Factors influencing the choice of parts feeding policies at Company C ....................... 33 4.3.3 Description of the parts feeding system at Company C ................................................ 34 4.3.4 Summary of Company C ............................................................................................... 36 4.4 Company D............................................................................................................................ 36 4.4.1 LVPs at Company D ...................................................................................................... 36 4.4.2 Factors influencing the choice of parts feeding policies at Company D ....................... 37 4.4.3 Description of the parts feeding system at Company D ................................................ 38 4.4.4 Summary of Company D ............................................................................................... 39 5 Analysis ........................................................................................................................................ 41 5.1 Definition of LVPs ................................................................................................................ 41 5.2 Parts feeding policies’ suitability for LVPs ........................................................................... 42 5.2.1 Continuous supply ......................................................................................................... 42 5.2.2 Kanban-based continuous supply .................................................................................. 43 5.2.3 Kitting ............................................................................................................................ 44 5.2.4 Sequencing .................................................................................................................... 45 5.3 The design of the parts feeding system related to parts feeding policies .............................. 46 5.3.1 Kanban-based continuous supply .................................................................................. 46 5.3.2 Kitting ............................................................................................................................ 48 5.3.3 Sequencing .................................................................................................................... 49 6 Results .......................................................................................................................................... 53 6.1.1 Definition of LVPs ........................................................................................................ 53 6.1.2 Parts feeding policies’ suitability for LVPs ................................................................... 53 6.1.3 The design of the parts feeding system related to parts feeding policies ...................... 54 7 Discussion ..................................................................................................................................... 57 7.1 Reflections on the findings .................................................................................................... 57 7.2 Contributions to academia and industry ................................................................................ 58 7.3 Generalizability of the findings ............................................................................................. 58 7.4 Trustworthiness of the study ................................................................................................. 58 7.5 Further research ..................................................................................................................... 60 8 Conclusions .................................................................................................................................. 61 References ............................................................................................................................................. 63 Appendix I. Material flows ................................................................................................................... 65 Appendix II. Questions for semi-structured interviews......................................................................... 69 1 1 INTRODUCTION This chapter will begin with a description of the constituents of a production system, which will then be narrowed down to explain the importance of parts feeding in the automotive industry, which will form a basis for the purpose of this study. The aim of the study as well as research questions are presented and the scope and limitations of the study are motivated. In the end of the chapter, an outline of the report is presented. 1.1 Problem background According to Finnsgård (2013), a production system in an assembly environment consists of a parts feeding system and an assembly system, see Figure 1. The parts feeding system supplies parts to the lineside presentation, which is the border to the assembly system. The delivered parts are within the assembly system assembled to an object and the output is end products (Finnsgård, 2013). These relations are presented in Figure 1 below. Figure 1. The production system and its constituents - Adapted from Finnsgård (2013) Studies show that some companies within the automotive industry offer up to billions of different product variants (Pil & Holweg, 2004). This is a consequence of the diverse customer demands that exist in the automotive industry. The high number of product variants impacts both the parts feeding system and the assembly system. The latter needs to maintain high efficiency while being capable of handling a high degree of flexibility. This is enabled by designing the assembly system into a mixed-model assembly line as it allows multiple products to be manufactured efficiently (Boysen, et al., 2009). However, as an increased number of parts needs to be handled, the complexity of the parts feeding system increases which could have a negative impact on the efficiency of the assembly system (Emde & Boysen, 2012). It is therefore important for a company to have a suitable design on their parts feeding system since it impacts the performance of the production system (Kilic & Durmusoglu, 2015). The parts feeding system can be divided into several constituents which all contribute to the performance of the system. According to Kilic & Durmusoglu (2015), the parts feeding system could be categorized into transport of parts, storage of parts, and parts feeding policy. However, previous studies highlight additional constituents which also could have an impact on the performance. An argumentation about the constituents of the parts feeding system will be further elaborated in the theoretical framework. This study will explore the constituent denoted as parts feeding policy in detail, due to its great influence on the entire parts feeding system. The parts feeding policy can be described as the method of delivering parts to the point-of-use (Kilic & Durmusoglu, 2015). However, as the constituents are interdependent it is necessary to consider the whole parts feeding system (Kilic & Durmusoglu, 2015). Production System Lineside Presentation Materials Flow Parts Feeding System Assembly System Supplier s 2 Emde & Boysen (2012) highlight the parts feeding to the mixed-model assembly line as a large challenge since a high amount of parts need to be handled. Caputo & Pelagagge (2011) state that an appropriate and cost efficient way of choosing among parts feeding policies is to use several different policies for different parts depending on their characteristics. Hua & Johnson (2010) concluded that part characteristics such as consumption volume, part size and variety within part family impact the design of the parts feeding system, meaning for example that the design for parts used in low volumes would differ from that of parts consumed in higher volumes. Furthermore, economic value of the part has also been identified as relevant to consider (Caputo & Pelagagge, 2011; Schwind, 1992). It has also been concluded that the design often is based on qualitative judgement where contextual factors such as company- specific practices, tradition, product structure and operational constraints impact the decision (Carlsson & Hensvold, 2008). Moreover, Carlsson & Hensvold (2008) concluded in their study of a production facility at Caterpillar that the organizational needs have a significant impact on the decision. For instance, the most appropriate design of the parts feeding policy depends on how the firm evaluates the importance of criteria such as lineside space reductions, operator walking time, and kitting time. Traditionally, continuous supply has been a common parts feeding policy to apply, where large quantities of all components are stored near the assembly line (Faccio, 2014). However, Caputo & Pelagagge (2011) state that continuous supply is not appropriate to use when there is a high amount of part variants, since it could cause high inventory costs and problems related to space limitations at the lineside presentation. As stated above, high amounts of part variants could often be seen in the automotive industry. Furthermore, the amount of non-value adding time for the operator can increase since excessive time is spent on fetching material due to longer walking distances (Medbo & Wänström, 2009). In order to handle these issues that could arise when using continuous supply, alternative parts feeding policies have been developed. Kitting, kanban-based continuous supply, and sequencing are some of the parts feeding policies that can be found in existing literature (Faccio, 2014; Sali, et al., 2015). Several comparisons between kitting and continuous supply can be found in previous research while there is a limited amount of studies regarding the other policies. As companies within the automotive industry experience more diverse customer demands, an increasing amount of parts needs to be handled by their parts feeding systems. Therefore, it is reasonable to argue that a large amount of parts could be categorized as low-volume parts (LVPs) today. While existing literature has not provided a definition for LVPs, the authors of this thesis refer to them as parts that are used to a low extent compared to other parts within the production system. Furthermore, to provide a definition of LVPs is one of the research questions that this thesis will treat. As the amount of LVPs increase, the importance of having an appropriate design of the parts feeding system becomes even more critical in order to reduce complexity, inventory costs, and issues related to space restrictions. However, even though several studies exist where parts feeding policies have been compared, literature focusing on policies that are appropriate to use for LVPs have not been found. As a consequence, companies cannot receive support from the academia about guidelines for how to choose the most suitable parts feeding policies for LVPs that fit their needs. Therefore, this study is focusing on filling this gap in the academia through a multiple case study. 3 1.2 Aim The aim of this study is to develop guidelines for when different parts feeding policies are suitable to apply for LVPs and highlight the effects of design options related to the parts feeding system. Since existing literature has identified that contextual factors impact the performance of the parts feeding system, the study will also highlight which factors that are most important for companies to take into consideration. In addition, the guidelines will be developed for the automotive industry, based on the fact that the automotive industry handles a large amount of parts in order to meet the diverse customer demands. The findings for LVPs based on contextual factors and part characteristics would be applicable for companies within the automotive industry to use as support when designing their parts feeding systems in order to choose the most appropriate policy, thus enhancing the performance of their production system. 1.3 Research questions In order to develop these guidelines for designing the parts feeding policies for LVPs in the automotive industry, three research questions have been developed. The first research question relates to the lack of definition for LVPs in existing literature. It is considered relevant to have a definition for LVPs, which should be adapted for the automotive industry and answer which parts the guidelines are applicable for. Research question 1: How can LVPs be defined for parts feeding in the automotive industry? The second research question is related to the various parts feeding policies that can be found in existing literature and to highlight the advantages and disadvantages with each policy. The parts feeding policies’ suitability for LVPs should also be answered with this research question. To further be able to give recommendations about what parts feeding policy to use for parts within the LVP scope, the different policies’ suitability based on part characteristics - such as consumption volume, size and variety within the part family - will be examined. This is relevant to evaluate since it has been stated in previous literature that it is beneficial to assign different policies based on part characteristics (Caputo & Pelagagge, 2011). Contextual factors influencing the choice of parts feeding policy will also be addressed. Research question 2: Which parts feeding policies are suitable to use for LVPs, and for what part characteristics does each policy fit best? As mentioned in the previous section, there are several constituents within the parts feeding system that impact the performance, thus it is not possible to isolate this study to only consider the design of the parts feeding policy. The third research question should therefore cover how design choices within the other constituents in the parts feeding system could impact the performance of the parts feeding system, as well as how contextual factors influence the design. Research question 3: How should the parts feeding system related to each parts feeding policy for LVPs be designed? 1.4 Scope and limitations The scope of this study is to investigate parts feeding policies and design options related to the parts feeding system that can be found in existing literature and that are applied by the studied companies. The reason is that the guidelines should be based on both theoretical knowledge and take into consideration practical issues, such as contextual factors. Furthermore, parts 4 feeding policies relevant for LVPs are in focus, and therefore the various policies and design of the parts feeding systems will be evaluated accordingly. The scope of this study has also been further narrowed down to solely include companies within the automotive industry, since the importance of parts feeding within this industry has been identified as critical. Moreover, the multiple case study is constituted of Swedish manufacturers. It could be argued that a mix of facilities located in different countries would increase the generalization of the recommendations from this study. This will therefore be further treated in the discussion chapter. Due to limited time and resources, at most one visit to each company’s production facility has been conducted. However, it is believed that the time was sufficient to identify and analyze the various parts feeding policies applied at the companies as well as relevant contextual factors. Furthermore, several of the companies participating in the study have multiple production facilities in Sweden and it is possible that different parts feeding policies are used at the various sites. However, one facility per company has been studied, which could lead to that we miss out on valuable information. In order to visit the most appropriate sites, it was decided in accordance with contact persons at each company regarding which facility that was most relevant to study. It is therefore believed that sufficient information was gathered during the visits in order to develop the guidelines. In addition, the companies participating in the study are considered to contribute with a good mix since they manufacture different types of products, with various production volumes as well as differences in design of their production systems. It is believed to be a requirement to have this mix of companies in the automotive industry since it increases the possibilities to capture a larger amount of different parts feeding policies. In addition, a higher degree of relevant contextual factors existing among the studied companies could be examined. 1.5 Thesis outline The overall structure of the report starts in Chapter 2 with a theoretical framework introducing concepts that will be used throughout the report, mainly covering the different constituents of the parts feeding system. Chapter 3 describes the methodology used in the study for the reader to get a clear picture of how the study has been conducted. Chapter 4 covers the empirical studies, treating the four case companies that were studied for this report, how they have designed their parts feeding systems and how they currently handle LVPs in this aspect. Chapter 5 connects the theory with the empirical findings into an analysis, where arguments will be put forward to be able to answer the research questions about how companies should define LVPs, what policies that are suitable for these parts, as well as what a company should think of when designing a parts feeding system for LVPs. In Chapter 6, the most important findings from the analysis will be presented. A discussion of the results is held in Chapter 7, treating reflections of the findings, contributions to the industry, generalizability and trustworthiness of the study, as well as input for further research that could be needed to provide further insight in the area of LVPs and parts feeding. The conclusions are presented in Chapter 8, and the report ends with references and appendices. 5 2 THEORETICAL FRAMEWORK The theoretical framework begins with describing methods for how classifications of parts can be carried out. This section should form a base for how the classification of LVPs can be performed within the parts feeding system, followed by a description of the parts feeding system and its constituents. The framework narrows down on theory related to parts feeding policies that will be used to evaluate when the policies are suitable for LVPs. In addition, contextual factors impacting the parts feeding system as well as relevant performance areas that the parts feeding system can be evaluated according to are presented. 2.1 Classification of LVPs Within existing literature, there is no unanimous definition of what should be categorized as an LVP. However, van Kampen et al. (2012), state that it is common to perform a classification of a company’s parts since it facilitates a systematic approach for decision-making based on part characteristics such as volume, value, and storage requirements. The usage of classifications could be found in many fields, for instance within the areas of production and operations management, and could be used for several different purposes (van Kampen, et al., 2012). The authors describe that it can be advantageous to use a part classification when deciding production and inventory policies since each class contains parts with similar characteristics, thus reducing the complexity of the control. Hence, LVPs could be grouped together in order to make the parts feeding of low-volume parts as efficient as possible. The ABC classification is a well-known method that is simple to apply since the parts are usually divided according to one criterion (Teunter, et al., 2010). For instance, demand value or consumption volume are measures which the division into groups could be based on (Teunter, et al., 2010; de Koster, et al., 2007). According to Caputo & Pelagagge (2011), a suitable criterion to perform an ABC classification for parts feeding takes volume and cost of the parts into consideration. However, van Kampen et al. (2012) state that the decision of which characteristic to base the ABC classification on depends on the purpose of the classification, but also contextual factors related to specific industries could have an impact. 2.2 The parts feeding system and its constituents In any company, the management of parts plays a large role in how the whole production system operates, in for instance ensuring the optimal flexibility and efficiency for the assembly system (Battini, et al., 2009). Accordingly, Baudin (2002) states that the factor that impacts the productivity of assembly systems the most is parts feeding. Further important parameters that can be significantly impacted by how parts feeding is managed are the costs related to materials handling and inventory (Battini, et al., 2009). The constituents of the parts feeding system differ between various studies. For instance, Sali et al. (2015) define parts feeding as an in-plant logistics process that involves preparation of parts at storage areas and transportation to the lineside presentation. Battini et al. (2009) also include the issue of how many storage areas that is optimal and where they should be placed. However, the theoretical framework treating parts feeding systems in this report is based on the framework presented by Kilic & Durmusoglu (2015), although with some modifications. Kilic & Durmusoglu (2015) have divided the parts feeding system into three constituents, namely storage of parts, transport of parts, and parts feeding policy. This framework has been modified in this thesis by breaking out picking of parts as a separate constituent, which in the framework 6 from Kilic & Durmusoglu (2015) is included in storage of parts, and with the addition of packaging of parts as a constituent, since this has been identified as an important factor impacting the parts feeding system. An illustration of the framework adapted for this study is presented in Figure 2 below. The choice of adding packaging to our description of the parts feeding system is due to the big impact that choosing the correct packaging size can have on both the parts feeding system, as well as on the assembly system. Medbo & Wänström (2009) express that choosing a smaller packaging size fitting the assembly context for the part could significantly improve flexibility and efficiency in the system. Picking of parts is described as a separate constituent due to its importance especially for the parts feeding policies of kitting and sequencing that will be presented in the following sections. Brynzér & Johansson (1995) state that picking of parts accounts for a large share of the time that operators spend to prepare the material, and therefore is important to consider when designing the parts feeding system. All these five constituents are intertwined and affect each other as well as the performance of the total parts feeding system and production system. Therefore, it is necessary to address all constituents of the parts feeding system when choosing and designing parts feeding policies for LVPs. Figure 2. The parts feeding system and its constituents – Adapted from Kilic & Durmusoglu (2015) 2.2.1 Parts feeding policies This chapter will treat the focus area of this study, namely different parts feeding policies. According to Kilic & Durmusoglu (2015), the parts feeding policy is defined as “the method of delivering parts to the usage points”. By studying the existing literature, several parts feeding policies have been found that are used in industrial companies. Some of the parts feeding policies described below can certainly be used on their own as the only method for delivering parts to the lineside presentation, and many authors have treated their reviews and analyses on this premise. While this could probably work for some companies, a more appropriate and cost Parts Feeding System Parts Feeding Policies Storage of Parts Packaging of Parts Transportation of Parts Picking of Parts 7 efficient way of choosing among parts feeding policies is to use several different policies for different products depending on the part characteristics, or by dividing the products through an ABC classification (Caputo & Pelagagge, 2011). There are several part characteristics that have been mentioned in previous studies as relevant to consider when choosing the parts feeding policy. Some that have been identified are variety in part family (Caputo & Pelagagge, 2011; Hua & Johnson, 2010), part size (Caputo & Pelagagge, 2011; Ding, 1992; Hua & Johnson, 2010; Limère, et al., 2012), consumption volume (Caputo & Pelagagge, 2011; Hua & Johnson, 2010), and part value (Caputo & Pelagagge, 2011; Schwind, 1992). In the following sections, the most commonly used parts feeding policies will be presented. They are, in order of appearance in the report, continuous supply, kanban-based continuous supply, kitting and sequencing. Each policy will be described with their characteristics as well as some advantages and disadvantages, and in what context and for what part characteristics the policy is most suitable to use. Since existing literature related to parts feeding policies lacks a focus on LVPs, the descriptions of the policies will mostly be on a general level. In the end of this section of the report, a table that summarizes the described policies’ suitability based on the above-mentioned part characteristics is presented. Continuous supply Continuous supply, also called line stocking, is a very common parts feeding policy in companies (Limère, et al., 2012), and is often used in practice to feed parts to assembly lines (Sali, et al., 2015). Sali et al. (2015) explain continuous supply as that all parts are stored in both a central warehouse or a preparation area, as well as at the lineside presentation, and the parts are replenished at a given time interval corresponding to a certain amount of takts. The term also includes how material is presented at the assembly line, with the main feature that parts are stored according to their part numbers, and that all part numbers are always presented at the assembly line (Hanson, 2012; Johansson, 1991). Most often, parts fed by continuous supply are stored at the lineside presentation in their original packaging (Limère, et al., 2012). Parts that are needed for a specific assembly process are then picked by the assembly operators, directly from these unit loads (Johansson, 1991). The downsides of continuous supply are mainly the space requirement for storing all parts at the assembly line and that some parts might need to be stored at multiple workstations at the assembly line, requiring even more space (Hua & Johnson, 2010). There will also be problems when there are increasing cycle times, due to more material having to be presented at each assembly workstation, making continuous supply a less useful parts feeding policy (Johansson & Johansson, 2006). Furthermore, this parts feeding policy often requires handling by forklift, which could be a risk in areas where there are many operators. Additionally, due to the nature of continuous supply where parts are presented in their original packaging and not suitably presented for the assembly operations, the picking time for operators could be long and the proportion of non-value adding work is at risk of becoming high, especially for parts in large packaging such as pallets (Medbo & Wänström, 2009; Finnsgård, et al., 2011). However, continuous supply is very efficient in terms of that the parts usually not needing any rework or repackaging from the supplier, but can be put in place directly at the lineside presentation (Hua & Johnson, 2010). Some work could be needed though, depending on how the material is received from the supplier and how it is wanted to be exposed at the assembly line. For example, 8 parts being received in pallets but presented in smaller boxes at the assembly line need to be repackaged, a procedure that is called downsizing (Sali, et al., 2015; Johansson, 1991). According to Sali et al. (2015) continuous supply fits best for supplying an assembly line with larger parts with low variety in part family. Hua & Johnson (2010) argue that continuous supply is more suitable for parts consumed in high volume. Furthermore, the policy is not recommended for high value goods since it increases the work-in-process at the assembly line (Caputo & Pelagagge, 2011). Kanban-based continuous supply Another variant, or further development, of continuous supply is the kanban-based continuous supply. In this policy, parts are fed in standardized storage containers, such as bins, from storage areas located close to the assembly line to the lineside presentation (Faccio, 2014). The replenishment can be triggered in many ways, for instance through empty bins or kanban cards, which could either be physical or electronic (Jonsson & Mattsson, 2009). The greatest advantage of kanban-based continuous supply is that parts can be frequently delivered in small quantities from the storage areas located close to the assembly line, which makes it possible to reduce inventory at the lineside presentation by using smaller unit loads, and avoid long delivery distances from central warehouses (Faccio, 2014). Furthermore, is it a highly visible and reliable system, and cheap to implement (Kouri, et al., 2008). A downside for kanban-based continuous supply is the stock-out risk resulting from high variability in the consumption of parts, which further requires higher safety stock at the assembly stations, inducing higher costs (Faccio, 2014). Jonsson & Mattsson (2009) also argue that kanban should be used in a stable environment in terms of e.g. lead times and demand. Caputo & Pelagagge (2011) argue that kanban-based continuous supply would not be suitable in a mixed-model assembly system, where the variation of products is high and the demand is low. Additionally, the same authors state that kanban-based continuous supply enables reduced work-in-process compared with continuous supply, thus it can be argued that the policy is more appropriate for parts with higher value. Kitting The parts feeding policy referred to as kitting is rather different from the above-mentioned policies. Kitting treats parts feeding to assembly line by delivering a so called kit with parts corresponding to the exact requirement at one or several assembly stations or for a complete end product (Limère, et al., 2012). The kit is prepared by putting the required parts into a kit carrier through picking from storage in a decentralized storage area or other kitting area (Faccio, 2014). The kit is then delivered to the assembly line in the sequence the kits should be assembled. There are two types of kits that can be used at the assembly lines, namely stationary kits and travelling kits. Kits are consumed in accordance to takt time, and for stationary kits one takt equals one kit. This makes it easy to forecast the consumption of parts and need for replenishment of kits (Limère, et al., 2012). A stationary kit is sent to a certain workstation after it has been put together and stays at the workstation until all parts have been used for the product, which then travels along to the next workstation where another kit awaits (Bozer & McGinnis, 1992). A travelling kit, however, follows the product through several workstations 9 where parts are taken from the kit and put on the product in a certain sequence, until it is emptied (Limère, et al., 2012). According to Bozer & McGinnis (1992), advantages with kitting is the increased visibility and control over work-in-process and the possibility to reduce storage space at the lineside presentation. Further advantages involve a possible increase in product quality and productivity at the workstation, since kits can present the parts in an effective way as an instruction, which induces learning for the assembly operators (Johansson, 1991). This also means that assembly operators could spend less time on fetching and searching for the required parts (Johansson, 1991; Medbo & Wänström, 2009). A main property of kitting is that it facilitates a wide variety of products, due to the ease of changing parts in every kit (Bozer & McGinnis, 1992), which also facilitates part introductions. Kitting is also advantageous to use when there are several workstations requiring the same parts, when the assembly system handles a large number of different parts, when the parts handled are of high economic value (Caputo & Pelagagge, 2011), and when the individual part volumes are low (Hua & Johnson, 2010). The main disadvantage of kitting on the other hand is the required labor-intensive assembly of the kits before moving them to the assembly line (Bozer & McGinnis, 1992). The kit preparation may also require additional storage space (Bozer & McGinnis, 1992). Furthermore, problems in assembly could occur when the quality of the kit is not satisfactory, since kitting entails that a defect product is not easily replaced (Limère, et al., 2012). In addition, there are restrictions concerning weight, volume and size for the parts delivered through kits, so not all parts are able to be kitted (Limère, et al., 2012). However, this depends on what materials handling equipment and kit containers the companies are using, or are able to implement into their parts feeding system. Sequencing The last parts feeding policy that will be described in this study is sequencing, or sequential supply. This policy could be seen as a special kind of stationary kit, where each kit consists of only one part (Sali, et al., 2015). The parts needed for a determined number of products are presented in the correct order they should be assembled on the products (Johansson & Johansson, 2006). The sequenced parts are typically transported and displayed at the lineside presentation in specially adapted unit loads (Sali, et al., 2015). Since sequencing can be seen as a certain kind of kitting, the policies share several benefits and drawbacks. For instance, sequencing is often preferred for parts that have a high number of varieties. However, in contrast to kitting, sequencing is more suitable for larger parts, which was shown in the study by Sali et al. (2015), where factors such as walking distances for assembly operators and preparation costs were considered. In addition, sequencing can be used if there are only a few components being assembled at every workstation (Johansson & Johansson, 2006). An advantage with sequencing compared to continuous supply, is that a higher space efficiency could be achieved at the lineside presentation for the sequenced parts (Hanson, 2012). Disadvantages with sequencing include the need for a specific preparation area as well as the requirement of specific operators for sequencing the parts into the transportation units. In addition, some sequentially fed parts might require special packaging solutions (Johansson & Johansson, 2006). 10 Summary of parts feeding policies’ suitability based on part characteristics Table 1 below summarizes the findings from previous studies regarding when the different parts feeding policies could be suitable to use based on part characteristics that have been identified in existing literature, namely variety in part family, part size, consumption volume and part value. It should be noted that the table is a simplified view of the reality, and should only be seen as a guideline, since contextual factors might influence the suitability of the parts feeding policies. Each part characteristic has been categorized into three segments. The table shows which segments of the part characteristics that each parts feeding policy is most suitable for according to findings from existing literature, which was mentioned within each section above. The grading has been performed by the authors in order to make the comparison more standardized, however, it should reflect the conclusions from previous findings. The evaluation of whether the policies are suitable for LVPs is mainly based on part volume, where continuous supply tends to be less preferable to use as parts feeding policy for LVPs, while kitting and sequencing and to some extent kanban-based continuous supply could be more suitable. This reasoning will be further developed in chapter 5.2 where empirical findings also are taken into consideration. Table 1. Parts feeding policies’ suitability based on part characteristics according to findings in existing theory Parts feeding Part policy characteristics Continuous supply Kanban-based cont. supply Kitting Sequencing Variety in part family Low Low-Medium High High Size Medium-Large Small-Medium Small-Medium Large Consumption volume High Medium-High Low Low Value Low Medium-High High High Suitability for LVPs Low Medium High High 2.2.2 Packaging of parts The second constituent in the framework related to parts feeding system is packaging. Hanson & Finnsgård (2014) mention that there is often thought to be a tradeoff between large unit loads, which enable a high efficiency in the parts feeding system versus smaller unit loads which enable a more efficient assembly system. Unit loads can for instance be containers or pallets. By having larger unit loads, the frequency of replenishment and thereby the number of transports of parts being made can be reduced (Hanson & Finnsgård, 2014). However, large unit loads will also lead to higher work-in-process and a requirement of more space being available at the point-of-use (Hales & Andersen, 2001). Most companies in Sweden mainly base their choice of packaging on cost and transportation efficiency, resulting in large unit loads (Medbo & Wänström, 2009). On the other hand, Hanson & Finnsgård (2014) conclude that the efficiency of the parts feeding system doesn’t necessarily depend on unit load size, since smaller unit loads also can provide a high efficiency of the parts feeding system. This is due to the possibility to transport a high number of different parts simultaneously with a tow train. The use of smaller containers further allows for easier and faster replenishment of unit loads, since it is easy to replace the containers at the lineside presentation. This is not possible when using pallets as unit loads, since the 11 replacement has to be made with a forklift or other material handling equipment (Hanson & Finnsgård, 2014). Medbo & Wänström (2009) argue that another advantage with small unit loads is that a higher flexibility at the lineside presentation can be achieved, since a higher variety of material can be presented at the same time in close proximity to the assembly operators, allowing more efficient fetching of parts. The process of repacking parts, mainly by downsizing to smaller unit loads, is a very inefficient activity that should be avoided (ten Hompel & Schmidt, 2007). It is therefore important that companies, together with the suppliers, strive towards choosing the proper packaging for parts in which they will be presented at their place of consumption. However, supplier contracts and fill-rates in the delivery trucks must also be considered, and might limit the choices of packaging solutions (Hanson & Finnsgård, 2014). Table 2 summarizes the relevant design options related to packaging. Table 2. Description of design options related to the packaging constituent Constituent Design area Design option Packaging Unit load size Choosing an appropriate packaging size for the part. For instance, pallets, boxes or cartons. Repacking Moving parts from one unit load to another. Inefficient process that should be avoided. 2.2.3 Storage of parts The third constituent in the framework related to parts feeding system is storage of parts. Storage space at the assembly lines is mostly a great restriction for companies today as a consequence of the high variety of parts, and it is also the most expensive storage area (Emde & Boysen, 2012). There is a trade-off between having material available at the assembly stations when it is needed to avoid stoppages, versus keeping stock and traffic at the assembly line to a minimum level to avoid high material handling and holding cost (Emde & Boysen, 2012). Storage of parts has traditionally been centralized in production facilities, and parts have been transported one unit load at a time when replenishment has been needed at the assembly stations (Emde & Boysen, 2012). For this to work and not to have excessive traffic in the facility, the parts have to be delivered in large lots, which increases work-in-process and takes up a lot of space at the assembly stations (Emde & Boysen, 2012). Battini et al. (2009) argue that the centralized storage configuration can reduce inventory cost by having most material in only a few places, but that material handling costs will increase due to a high amount of transport from the centralized storage area. This will also impact the flexibility at the assembly line negatively. Another possibility is to instead have a decentralized storage configuration where parts are supplied from supermarkets to the assembly line. This entails shorter delivery times when material could be stored closer to the assembly line, consolidating material to match the need at the assembly line and at the correct time that it is needed (Emde & Boysen, 2012). By having a decentralized storage configuration, material can be presented in smaller unit loads, which will increase the accessibility for the assembly operators, as well as decrease picking time and improve ergonomics (Emde & Boysen, 2012). The downsides of supermarkets are mainly that they require space in the facility and that the material is stored in a less efficient manner than 12 in a central storage area, due to that most material should be easily accessible for the operators (Emde & Boysen, 2012). As a contrast to the centralized storage configuration, decentralization leads to higher flexibility at the assembly line and less transportation, but higher inventory costs (Battini, et al., 2009). There are several storage policies mentioned in existing literature for how parts should be stored in the different storage areas, where random storage, class-based storage and family grouping are three highlighted policies mentioned by de Koster et al. (2007). Random storage means that a unit load can be stored in any empty slot in the storage area, allowing for high space utilization but an increased travelling distance, when parts consumed in high volumes could be stored in less preferred storage locations. The class-based storage policy divides the parts into different categories based on for instance an ABC classification and assigns them to a dedicated storage area for each class. The parts that are picked most often could be stored closest to the depot in order to reduce travel distance, while LVPs could be placed in less easily accessible areas. When applying family grouping, parts that are often picked together are stored close to each other. This method could be related to Brynzér & Johansson’s (1995) recommendation that the product structure could be used as support when assigning storage policies. Table 3 summarizes the design options related to storage of parts. Table 3. Description of design options related to the storage constituent Constituent Design area Design option Storage Storage configuration Centralized or decentralized storage of parts Storage policy Random, class-based or family grouping 2.2.4 Transportation of parts The next constituent of the parts feeding system is transportation of parts. The transportation process is defined by Sali et al. (2015) as “collecting items prepared and delivering them to their point-of-use…”. Baudin (2004) makes a distinction between in-plant transportation and inbound & outbound transportation regarding where the most improvement potential can be achieved. For in-plant transportation, which this study focuses on, Baudin (2004) states that it is of higher relevance to reduce the number of trips rather than reducing the distance travelled on the trips. Hence, it is beneficial to perform transports of multiple parts in the same delivery tour instead of having direct transports. According to Baudin (2004), some common material handling equipment used in production facilities are forklifts, tow trains, push carts, and pallet jacks. The various handling equipment have their advantages and disadvantages and therefore the usage should reflect the requirements in terms of load size and frequency. According to Cottyn et al. (2008), the forklift has traditionally been the most common equipment used for in-plant transportation due to its flexibility. However, the forklift is adapted for carrying large and heavy parts in pallets directly from a storage area to the point-of-use, and is less appropriate to use for smaller unit loads (Baudin, 2004). In addition, the forklift is a safety risk for both employees and material causing many companies to evolve towards having a forklift free environment (Cottyn, et al., 2008). In comparison with the forklift, the tow train has higher capacity and is adapted for smaller unit loads (Baudin, 2004). They are common to use for indirect transport routes called milk runs, where the tow train travels in a loop on a fixed route from a decentralized storage location to 13 several workstations, transporting multiple parts on each trip. The tow train is operator driven and consists of a tractor and several wagons carrying the materials, arranged in the order of delivery and usage at the assembly line (Sali, et al., 2015). The push cart is a low cost alternative that is adapted for transporting smaller unit loads such as bins (Baudin, 2004). The pallet jack is another equipment which is cheap compared to the traditional forklift. However, it is primarily designed for short movements and it cannot perform vertical lifts (Baudin, 2004). Table 4 summarizes the design options related to transportation of parts. Table 4. Description of design options related to the transportation constituent Constituent Design area Design option Transportation Material handling equipment Forklift, tow train, push cart, pallet jack etc. Routing Direct transport or indirect transport (milk-run) from storage area to lineside presentation 2.2.5 Picking of parts The last constituent of the parts feeding system being presented in this report is picking of parts, which could be performed both in warehouses and in specific picking areas in assembly facilities. The picking process in warehouses is important to consider because it is an expensive process that is labor-intensive (de Koster, et al., 2007). In the context of parts feeding systems, picking is primarily related to the kitting and sequencing processes. According to Brynzér & Johansson (1995) the design of the picking system is an important factor that impacts the overall performance of a kitting system. Brynzér & Johansson (1995) evaluate the performance of the picking process in terms of picking productivity and picking accuracy. The study highlights that the industry has put more emphasis on improving the productivity aspect while the picking accuracy has had lower focus, even though it could cause more severe problems in the production process. In the context of picking in warehouses, approximately half of the order picker’s time in a manual picking system consists of travelling (de Koster, et al., 2007). A method to reduce the total travelling distance is to apply batching of orders. It aligns to some extent with Brynzér & Johansson’s (1995) conclusions as they state that batching of multiple kits could improve the picking efficiency unless it does not increase the amount of sorting and administrative work. Another recommendation is to implement storage policies within the area where picking is performed. A few storage policies mentioned in the literature, and their advantages, have been presented in section 2.2.3. Within the study performed by Brynzér & Johansson (1995), frequently occurring issues related to picking accuracy were reading mistakes, interruptions of the picker, and mixing of parts in the batch. Thus, batching could increase the picking productivity, while it decreases the picking accuracy. A method which could improve the picking accuracy is to manually count the picked parts after the picking tour, however, there are more appropriate arrangements for improving the quality, such as improved picking information (Brynzér & Johansson, 1995). This aligns with the conclusions of Fager et al. (2014), stating that the picking information impacts the picking quality for materials preparation for kitting and sequencing. Traditionally, the picking information has consisted of a pick list, which informs the picker which parts that should be 14 collected (Brynzér & Johansson, 1995), but other information methods exist such as pick-by- voice and pick-by-light (Fager, et al., 2014). According to ten Hompel and Schmidt (2007), the pick-by-voice method is superior from a picking accuracy perspective, while the pick-by-light and pick list shows approximately similar accuracy levels. From a picking efficiency perspective, ten Hompel & Schmidt (2007) state that a disadvantage with the usage of pick list is the time needed to identify the next part to pick. The picking of parts for kit preparation could be performed either by specific kitting personnel or by the assembly operators. Brynzér & Johansson (1995) argue that it is generally more beneficial to assign the kit preparation to the assembly operators if the process is located close to the assembly line since balancing problems can be reduced and higher picking accuracy can be obtained as the operators are responsible for the whole job. However, if the process is far away from the assembly line, cost could often be held to a minimum if specific kitting personnel have responsibility for the assembly of the kits, since assembly and administrative work could be bundled with other storage and material handling activities (Johansson, 1991). Table 5 summarizes the design options related to picking. Table 5. Description of design options related to the picking constituent Constituent Design area Design option Picking Picking quantity Picking for one or several orders during one tour Picking information Pick list, Pick-by-light, Pick-by-voice Picking location Close to or far away from lineside presentation Responsible for picking Specific department or assembly operators 2.3 Contextual factors impacting the design of the parts feeding system Apart from what has been described in earlier sections as part characteristics, there are also further contextual factors affecting the design of parts feeding policies. These are all important to consider, as e.g. cost could differ much depending on in what context a certain policy is used (Sali, et al., 2015). In addition to the part characteristics, Hanson (2012) divides contextual factors into production-related factors and layout-related factors. A similar structure will be adapted in this section. Limère et al. (2012) state that the parts feeding system should be adjusted to the assembly system in order to enhance the performance, making the design of the assembly system an important contextual factor to consider. One production-related factor is the assembly cycle time which can vary between assembly systems. According to Medbo (2003), a longer cycle time increases the space requirements if all parts are supplied with continuous supply. Another important factor that influences the choice of parts feeding policy is the use of a certain part at several different workstations and for several end products, and the deduced need of presenting parts at multiple workstations at the assembly line (Hua & Johnson, 2010). Furthermore, the fact that the companies are operating within the automotive industry is an important contextual factor in itself. As has been said before, customer demand in the automotive industry is highly variable which means that companies in this industry must be very flexible (Pil & Holweg, 2004). This influences how they design their assembly and parts feeding system, with for example mixed-model assembly lines and extensive customization choices. 15 Layout-related contextual factors can also impact the design of the parts feeding system where space restrictions could cause limitations in the facility (Hua & Johnson, 2010). Battini et al. (2009) mention distances between storage areas and the assembly line as another aspect which could be restricted by the production facility. The design of the assembly stations may impact the space restrictions for displaying all parts at the lineside presentation. For instance, if there is not sufficient space available at the lineside presentation to display all parts, it is not possible to use continuous supply consistently (Caputo & Pelagagge, 2011). Additional factors that are not to forget, is firstly that the personnel might be resistant to change, and that they might have certain preconceptions about the different parts feeding policies (Hua & Johnson, 2010). Secondly, company goals may play a role, where for example wanting to reduce non-value adding tasks might lead to the company choosing a certain parts feeding policy (Hua & Johnson, 2010). 2.4 How to assess the performance of the parts feeding system for LVPs Performance assessment is important for the evaluation of how the parts feeding system is working, as well as to be able to assess how changes in the system affect the performance, in order to be able to improve the system further (Hua & Johnson, 2010). To evaluate how well a parts feeding system operates, existing literature have several different suggestions on performance areas that can be used. Space requirements and work-in-process in the facility are mentioned by several authors as relevant performance areas to evaluate since different parts feeding policies yield different advantages and disadvantages in these areas (Bozer & McGinnis, 1992; Field, 1997; Hanson & Brolin, 2013; Hua & Johnson, 2010; Limère, et al., 2012). Another performance area that can be assessed is the flexibility in the assembly system with regards to handling a wide variety of parts and fluctuating production volumes (Hanson & Brolin, 2013; Medbo & Wänström, 2009). According to Baudin (2002) an important factor to account for and measure, especially in mixed-model assembly systems, is picking errors, which could be seen as the most common source of defects in assembly operations. This is in line with Hua & Johnson’s (2010) performance area of component selection error. Hua & Johnson (2010) also mention the amount of material handling required as a relevant performance area to assess. Hanson & Brolin (2013) also evaluate the man-hour consumption required between usage of different parts feeding policies. 16 17 3 METHODOLOGY In this chapter of the report, theory about research methods and their application is going to be discussed, such as different approaches to theory creation, quantitative versus qualitative studies and different designs available to perform a research study. The chapter aims to investigate what different research methods there are and how they should be used as well as to, from this information, draw conclusions about the correct and fitting methods for this particular study and to give the full description of the methodology that has been used. 3.1 Research approach The research approach used in a research study describes in what way the theory used has been created, collected or evolved. Holme et al. (1997) explain that theory could be seen in two ways; either the theory is very vague and it must be evolved throughout the study, or the theory might be very precise and treats exactly what is meant to be investigated in the study. There are mainly two terms present to describe the creation of theory and its connection to empirical findings, called deductive method and inductive method respectively. In the deductive method, the theory lies as a basis for the empirical observations and findings (Bryman & Bell, 2011), and hypotheses are often made from theory in order to be proven by the empirical studies (Holme, et al., 1997). This also means that further studies and research could be made from the results. Borrego et al. (2009) put it that theory is meant to justify variables used in the study as well as the purpose statement and research questions, which should be narrowly specified. In the other end of the spectrum, there is the inductive method, where the empirical data and findings mainly are the basis for the theoretical framework (Bryman & Bell, 2011). Here, it is important that the framework matches the specific empirical situation and context that has been studied (Holme, et al., 1997). The deductive method is the more formalized theory of the two, and is easier to create specific guidelines for (Holme, et al., 1997). However, most studies are not exclusively following only one of the two methods, but instead there is often some interaction between the two, perhaps having the deductive method with influences from the inductive method as the most commonly used mix. Holme et al. (1997) express that the combination of the two methods is very important, and that it is when they meet that creativity is created. Bryman & Bell (2011) also mention a mix between deductive and inductive methods, where data and theory is weaved back and forth and simultaneously changing each other to create a better focus for the study. They call this method iterative, as theory and data are continuously revisited. A very similar approach, although with a different name, is proposed by Dubois & Gadde (2002). This approach, called systematic combining, is according to them mainly used in case studies, and they describe the approach as continuous movement between theory and empirical findings, together influencing the process of the research study. The method usually means a better understanding of both the theory and the empirical observations, which can be used for a higher level research study (Dubois & Gadde, 2002). The method mainly used in this study is the deductive method, but with some influences from inductive method, which also is the most common approach in many other case studies. The study started with a literature review in order to enhance the knowledge within the research 18 area, and to be able to develop the aim and the research questions. After this was done, data was collected from case companies to see whether the reality corresponded to theory. In addition, the process was then somewhat iterative, as findings in the case companies affected the formatting of the research questions, and further studies of literature. 3.2 Research strategy Another expression in research methodology is research strategy, treating the distinction between quantitative and qualitative studies. The two strategies have the same purpose in a study, which is to propose a description of our society and how entities in the society affect and influence each other, but the way to fulfill this purpose is different (Holme, et al., 1997). The overall definition of quantitative studies is that it involves measurement in one way or another (Bryman & Bell, 2011), and that information is converted into numbers in order to be able to make statistical analyses (Holme, et al., 1997). Borrego et al. (2009) express the purpose of a quantitative study to generalize findings in a sample over a larger area of research through statistical analysis. Holme et al. (1997) mean that a strength of the quantitative study is that the researcher is able to cover a large amount of subjects by having standardized questions. This is often required to increase the generalizability of the study, since research subjects are often chosen randomly in a quantitative study (Eisenhardt, 1989). Furthermore, a weakness with this kind of strategy is that the study won’t give any data on the social processes in the studied units (Holme, et al., 1997). According to both Bryman & Bell (2011) and Borrego et al. (2009), the quantitative research strategy is usually a good fit for a strict deductive approach described above, where the theory through several hypotheses is tested with help from empirical data. The qualitative research strategy, which is defined by Holme et al. (1997) as that information is interpreted by the researcher, that his/her apprehension of the studied subjects is in focus, and that this information cannot or should not be converted into numbers. According to Borrego et al. (2009), a qualitative study is characterized by the collection and analysis of mainly textual data, meaning activities such as conducting and analyzing surveys and interviews, and attending at and observing the research subjects. The authors further explain that qualitative studies also can be used to generalize findings, but that the descriptions are connected to specific contexts, and that the readers can make their own conclusions about how the findings can be applied in their situation. Strengths with qualitative studies are that they give the overall picture of the case and that they are flexible in their design (Holme, et al., 1997). A weakness, however, is that because case studies in a qualitative manner might be very time consuming, not as many units can be investigated as in a quantitative study (Holme, et al., 1997). Bryman & Bell (2011) mean that qualitative studies are mostly suited for an inductive approach, where theory is generated through the findings of the empirical data, but also say that there are several examples where qualitative studies have tested the theory rather than to generate it. Borrego et al. (2009) mean that, in order to be able to completely and correctly answer the research questions in a qualitative case study, the data has to go through a so called thick description, meaning rich, contextual descriptions. Yin (2014) on the other hand, means that this does not have to be the case, because a case study might be in a mix of qualitative and quantitative strategy, each providing their explanations of the case. The research strategy used in this study is mainly qualitative, where the researchers’ interpretation and analysis stand as a basis for answering the research questions. The number 19 of companies taking part in this study is set at four, which makes it hard to perform a quantitative study, where a high number of participants would be preferred in order to give a bigger data sample. Another reason to choose a qualitative strategy is that due to the nature of the research questions, it is hard to find specific quantitative measurements answering the question of what contextual factors influence the decision of what parts feeding policy to use. The study’s aim is instead to through observation assess whether the right feeding policies are used or what feeding policy would fit best in the specific context. This is clearly a qualitative process, where interpretation and apprehension are in focus. 3.3 Research design Moving further within research methods, there is the question of what research design to use in order to best answer the research questions. According to Bryman and Bell (2011), there are five different kinds of research designs, namely: experimental design; cross-sectional or social survey design; longitudinal design; case study design; and comparative design. All of these designs have their advantages and disadvantages and all fit in different situations and contexts. Since this study will treat four different cases, at four separate companies, the most suitable research designs for this specific study are case study design and comparative design. These will be described below. A case study is a detailed and thorough analysis of a single case, such as an organization, a location, a person or an event, and case studies are extensively used in business research (Bryman & Bell, 2011). Eisenhardt (1989) describes a case study as focusing on understanding the dynamics of a specific context, and that they usually combine several data collection methods such as interviews and observations. Case studies could generate results that are both qualitative and quantitative, or a mix of both, and could be used to provide a description of the research subject, to test theory or to generate theory (Eisenhardt, 1989). A weakness with a case study could be that due to the specificity of this design, it does not give enough information for a generalization to be made to more than perhaps a specific situation (Dubois & Gadde, 2002; Yin, 2014). Case studies can involve both a single case or multiple cases, and several levels of analysis (Yin, 2014). When multiple cases are involved in the study and the strategy is mainly qualitative it is called a multiple-case study (Bryman & Bell, 2011). This is an extension of the case study design, and could also be seen as a comparative design. The comparative design improves the creation of theory and makes it easier to see whether a theory holds or not. There could, however, be a weakness here resulting in that the researcher only focuses on the differences between the cases and does not see the specific contexts of each case (Bryman & Bell, 2011). This study consists of, as mentioned before, four separate case studies, and is primarily of qualitative nature. Apart from studying each case for itself to evaluate every company in its own context, a comparative study has been performed to find similarities and differences among the companies as well as between the companies and the theory. 3.4 Work procedure This section will describe general views of how a research work procedure is made, as well as how the specific work procedure of the case studies looked like. The different research methods used in this study will be treated and presented in the following order: firstly, the literature 20 review and how current theory has been processed; secondly, data collection with information regarding the case companies and how this has been obtained; and lastly, data analysis concerning how the case companies have been analyzed and how the results have been connected to the current theory. 3.4.1 Literature review To be able to make a thorough investigation of the situation regarding parts feeding policies, an extensive literature review has been made. It is, according to Eisenhardt (1989) important to process a wide range of literature to be able to build a theoretical framework and to compare to the empirical findings for an interesting analysis of the study. Literature with both similar and contradictory findings are important to include in the framework, in order for the researchers to draw their own conclusions about the situation in the specific case and to show the readers that all options have been considered, and to make the study more valid and easier to generalize (Eisenhardt, 1989). A literature study has been performed in order to fully understand the theory related to parts feeding systems and what design options there are, especially for LVPs. The theoretical framework does also provide a greater knowledge of what conclusions that previously have been drawn regarding the existing parts feeding policies. 3.4.2 Data collection According to Yin (2014) there are six sources of evidence for data collection: documentation, archival records, interviews, direct observations, participant observation, and physical artifacts. These sources give different kinds of information, and a good case study should use as many of them as possible (Yin, 2014). The sources relevant for this study will be discussed below. In this study, an empirical study has been performed in order to build up the awareness of what kind of parts feeding policies that are applied for LVPs in companies today. Four Swedish companies within the automotive industry have been analyzed in the study. Data from the involved companies has been gathered through semi-structured interviews, both via telephone and in person, and visits at three of the four companies’ production facilities. In addition, follow-up questions were asked to the participating companies over telephone and email after the visits at the companies’ production facilities. At each participating company, the interviewed employee has been considered to be well familiar with the current design of the respective parts feeding system. The data collection was also complemented by internal documents such as existing guidelines related to parts feeding. This data has given an insight in how the parts feeding is performed in the studied facilities, as well as the reasons behind their choices of the parts feeding policies and whether the current policies are actually suitable for their specific context. This data has also made it possible to draw conclusions about similarities and differences between the studied companies, conducted as a comparative study. Interviews The initial contact with the companies was made via e-mail and a first phone meeting was held with each company taking part in the study. During this meeting, the researchers introduced themselves and the project, and the company contact was given the chance to explain how they perceived the project, what results they were expecting and how they could support the researchers in the work of the study. This first phone meeting was mainly unstructured, with 21 only some notes guiding the conversation. No real interview questions were asked, since the meeting was primarily for the researchers and the company contacts to get acquainted. During this phone meeting, the structure for future contact was discussed, where the researchers presented a request of having a phone interview within a couple of weeks, as well as a study visit at each company’s production facility a few weeks after that. The interviews were scheduled a few weeks in advance to give time for the researchers to make themselves acquainted with the current theory, to make sure that the correct questions would be asked. This aligns with Leech (2002), stating that what information you currently possess will decide what questions you will ask. Hence, starting with a wide knowledge base means that it will be easier to attain relevant and valid information from the interviewees, which will enable more thorough analysis and conclusions to be made. Interviews conducted in case studies are often of the semi-structured type, as were the phone interviews in this study. Yin (2014) describes semi-structured interviews as guided conversations where the researchers both have to follow a question protocol and at the same time ask further questions to deepen the answers from the interviewee, and make sure that the conversation has a friendly tone. Semi-structured interviews are a mix between unstructured interviews (conversations) and structured interviews (surveys), and are helpful to provide detail, depth and a perspective from within the interviewee’s situation (Leech, 2002). The questions were sent to the companies prior to the phone interviews for the interviewees to be able to prepare, and the questions for the semi-structured interviews can be found in Appendix II. Direct observations To better help the researchers understand the full picture of the parts feeding systems at the companies, and to be able to see what might not have been brought up or kept unsaid during interviews, direct observations were made in the form of study visits at three of the four companies. It was not possible to visit the facility of one company, later denoted as Company B, due to restrictions regarding visits for non-employees during the period this study was conducted. Direct observations carried out in proximity to the interviews are according to Yin (2014) useful to provide additional information about the studied topic, and to increase the reliability of the observations it is important that multiple researchers are present for the observations. During the study visits, both researchers were always present, in order to get the most from the visits, and to be able to discuss and analyze what had been observed. During the study visits, the researchers followed the parts feeding process of primarily LVPs, and how the most important constituents of the parts feeding system were designed at each case company. The contact person at each studied company led the walking tour in the respective visited facilities, and as specific questions arose, personnel in the different areas of the facilities were asked in order to get the most qualitative data for the study. Documentation In order to get a better understanding of the companies as such, and specifically about the parts feeding system and the policies used for LVPs, the companies shared documents and presentations with the researchers during and after the study visits. This data was studied as a complement to the interviews and study visits in order to get more detailed knowledge of the companies and their processes. 22 Yin (2014) illuminates the importance of knowing that documents do not show all the truth, since they may have been edited or describe something a bit different from how the reality actually looks like. Company documentation should therefore be used carefully, and only as a support to other sources of evidence. These recommendations were followed by the researchers, as documents were only complements to the direct observations and semi-structured interviews, and were continuously discussed with the interviewees. 3.4.3 Data analysis An analysis approach described by Eisenhardt (1989) for case studies explicitly is the so-called within-case analysis. It involves the detailed description of each case, which is very important to create insight, and to become very familiar with each case as its own unit. This is helpful for the researchers to see the contexts of each company and its situation, to compare between the cases, and later also to generalize patterns across the cases (Eisenhardt, 1989). To help get the data sorted and to make the analysis easier and higher in quality, it could be a good idea to compare answers from interviewees, to put collected data into arrays and matrices, to create flowcharts and other graphics, and to sort data into chronological order to easier see contexts and sequences in events (Yin, 2014). This study will, as described above, give a thorough description of each case company – how they work, how their parts feeding policies are designed, and which contextual factors that can be identified. Given this data from four different companies, the analysis will compare current theory from the theoretical framework with the reality in all the case companies, as well as a comparison between the four companies. The analysis will also treat what the most appropriate parts feeding policy would be, based on different part characteristics by connecting current theory with empirical findings and the interpretations and thoughts from the researchers. 3.5 Research quality To make sure that the quality of the research study and report is at a high level and that the results could be trusted, it is important that the reliability and validity of the study are robust. The terms reliability and validity primarily concern quantitative studies, although it could be applied to qualitative studies through some re-work of the definitions (Bryman & Bell, 2011). Reliability refers to if the results in the study are repeatable, if the measures are stable and if the results are consistent (Bryman & Bell, 2011). This is important in a quantitative study, but not as much in a qualitative study, since it’s not as important to get the statistical average as it is to understand the whole picture (Holme, et al., 1997). Validity is according to Bryman & Bell (2011) the most important criterion, and refers to the level of integrity of the conclusions that are generated in the study. Also this criterion, however, primarily applies to quantitative studies, and it is not such a big problem to collect valid information in a qualitative study (Holme, et al., 1997). There may, however, be a problem that even if the information received and collected in the study might be valid, the researchers may interpret the information wrong, and the results will not mirror the reality (Holme, et al., 1997). Another problem may be that the studied subject, if it is a person, might behave in a different way when he/she is studied, according to what he/she thinks the observer want him/her to do. To make sure that the collected data is valid, it is important for the researchers to have a back- 23 and-forth strategy with data collection, meaning that the information is processed several times by both the researchers and the studied subjects (Holme, et al., 1997). To better apply these concepts to qualitative studies, some authors have adapted them into other concepts. Borrego et al. (2009) explain that the term trustworthiness is often used for qualitative studies instead of reliability and validity. For a study to be trustworthy some criteria are to be fulfilled, for example: a clear statement of the theoretical perspective; asking participants to review research findings; using multiple data sources; and providing thick description of the cases (Borrego, et al., 2009). By making a comparative study with four different cases in which several different data collection methods are used, this study will provide a high trustworthiness, and the results will be easier to generalize than if only one company would have been studied. Furthermore, the theoretical framework was structured and literature was reviewed before starting to interview the companies. All study visits involved both researchers, and any questions that arose during data collection or analysis were revisited and checked with corresponding interviewee or company contact. All of these actions have made the study and the results more trustworthy. 24 25 4 EMPIRICAL FINDINGS In this chapter, general descriptions of the studied facilities of the participating case companies will be presented. The companies’ current view of LVPs will also be described, as well as a description of the parts feeding policies that are used for LVPs and which factors that impact the choice. Furthermore, the design of the parts feeding systems related to each parts feeding policy applied for LVPs will be presented. A short summary where key findings are presented can be found after each company description. Due to confidentiality agreements, the companies will be denoted as Company A through Company D. 4.1 Company A Company A is a global company within the automotive industry that manufactures heavy-duty vehicles, such as articulated haulers, wheel loaders, road rollers, and excavation equipment. The facility that has been studied manufactures cabs to the vehicles as well as complete fuel tanks and hydraulic tanks for wheel loaders and articulated haulers. Within this study, however, only the parts feeding system related to the final assembly of cabs will be examined. The cabs are delivered to other production facilities within the same company, located in Sweden and Germany, for final assembly of the complete heavy-duty vehicles. Within the production facility that was studied, approximately 400 people are employed and 35 people work within the materials handling department. 4.1.1 LVPs at Company A Company A offers a high degree of customer adaptation. For instance, the customers are able to choose from 350 different single options for the cab alone when purchasing a wheel loader, making almost every assembled cab unique. Hence, a high degree of flexibility is required in the production process in order to meet the customer requirements. The interviewee indicated that a pareto relationship exists among the various product options that the customers can choose from, which means that there is a large variation in demand for the different parts managed in the facility. The diverse customer demands result in that many parts in the facility have low consumption volume and therefore can be considered as LVPs. Currently, the company does not have any general guidelines for deciding which material flow to apply for parts based on consumption volume or usage frequency. The LVPs are therefore handled with the same parts feeding policies as parts with higher consumption volume. This is seen as troublesome to the company, since LVPs take up a lot of space at the lineside presentation, even though they are not used that regularly, and prohibit efficient use of the more frequently used parts. Therefore, Company A is looking at how to improve parts feeding for LVPs, and one option is to categorize the parts in the near future. The interviewee expressed that the pareto relationship could be used to create classes based on consumption volume for the parts. The classes should then be treated differently regarding parts feeding, resulting in specific choices for parts feeding policies and the design of the parts feeding system for LVPs. 4.1.2 Factors influencing the choice of parts feeding policies at Company A The parts feeding system at Company A primarily consists of two material flows that are used to feed parts to all three assembly lines within the facility. Both flows can be categorized as continuous supply, and the biggest reason for the company to be using this parts feeding policy for all parts is according to the interviewee mainly due to tradition, and that the current design of the parts feeding system has worked fairly well in the past. Furthermore, the use of three 26 assembly lines reduces the part variety of each assembly line, making it feasible to display all parts at the lineside presentation. Fasteners and other components that are prepared in pre- assembly processes have different parts feeding policies. However, they are only a small share of the total amount of parts handled in the facility, and therefore, these flows will not be considered in this study nor described in this section. The two flows in the assembly facility have different designs, and what flow a part is fed through depends on what packaging the part is presented in at the lineside presentations. One flow consists of parts feeding of pallets and the other of boxes and cartons. The majority of the parts are presented at the assembly lines in these standardized packages. More than 5,000 parts are handled within the facility and approximately 70% of them are being used at the final assembly lines, whereas the rest are used at the several pre-assembly workstations in the facility. The decision of what packaging a part is presented in at the assembly lines is made specifically for each part. The unit load that the part is received in from the supplier, available space at the lineside presentation, as well as sufficient cover time of the part are factors which are taken into consideration when deciding which packaging that should be used at the lineside presentation. The interviewee described that Company A is a relatively small actor compared with many of its suppliers, leading to difficulties to receive the parts in the most appropriate unit load as it could impact the purchasing cost of the parts. As a consequence, a large share of parts is repacked internally before being presented at the lineside presentation. When new parts are introduced, and there is not sufficient space available at the lineside presentation, the unit load of another part could be reduced in order to access the needed space. The part that is repacked into a smaller packaging is often an LVP. As a result, the choice of unit load that parts are presented in at the lineside presentation is not performed in a structured way and it is possible that LVPs are presented in inappropriate packaging leading to excessive space usage and inventory costs. 4.1.3 Description of the parts feeding system at Company A Cabs for different heavy-duty vehicles are manufactured on the three assembly lines. The cabs for wheel loaders and road rollers are assembled on a continuously moving assembly line while the cabs for haulers and excavation equipment are manufactured on two separate assembly lines, where the cabs are manually transferred between the workstations. The continuously moving assembly line is elevated approximately 1.5 meters from the floor and therefore the parts feeding becomes more complex since the replenishment of parts to the lineside presentation has to be managed with materials handling equipment that can perform vertical lifting. As a consequence, various variants of forklifts are used for parts feeding to the assembly lines. The high degree of forklift traffic in areas where people are operating is considered to be an issue since the safety of the operators is at risk. However, it is difficult to overcome with the current design of the assembly lines. A measurement used at Company A related to workplace safety is the number of risks and accidents identified in the facility, and according to the interviewee, the high degree of forklift traffic is considered to be a large issue for the improvements related to a safe workplace. The elevated assembly line has 14 workstations, each at about 4 meters in length, and material is mainly presented on one side of the line. The lineside presentation on the other side contains parts needed for the pre-assembly operations that are located in parallel with the assembly line. The takt time is usually 17 minutes, which limits the maximum time allowed for replenishment 27 of pallets to the lineside presentation. The two other assembly lines have similar designs but with fewer workstations and shorter takt times. With regard to the quality aspect related to the large number of parts being stored at the lineside presentation, the company experiences a risk of picking errors at the assembly line, due to similar parts being stored close to each other. This could be seen as a large disadvantage, since the quality aspect of the offered products is considered to be of high importance for the customers. However, an advantage experienced at Company A regarding all parts being stored at the line is that everything is accessible when it is needed. A measurement that is used for materials handling related to this is internal delivery precision, which is evaluated in terms of how often the parts are available at the lineside presentation when there is a requirement for them. According to measurement data provided by Company A, only a limited amount of shortages is experienced at the lineside present