Report No E2014:084 Department of Technology Management and Economics Division of Quality Sciences CHALMERS UNIVERSITY OF TECHNOLOGY Gothenburg, Sweden 2014 Towards the integration of Quality Management and Business Analytics A case study at Volvo GTT PE Master’s thesis in Quality and Operations Management Neda Abdolrashidi Niklas Glaerum I Report NO E2014:084 Towards the integration of Quality Management and Business Analytics A case study at Volvo GTT PE Neda Abdolrashidi Niklas Glaerum Department of Technology Management and Economics Division of Quality Sciences CHALMERS UNIVERSITY OF TECHNOLOGY Gothenburg, Sweden 2014 II Towards the integration of Quality Management and Business Analytics A case study at Volvo GTT PE Neda Abdolrashidi & Niklas Glaerum, 2014 ©Neda Abdolrashidi & Niklas Glaerum, 2014 Technical report no E2014:084 Department of Technology Management and Economics Division of Quality Sciences Chalmers University of Technology SE-412 96 Göteborg Sweden Telephone + 46 (0)31-772 1000 III Towards the integration of Quality Management and Business Analytics A case study at Volvo GTT PE Neda Abdolrashidi & Niklas Glaerum, 2014. Department of Technology Management and Economics Division of Quality Sciences Chalmers University of Technology SUMMARY With the increase of digital data and the rise of concepts like big data, the need for business analytics is assumed to increase. Business analytics relationship to other research areas is yet to be investigated. This thesis will therefore contribute to bridging the research gap by focusing on quality management and its support to business analytics. The relationship is discussed in general terms and a quality management practice is investigated for its ability to support the business analytics process. A literature review is conducted in order to display the relationship between the two research areas. Quality management is presented as a system of principles, practices and techniques. Several business analytics processes are presented and compared and the Knowledge Discovery in Database process is chosen as a representative process. A case study is conducted at Volvo GTT PE and through an abductive research approach a customized version of Quality Function Deployment is developed in order to support the business analytics process. The proposed methodology consists of four stages; Requirements investigation, Outcome planning, Process planning and Taking action based on findings, each involving several steps. The methodology is explained in the context of the case study. The quality management principles, practices and techniques that can support business analytics are investigated and displayed in a framework. The framework shows that the quality management principles should be considered in all phases of the business analytics process. The case study has also shown that the customized version of Quality Function Deployment can support all phases while the quality management techniques can be used in specific phases. Keywords: Business analytics, Quality management, Quality Function Deployment, House of Quality. IV Acknowledgements This study was conducted as a master thesis by two students in the Master’s Programme in Quality and Operations Management at Chalmers University of Technology, Sweden. The study was enabled by the help and support from people around us and we would like to extend our gratitude to all of them. First of all we would like to thank our supervisor at Chalmers, Hendry Raharjo, for his invaluable support and directions. We would also like to thank our examiner, Ida Gremyr, and our opponents, Helena Hellerqvist and Mafalda Svensson de Brito, for excellent feedback on the report helping us to increase the quality and clarity of our message. This study required support from a company and we received a lot from Volvo GTT PE in Gothenburg. We would like to thank our supervisors Hans Berggren and Per Johansson for educating us about the world outside academia as well as for their availability and helpful advices. The study also included 30 interviewees and many other informants without whose help the study would have been impossible. Thank you for your time and friendly advices. Finally we would like to thank each other for a great collaboration and lessons outside the scope of the thesis. Gothenburg, July 2014 ________________________ ________________________ Neda Abdolrashidi Niklas Glaerum V Table of content 1. INTRODUCTION........................................................................................................................................ 1 1.1. INTRODUCTION ................................................................................................................................................. 2 1.2. PURPOSE ......................................................................................................................................................... 3 1.3. RESEARCH QUESTIONS........................................................................................................................................ 3 1.4. DELIMITATIONS ................................................................................................................................................ 3 2. RESEARCH METHODOLOGY ...................................................................................................................... 5 2.1. RESEARCH STRATEGY ......................................................................................................................................... 6 2.2. RESEARCH DESIGN ............................................................................................................................................. 6 2.3. RESEARCH METHOD .......................................................................................................................................... 6 2.3.1. Understanding the case and test procedures ....................................................................................... 7 2.3.2. Literature review .................................................................................................................................. 7 2.3.3. Interviews ............................................................................................................................................. 8 2.4. DATA ANALYSIS ................................................................................................................................................. 9 2.4.1. Analysis of interview data .................................................................................................................... 9 2.5. RESEARCH QUALITY ......................................................................................................................................... 11 2.6. ETHICS .......................................................................................................................................................... 12 3. THEORETICAL FRAMEWORK ................................................................................................................... 13 3.1. QUALITY MANAGEMENT ................................................................................................................................... 14 3.1.1. Collecting information about the customer ........................................................................................ 16 3.1.2. Quality Function Deployment ............................................................................................................. 17 3.1.3. The Kano model .................................................................................................................................. 19 3.1.4. Improvement and management tools ................................................................................................ 19 3.1.5. Summary ............................................................................................................................................ 20 3.2. BUSINESS ANALYTICS........................................................................................................................................ 21 3.2.1. Big data .............................................................................................................................................. 26 3.2.2. Data analysis ...................................................................................................................................... 26 3.2.3. Presentation ....................................................................................................................................... 27 3.3. SYNTHESIS OF THEORETICAL FRAMEWORK ............................................................................................................ 27 4. RESULTS AND ANALYSIS ......................................................................................................................... 31 4.1. THE CASE – VOLVO .......................................................................................................................................... 32 4.1.1. The COP and Hot test.......................................................................................................................... 33 4.2. QFD AS A SUPPORTIVE PRACTICE FOR BUSINESS ANALYTICS ..................................................................................... 33 4.3. REQUIREMENTS INVESTIGATION ......................................................................................................................... 34 4.3.1. Determine who the customers are ..................................................................................................... 34 4.3.2. Understanding the current situation .................................................................................................. 36 4.3.3. Determining customer needs .............................................................................................................. 39 4.3.4. Prioritize customer needs ................................................................................................................... 39 4.3.5. Analyzing correlations ........................................................................................................................ 41 4.4. OUTCOME PLANNING ....................................................................................................................................... 42 4.4.1. Identify quality attributes ................................................................................................................... 42 4.4.2. Relationship matrix............................................................................................................................. 42 4.4.3. Planning and deploying customer needs ............................................................................................ 44 4.4.4. Analyzing correlations ........................................................................................................................ 44 VI 4.5. PROCESS PLANNING ......................................................................................................................................... 45 4.5.1. Identify actions ................................................................................................................................... 45 4.5.2. Drawing relationship matrix ............................................................................................................... 45 4.5.3. Planning and deploying quality attributes ......................................................................................... 47 4.5.4. Analyze correlations ........................................................................................................................... 47 4.6. TAKING ACTION BASED ON FINDINGS ................................................................................................................... 48 4.6.1. Sort actions in order of importance .................................................................................................... 48 4.6.2. Divide actions based on BA phase ...................................................................................................... 48 4.7. GENERAL QFD METHODOLOGY FOR SUPPORT OF BA PROCESSES .............................................................................. 50 4.8. SUPPLEMENTS TO QM´S SUPPORT OF BA ............................................................................................................ 53 4.8.1. Selection ............................................................................................................................................. 53 4.8.2. Preprocessing ..................................................................................................................................... 54 4.8.3. Transformation ................................................................................................................................... 54 4.8.4. Data mining ........................................................................................................................................ 55 4.8.5. Interpretation/Evaluation ................................................................................................................... 55 4.8.6. Update of the framework ................................................................................................................... 55 5. DISCUSSIONS AND CONCLUSION ........................................................................................................... 57 5.1. DISCUSSIONS.................................................................................................................................................. 58 5.2. CONCLUSION .................................................................................................................................................. 60 5.3. FUTURE RESEARCH .......................................................................................................................................... 60 REFERENCES ................................................................................................................................................... 61 APPENDICES ................................................................................................................................................... 67 APPENDIX A – INTERVIEW GUIDE MANAGERS .............................................................................................................. 68 APPENDIX B – INTERVIEW GUIDE SPECIALISTS .............................................................................................................. 70 VII List of figures FIGURE 1 SYSTEMATIC COMBINING FRAMEWORK (DUBOIS & GADDE, 2002) .......................................................................... 6 FIGURE 2 RESEARCH PROCESS ......................................................................................................................................... 7 FIGURE 3 DEFINITIONS OF QUALITY (BERGMAN & KLEFSJÖ, 2011) ...................................................................................... 14 FIGURE 4 QM FRAMEWORK (DEAN & BOWEN, 1994) ..................................................................................................... 15 FIGURE 5 THE CORNER STONE MODEL (BERGMAN & KLEFSJÖ, 2011) .................................................................................. 15 FIGURE 6 PRINCIPLES, TECHNIQUES AND TOOLS ACCORDING TO HELLSTEN AND KLEFSJÖ (2000) ............................................... 16 FIGURE 7 THE HOUSE OF QUALITY (GOVERS, 2001) ......................................................................................................... 17 FIGURE 8 THE FOUR PHASES IN QFD ACCORDING TO HAUSER AND CLAUSING (1988) ............................................................. 18 FIGURE 9 THE KANO MODEL (MATZLER AND HINTERHUBER, 1996) .................................................................................... 19 FIGURE 10 THE SEVEN IMPROVEMENT TOOLS (BERGMAN & KLEFSJÖ, 2011) ........................................................................ 20 FIGURE 11 THE SEVEN MANAGEMENT TOOLS (BERGMAN & KLEFSJÖ, 2011) ......................................................................... 20 FIGURE 12 A SUMMARY OF THE PRINCIPLES, PRACTICES AND TECHNIQUES OF QM .................................................................. 20 FIGURE 13 THE BA PROCESS ACCORDING TO SAXENA AND SRINIVASAN (2013) ..................................................................... 21 FIGURE 14 THE KDD PROCESS (FAYYAD, 1996) .............................................................................................................. 22 FIGURE 15 THE CRISP-DM PROCESS (SHEARER, 2000) ................................................................................................... 23 FIGURE 16 THE BA PROCESS ACCORDING TO RUNKLER (2012) ........................................................................................... 24 FIGURE 17 THE ORGANIZATIONAL BA FRAMEWORK ( GROSSMAN & SIEGEL, 2014) ............................................................... 24 FIGURE 18 THE BA PROCESS ACCORDING TO LAURSEN AND THORLUND (2010)..................................................................... 25 FIGURE 19 COMPARISON BETWEEN BA PROCESSES .......................................................................................................... 28 FIGURE 20 THE SUGGESTED BA PROCESS (FAYYAD, 1996) ................................................................................................ 28 FIGURE 21 INITIAL FRAMEWORK INTEGRATING QM AND BA .............................................................................................. 29 FIGURE 22 ORGANIZATIONAL STRUCTURE AT VOLVO GTT PE GOTHENBURG ......................................................................... 32 FIGURE 23 STAKEHOLDER RANKING ............................................................................................................................... 35 FIGURE 24 BARCHART OVER CUSTOMER RANKING ............................................................................................................ 35 FIGURE 25 CURRENT USAGE OF THE TEST RESULTS ............................................................................................................ 37 FIGURE 26 CURRENT USAGE SPLIT BY SECTION ................................................................................................................. 37 FIGURE 27 PERCEIVED IMPACT ON ACTIVITIES .................................................................................................................. 38 FIGURE 28 REASONS FOR NOT USING THE TEST RESULTS..................................................................................................... 38 FIGURE 29 EMISSION AND PERFORMANCE PARAMETERS OF INTEREST .................................................................................. 38 FIGURE 30 HOUSE OF QUALITY 1 .................................................................................................................................. 40 FIGURE 31 PRIORITIZATION OF CUSTOMER NEEDS ............................................................................................................. 41 FIGURE 32 HOUSE OF QUALITY 2 .................................................................................................................................. 43 FIGURE 33 HOUSE OF QUALITY 3 .................................................................................................................................. 46 FIGURE 34 PRIORITIZED ACTION PLANS ........................................................................................................................... 47 FIGURE 35 ACTIONS SPLIT BY BA PROCESS PHASE ............................................................................................................. 49 FIGURE 36 GENERAL QFD METHODOLOGY .................................................................................................................... 51 FIGURE 37 FINAL FRAMEWORK FOR INTEGRATING QM AND BA .......................................................................................... 56 file:///E:/Backup/New%20backup/Report/Final%20report/Towards%20the%20integration%20of%20QM%20and%20BA%20-%20Neda%20Abdolrashidi%20and%20Niklas%20Glaerum.docx%23_Toc397504510 file:///E:/Backup/New%20backup/Report/Final%20report/Towards%20the%20integration%20of%20QM%20and%20BA%20-%20Neda%20Abdolrashidi%20and%20Niklas%20Glaerum.docx%23_Toc397504511 file:///E:/Backup/New%20backup/Report/Final%20report/Towards%20the%20integration%20of%20QM%20and%20BA%20-%20Neda%20Abdolrashidi%20and%20Niklas%20Glaerum.docx%23_Toc397504512 file:///E:/Backup/New%20backup/Report/Final%20report/Towards%20the%20integration%20of%20QM%20and%20BA%20-%20Neda%20Abdolrashidi%20and%20Niklas%20Glaerum.docx%23_Toc397504513 file:///E:/Backup/New%20backup/Report/Final%20report/Towards%20the%20integration%20of%20QM%20and%20BA%20-%20Neda%20Abdolrashidi%20and%20Niklas%20Glaerum.docx%23_Toc397504514 file:///E:/Backup/New%20backup/Report/Final%20report/Towards%20the%20integration%20of%20QM%20and%20BA%20-%20Neda%20Abdolrashidi%20and%20Niklas%20Glaerum.docx%23_Toc397504515 file:///E:/Backup/New%20backup/Report/Final%20report/Towards%20the%20integration%20of%20QM%20and%20BA%20-%20Neda%20Abdolrashidi%20and%20Niklas%20Glaerum.docx%23_Toc397504516 file:///E:/Backup/New%20backup/Report/Final%20report/Towards%20the%20integration%20of%20QM%20and%20BA%20-%20Neda%20Abdolrashidi%20and%20Niklas%20Glaerum.docx%23_Toc397504518 file:///E:/Backup/New%20backup/Report/Final%20report/Towards%20the%20integration%20of%20QM%20and%20BA%20-%20Neda%20Abdolrashidi%20and%20Niklas%20Glaerum.docx%23_Toc397504519 file:///E:/Backup/New%20backup/Report/Final%20report/Towards%20the%20integration%20of%20QM%20and%20BA%20-%20Neda%20Abdolrashidi%20and%20Niklas%20Glaerum.docx%23_Toc397504520 file:///E:/Backup/New%20backup/Report/Final%20report/Towards%20the%20integration%20of%20QM%20and%20BA%20-%20Neda%20Abdolrashidi%20and%20Niklas%20Glaerum.docx%23_Toc397504521 file:///E:/Backup/New%20backup/Report/Final%20report/Towards%20the%20integration%20of%20QM%20and%20BA%20-%20Neda%20Abdolrashidi%20and%20Niklas%20Glaerum.docx%23_Toc397504522 file:///E:/Backup/New%20backup/Report/Final%20report/Towards%20the%20integration%20of%20QM%20and%20BA%20-%20Neda%20Abdolrashidi%20and%20Niklas%20Glaerum.docx%23_Toc397504524 file:///E:/Backup/New%20backup/Report/Final%20report/Towards%20the%20integration%20of%20QM%20and%20BA%20-%20Neda%20Abdolrashidi%20and%20Niklas%20Glaerum.docx%23_Toc397504525 file:///E:/Backup/New%20backup/Report/Final%20report/Towards%20the%20integration%20of%20QM%20and%20BA%20-%20Neda%20Abdolrashidi%20and%20Niklas%20Glaerum.docx%23_Toc397504526 file:///E:/Backup/New%20backup/Report/Final%20report/Towards%20the%20integration%20of%20QM%20and%20BA%20-%20Neda%20Abdolrashidi%20and%20Niklas%20Glaerum.docx%23_Toc397504527 file:///E:/Backup/New%20backup/Report/Final%20report/Towards%20the%20integration%20of%20QM%20and%20BA%20-%20Neda%20Abdolrashidi%20and%20Niklas%20Glaerum.docx%23_Toc397504531 file:///E:/Backup/New%20backup/Report/Final%20report/Towards%20the%20integration%20of%20QM%20and%20BA%20-%20Neda%20Abdolrashidi%20and%20Niklas%20Glaerum.docx%23_Toc397504532 file:///E:/Backup/New%20backup/Report/Final%20report/Towards%20the%20integration%20of%20QM%20and%20BA%20-%20Neda%20Abdolrashidi%20and%20Niklas%20Glaerum.docx%23_Toc397504533 file:///E:/Backup/New%20backup/Report/Final%20report/Towards%20the%20integration%20of%20QM%20and%20BA%20-%20Neda%20Abdolrashidi%20and%20Niklas%20Glaerum.docx%23_Toc397504534 file:///E:/Backup/New%20backup/Report/Final%20report/Towards%20the%20integration%20of%20QM%20and%20BA%20-%20Neda%20Abdolrashidi%20and%20Niklas%20Glaerum.docx%23_Toc397504535 file:///E:/Backup/New%20backup/Report/Final%20report/Towards%20the%20integration%20of%20QM%20and%20BA%20-%20Neda%20Abdolrashidi%20and%20Niklas%20Glaerum.docx%23_Toc397504536 file:///E:/Backup/New%20backup/Report/Final%20report/Towards%20the%20integration%20of%20QM%20and%20BA%20-%20Neda%20Abdolrashidi%20and%20Niklas%20Glaerum.docx%23_Toc397504537 file:///E:/Backup/New%20backup/Report/Final%20report/Towards%20the%20integration%20of%20QM%20and%20BA%20-%20Neda%20Abdolrashidi%20and%20Niklas%20Glaerum.docx%23_Toc397504538 file:///E:/Backup/New%20backup/Report/Final%20report/Towards%20the%20integration%20of%20QM%20and%20BA%20-%20Neda%20Abdolrashidi%20and%20Niklas%20Glaerum.docx%23_Toc397504542 file:///E:/Backup/New%20backup/Report/Final%20report/Towards%20the%20integration%20of%20QM%20and%20BA%20-%20Neda%20Abdolrashidi%20and%20Niklas%20Glaerum.docx%23_Toc397504544 file:///E:/Backup/New%20backup/Report/Final%20report/Towards%20the%20integration%20of%20QM%20and%20BA%20-%20Neda%20Abdolrashidi%20and%20Niklas%20Glaerum.docx%23_Toc397504546 VIII List of tables TABLE 1 MATCHING RESEARCH QUESTIONS WITH RESEARCH METHOD..................................................................................... 7 TABLE 2 RESPONDENTS SPLIT BY SECTION .......................................................................................................................... 8 TABLE 3 RISKS WITH CHOSEN RESEARCH METHOD AND WAYS OF MITIGATING THE RISKS ............................................................. 9 TABLE 4 MATCHING QUALITATIVE AND QUANTITATIVE EVALUATION CRITERIA (BRYMAN & BELL, 2011) ..................................... 11 TABLE 5 THE STAGES IN QFD ACCORDING TO FRANCESCHINI (2001) ................................................................................... 18 TABLE 6 SUGGESTED STAGES AND STEPS FOR QFD WHEN SUPPORTING BA............................................................................ 33 1 1. Introduction This chapter introduces the research area and outlines the purpose as well as the research questions associated with this study. 2 1.1. Introduction According to Bergman and Klefsjö (2011) quality has always been important to customers. Quality management (QM) is therefore a mature and relatively old research field (Sousa & Voss, 2002). Although diversity in definitions of quality and QM still exists most studies show a positive correlation between QM efforts and operational and business performance (Sousa & Voss, 2002). This is best exemplified by the “Japanese miracle” between 1950 and 1985 when the Japanese industry went from having a very poor quality reputation to being world leading (Bergman & Klefsjö, 2011). This thesis use the definition of QM as an approach to management involving a system of principles, practices and techniques presented by Dean and Bowen (1994). The recent years have seen a large increase in the amount of digital data produced (Loshin, 2013) which has led to the rise of concepts like big data and business analytics (BA) (Mayer-Schönberger & Cukier, 2013). BA is defined as ensuring that the right users get the right information at the right time (Laursen & Thorlund, 2010). All collected data needs to be translated into information and knowledge for full understanding (Laursen & Thorlund, 2010). This translation is the output of BA (Davenport et al., 2001). BA has traditionally been performed manually but the increasing amount of data makes manual analysis slow, expensive and impractical (Fayyad et al., 1996). Meanwhile data left without analysis is a waste (Davenport et al., 2001) which is why an increase in the amount of data will lead to an increased need for BA. The adoption of BA comes with benefits in terms of better decision making (Davenport, 2009) as well as improved business performance (Bronzo et al., 2013; Kiron et al., 2011). As BA grows in importance other research areas need to reflect over the implications on their activities. This applies to QM as well as other research areas. If companies want to keep the competitive advantages they get from QM (Bergman & Klefsjö, 2011) while capitalizing on BA the support and conflicts with between the improvement concepts need to be fully understood. This thesis is a case study at Volvo GTT PE in Gothenburg. Just like many other companies (Davenport et al., 2001) this company struggles with analyzing the amount of data currently produced in their business processes. More specifically two processes, the Conformity of Production (CoP) and the Hot test, will be investigated. With quality being one of the company’s core values, the aim is that a part of the solution to this problem lies in the QM field. 3 1.2. Purpose The purpose of this research is to discover how quality management can support business analytics process in the organization. 1.3. Research questions The purpose will be fulfilled by answering the following research questions: RQ1: How can quality management principles support business analytics process? RQ2: How can quality management practices and techniques support business analytics process? 1.4. Delimitations At present, the CoP- and Hot tests are performed in several locations worldwide. This project is limited to the tests performed in Sweden. The aim is however to provide the results in a way that they are applicable to other sites and processes. The project will provide a framework on how quality management can support business analytics. The project is delimited from implementing the suggested guidelines into the organization. The BA process is long and stretches from decision framing to executing the decisions taken based on analytics. This study is delimited from the support of decision making and decision execution as decision making is a research area on its own. 4 5 2. Research Methodology In this chapter the methodology used in this study is described. The chapter also addresses research quality and ethical considerations. 6 2.1. Research strategy This study has utilized a qualitative research strategy. According to Hacohen (2004) research methodology is highly dependent on the distinction between induction and deduction. The qualitative research strategy involves induction where theory is developed from the research findings (Bryman & Bell, 2011). Deduction on the other hand is the testing of a hypothesis (Bryman & Bell, 2011). Both inductive and deductive research has elements of the other research stance (Bryman & Bell, 2011). Induction which has a deductive explanatory nature is called abduction (Kuipers, 2004). This study has used an abductive approach called systematic combining where theory, framework, empirical world and the case all influence the research process (see Figure 1) (Dubois & Gadde, 2002). The theoretical frame and empirical study thus evolved simultaneously. 2.2. Research design The systematic combining approach involves matching of several data sources (Dubois & Gadde, 2002). The study employs a case study design where a case and literature is examined. The case study design provides the opportunity to study an organization and its activities related to the research area in detail (Bryman & Bell, 2011). Yin (2009) suggest that a case study is suitable when the research questions are of an exploratory nature and explain that these research questions often begin with the words how or why. The two research questions in this study fits well into this description. According to Dubois and Gadde (2002) the case evolves during the research process as more theory and data are gathered. Corbin and Strauss (2008) emphasize that some theoretical knowledge can facilitate a researchers understanding of a case while too much theoretical knowledge inhibits it. This relates well to Gummesson’s (2000) ideas of Preunderstanding as a stepping stone to understanding. The authors’ preunderstanding is discussed further in the research quality chapter (Chapter 2.5). 2.3. Research Method According to Bryman and Bell (2011) a research method denotes the means of data collection. In this study several means of data collection were used such as literature review, interviews, observations and the study of internal company documents and test results from the two test procedures investigated (see results chapter for more information about the tests). Figure 1 Systematic combining framework (Dubois & Gadde, 2002) 7 Table 1 shows the connection between the research questions and the method used to answer them, the research process can be explained as in Figure 2 2.3.1. Understanding the case and test procedures In order to investigate the CoP and Hot test there was a need to understand the data that was collected and the processes that creates the data. Therefore two weeks were spent on observation of related processes in internal management systems and reviewing available documents. In addition, informal conversations as well as four unstructured interviews with one process owner, two test engineers responsible for collecting the test results and one analyst were conducted. The test rigs were also visited and similar tests observed in order to enhance the understanding of the test procedure. 2.3.2. Literature review In order to answer the first research questions the data collection method was mainly based on the literature review. In order to gain knowledge about quality management principles, tools and techniques and also the concepts and frameworks regarding business analytics, as two main areas of research, a literature review was performed using mainly Science direct (sciencedirect.com) and Web of Science (apps.webofknowledge.com). The following keywords were investigated: Quality management, Quality Function Deployment, Business analytics, Decision making, data analysis, data presentation, visualization. The articles and books found using these keywords were evaluated based on their relevance to the research. In total around 70 articles and books were found to be useful and read more thoroughly. The knowledge gained from the literature was then used as an input to data analysis part of the research and to run the case study. However, the data collection process was Research question Research method RQ1: How can quality management principles support a business analytics process Literature review, interviews RQ2: How can quality management practices and techniques support a business analytics process Literature review, interviews, observations and internal documents Develop framework Case study Planning Data collection Data processing Data analysis Literature review Understanding the process and organisation Observation Internal documents Figure 2 Research process Table 1 Matching research questions with research method http://www.sciencedirect.com/ 8 iterative and went back and forth between the case study and the literature review. The knowledge found in the literature was used as a guide to run the different phases of the case study and the findings from the case study was used as a guide to which areas were needed to be further investigated. 2.3.3. Interviews The data collection through interviews was initiated with the identification of stakeholders to the testing procedures. All stakeholders were internal customers working in the same part of the organization (Volvo GTT Powertrain Engineering). Throughout the report the names stakeholders and customers will be used interchangeably for this group. This was done through a snowball sampling where the customers at management level were identified by two persons currently performing analysis on the test results. The management group was chosen as customers based on that they are affected or interested in the test results. These managers were then during the interviews asked to identify specialists in their section that uses, or would benefit from using, the test results in their activities. This resulted in a total of 30 interviewees spread over eight sections. The distribution can be seen in Table 2. The interviews were based on an interview guide. Semi-structured interviews were chosen as research method since the method fits the inductive orientation better than structured interviews (Bryman & Bell, 2011) allowing more flexibility to the interviewer and interviewee. An interview guide was then developed where the authors first brainstormed areas of interest. After these were identified, interview questions that correspond to the research areas were then derived and improved. According to Bryman and Bell (2011) the language should be relevant to the interviewees and this was considered when improving the questions. An introduction that set the scene was also developed to ensure that all the interviewees had information about the purpose of the study and interview as well as relevant knowledge about the tests. Also instructions to minimize misunderstandings and faulty information were included in the introduction. The questions were then arranged in order of invasiveness starting with questions about actions, then knowledge and finally philosophy in accordance with Price’s (2002) theory of laddered questions. This facilitated the creation of rapport between the interviewee and researchers which according to Dundon and Ryan (2010) is a key factor to collecting rich data. The interview guide was then tested, both on imaginary customers and one of the already identified customers in a pilot study. A list of risks with the chosen research method was also brainstormed by the authors along with potential solutions (see Table 3). These solutions along with the feedback from the pilot study were then used to improve the interview guide. Slightly different interview guides were developed for the managers versus the specialists due to the fact that some questions only were relevant to one Table 2 Respondents split by section 9 of the groups. The final interview guides can be seen in appendices A and B. Table 3 Risks with chosen research method and ways of mitigating the risks The interviews were then conducted. With the permission of the interviewees all interviews were recorded. Both authors attended all the interviews and the interview guide was divided so that the same questions were asked by the same researcher, in the same way, in all interviews. Follow-up questions were asked when anything was unclear. The researcher not asking questions focused on taking notes that were used to support the summary and analysis of the interviews. All interviews took place in private rooms except for one interview conducted via Lync (an online solution). The lengths of the interviews were between 15 minutes to 3,5 hours depending on how much the interviewee had to say in relation to the research area. If the interview took longer time than expected a new time and place was arranged for the following interview. All interviews were summarized and sent to the interviewees for validation. The interviewees were given one week to change any answers that they felt did not reflect reality due to misunderstandings or a change of mind. Lincoln and Guba (1985) refer to this technique as member checking and presents it as a technique for increasing credibility in qualitative research. Cho and Trent (2006) warns that this technique requires that the respondents have integrity, an idea that is shared by Lincoln and Guba (1985). Buchbinder (2011) also notes that the power balance between the respondents and researchers shift when the researchers are reliant on the respondents to accept their work which in turn could affect the analysis. When summarizing an interview one of the researcher would listen through the recorded interview and use the interview guide to fill in the answers to all questions. The summary was then scrutinized by the other researcher who compared it to the notes taken during the interview as well as his or her memory of the interview. If there were any disagreements these were discussed between the authors and an agreement was reached. The summarized interview was then sent to the interviewee for validation along with any follow-up questions. 2.4. Data analysis 2.4.1. Analysis of interview data After receiving a validation from the interviewee or the passing of deadline for validation the interview data was copied into an excel sheet where each row corresponded to an interviewee and each column corresponded to a question. The sheet also included information about which section the interviewee belonged to as well as whether it was a manager or a specialist. The interview guide contained many questions that were not aimed at only finding the needs (Appendix A and Appendix B). These questions were instead used to understand the current situation. The authors then codified 10 the answers individually. The codes were written down on post-its and compared with the other researcher’s codes. In case any codes were identical one of them were discarded. The different codes were then explained and grouped with other similar codes with the aim of having 6-10 groups. No codes were forced into a group if it was not perceived to belong there. Each answer was then categorized as belonging to one of the decided codes. From the codes a number of requirements on the specific BA process (CoP and Hot test) were then identified. These were then evaluated on whether they were real needs or quality attributes to an underlying requirement. If they were considered to be a quality attribute the underlying requirement was identified by the authors and added to the list of needs. A large table inspired by the House of Quality in QFD (Bergman & Klefsjö, 2011) with every interviewee in a separate row and every requirement in a separate column (HoQ1, Figure 31) was constructed. Based on the interviews each interviewee was then matched with the needs he or she had expressed. In the case that any requirement was implied by another requirement these were also added. If the requirement was requested by an interviewee the corresponding cell was marked with a “1”. If it was not requested the cell was marked with a “0”. The number of interviewees mentioning a specific requirement was then summed up to give an indication of the importance of each requirement. Since the summarized values only show the frequency of mentioning they were not considered to give a good estimation of needs relative importance. The stakeholders were then evaluated based on their level of current usage, their interest in using the test results and the impact their activities had on the final outcome in order to give different weights to responses from different customers. This was incorporated into the HoQ and a new importance rating on needs were derived. The roof of the matrix was filled out to show correlations between needs. Each requirement was now considered in order to brainstorm quality attributes that reflected the needs. This was done individually by the authors and the quality attributes were then compared and a comprehensive list developed. A new table (HoQ2, Figure 32) was created with the needs from the first table corresponding to a row in the new table and the developed quality attributes corresponding to a column. The quality attributes were then matched with needs in the same way that the needs were matched with the stakeholders. The rating scale used in the relationship matrix was 0,1,3,9 as the relationship now could be of different strength. The importance of each requirement gave different weight to the quality attribute corresponding to that requirement. The multiplied numbers were summarized for each quality attribute. This was used as an importance rating of the different quality attributes. The roof of the second HoQ was also filled out to establish any correlations between quality attributes. The quality attributes and their summed up importance rating were then included in a third HoQ (HoQ3) as rows with the columns occupied by actions that corresponded to the quality attributes. The actions were brainstormed by answering the question “what needs to be done for this quality attribute to be present?”. The list of actions was validated by company representatives familiar with the test processes. The relationship matrix was filled in using the same rating scale as the previous HoQ (0,1,3,9) and the sum of each rating multiplied with the importance of the quality attribute it corresponded to was calculated. The three HoQ can be found in the results and analysis chapter. 11 2.5. Research quality According to Bryman and Bell (2011) the use of the same criteria when evaluating qualitative research as when evaluating quantitative research is insufficient. Lincoln and Guba (1985) instead present the concept trustworthiness. Trustworthiness consists of the four criteria credibility, transferability, dependability and confirmability (Lincoln & Guba, 1985). Bryman and Bell (2011) relate these criteria to the quantitative criteria in the following way (Table 4). Table 4 Matching qualitative and quantitative evaluation criteria (Bryman & Bell, 2011) Credibility relates to the extent that multiple researcher accounts of a social reality is similar (Bryman & Bell, 2011). There are several techniques for ensuring credibility in a research study (Lincoln & Guba, 1985). One of these is member checks which entails the validation of research findings with respondents (Lincoln & Guba, 1985). This technique was utilized in this research study as the summaries of interviews were sent to each respondent for validation. As explained earlier this technique and its benefits are debated. Another technique that was used to some extent in this study is triangulation. By interviewing several stakeholders with similar work assignments as well as reading internal documents some answers from respondents could be questioned and through the use of follow-up questions accepted or rejected. According to Lincoln and Guba (1985) this technique establishes credibility and thereby trustworthiness. Transferability relates to the ability to generalize the research findings to another time or to a larger population than the sample (Lincoln & Guba, 1985). Both Lincoln and Guba (1985) and Bryman and Bell (2011) agree that transferability is best established by a detailed description of the study subject. This way other researchers can read and decide whether the findings are applicable to their sample or not. In this case the authors have attempted to describe the case as detailed as possible for enhanced transferability. To what extent it was successful is for other researchers to evaluate. Dependability instead relates to the ability to audit the study as such (Bryman & Bell, 2011). This is according to Lincoln and Guba (1985) established through a detailed description of the research process. In this study it is attempted to explain the methodology in an exhaustive manner in order to satisfy this evaluation criteria. Confirmability is according to Bryman and Bell (2011) the degree of objectivity shown by the researchers. Lincoln and Guba (1985) mean that this should be audited by others and is hard for the researchers to evaluate themselves. All of the evaluations were made separately by the authors and later compared which is believed to reduce the risk of subjectivity in the research. As previously mentioned Gummesson (2001) emphasize the importance of preunderstanding in research programs. It is therefore relevant to explain the authors relation to the case company and research area. Both authors are studying QM at master level and are therefore familiar with the Qualitative criteria Quantitative criteria Credibility = Internal validity Transferability = External validity Dependability = Reliability Confirmability = Objectivity 12 research area while the BA research area was new to both authors although statistical analysis as a part of BA is also frequently used in QM. In terms of the company one of the authors has been working at the department where this study was conducted and therefore had knowledge about the organization and the people in the group where the study was conducted, while the other researcher was new to the organization without any previous knowledge of the specific industry. 2.6. Ethics Bryman and Bell (2011) presents four ethical principles to consider when conducting a research study. These areas are; harm to participants, lack of informed consent, invasion of privacy and deception. This study has attempted to consider these principles. No harm came to the respondents as no invasive questions were asked and all interviews were conducted on a voluntary basis. The interviews were recorded but the respondents were always asked for permission first which combined with the ability for respondents to read and validate all that had been written after the interviews addressed the issue of lack of informed consent. No questions were of a private nature and the respondents were informed that no anonymity was promised. Therefore it is believed by the authors that no invasion of privacy was committed. Before each interview the respondent was informed about the purpose of the research and interview along with other relevant information about the authors and the study (see Appendix A and Appendix B). This was an attempt to avoid deception. 13 3. Theoretical framework In this chapter the theoretical framework is presented. The two main research areas quality management and business analytics are presented individually before expressing the theory synthesis. 14 3.1. Quality management There are many definitions of quality available as can be seen in Figure 3. Garvin (1988) categorizes the definitions into five approaches to quality; the transcendent, user- based, manufacturing-based, value-based and product-based approaches. According to this approach, the transcendent refers to the quality as an entity beyond something that can be define, and according to the transcendent approach quality is a condition of reaching the excellence and achieving the highest standard. In addition, according to Garvin´s (1998) approach, the focus of user- based is on the consumer needs. He defines quality as something that fits to consumer preferences and satisfies their desires. Moreover, regarding the product-based approach, he emphasizes reaching the desired attributes and ingredients of the product as the definition of quality. According to the manufactured-based approach quality is conformity to the established specifications and any deviation from specifications lead to quality reduction, and regarding the value-based approach quality can be defined in terms of cost, prices or any other attribute (Garvin, 1988). This diversity in definitions enhances the importance of choosing a representative definition. Bergman and Klefsjö (2011) define quality as a product´s ability to satisfy, or preferably exceed, the needs and expectations of the customers. They further define customers as “Those we want to create value for” (Bergman & Klefsjö, 2011:28).The definition of customers is important since the customers, according to the above definition of quality, determines if we produce a product of good quality or not. In this research the definition of customer by Bergman and Klefsjö (2011) is used. Dean and Bowen (1994) view TQM as a system of principles, practices and techniques. This view is supported by Hellsten and Klefsjös (2000) view of TQM as a management system consisting of values, techniques and tools. The techniques are explicit ways of performing the practices which are activities to support the principles (Dean & Bowen, 1994). This explanation of practices and techniques show that they relate well to Hellsten and Klefsjö’s (2000) techniques and tools. The structure of these frameworks can therefore be viewed as in Figure 4. The QM system used in this research is based on the view of Dean and Bowen (1994) since the idea of principles, practices and techniques was first discovered by them and later on supported by Hellsten and Klefsjö (2000). Figure 3 Definitions of quality (Bergman & Klefsjö, 2011) 15 According to Hellsten and Klefsjö (2000) there are different viewpoints about which the principles of QM are but some are however generally agreed upon. These are presented as the corner stones of Total Quality Management (TQM) by Bergman and Klefsjö (2011) (Figure 5). TQM is defined by the same authors as “a constant endeavor to fulfill, and preferably exceed, customer needs and expectations at the lowest cost, by continuous improvement work, to which all involved are committed, focusing on the processes in the organization” (Bergman and Klefsjö, 2011:37). The corner stone model is a representation of the values behind TQM and involves focus on customers and processes, continuous improvements, decisions based on facts and committed leadership as well as letting everybody be committed (Bergman & Klefsjö, 2011). As previously stated, each principle in QM need to be performed through a set of practices. According to Dean and Bowen (1994), there are several practices that can be used to support different principles such as making direct contact with the customer and identifying the customer needs through collecting information are the proposed practices to support customer focus. In addition, there are a wide range of techniques that can be used for supporting different practices e.g. flowcharts, control charts, process maps, etc. Examples of tools and techniques are also presented by Hellsten and Klefsjö (2000) (Figure 6). Focus on customers Focus on processes Let everybody be committed Continuous improvements Base decisions on facts Top management committment Figure 4 QM framework (Dean & Bowen, 1994) Figure 5 The corner stone model (Bergman & Klefsjö, 2011) 16 In this research, a set of practices and techniques are used in order to support the QM principles in the cornerstone model. These practices and techniques are explained in the following sections. 3.1.1. Collecting information about the customer At the center of the corner stone model is the focus on customers, which relates well to the definition of quality as being determined by the customer. Bergman and Klefsjö (2011) mean that companies should determine the needs and wants of the customers and attempt to fulfill them in a systematic way. The process of investigating customer needs naturally start with identifying the customers. This task is not limited to the external customers but also include customers within the company (Bergman & Klefsjö, 2011). The notion that customers can be divided into internal and external is shared by Kondo (2001). Lengnick-Hall (1996) elaborates on this theory by presenting five roles that a customer can have and even say that a customer orientation requires an understanding of these roles. The roles are the customer as a resource, co-producer, user, buyer and product. The role a customer has influences the way that customer can contribute to increased quality (Lengnick-Hall, 1996). Maylor (2010) also present three groups from which the stakeholders come from; internal team, core externals and rest of the world which could be helpful when identifying the stakeholders. As customers are a form of stakeholders (Mitchell, Agle & Wood, 1997) the definition of what a stakeholder is becomes relevant. Freeman (2010, p.46) defines a stakeholder as “any group or individual who can affect or is affected by the achievements of the organizations objectives”. Not all stakeholders are of equal importance (Maylor, 2010). When identifying stakeholders Mitchell, Agle and Wood (1997) mean that the dimension stakeholders are evaluated upon should reflect who is really important. Further they suggest three dimensions to consider; power, legitimacy and urgency (Mitchell, Agle & Wood, 1997). A stakeholders position on these three dimensions also give an indication of how they will be treated by managers (Mitchell, Agle & Wood, 1997). Maylor (2010) instead present power and interest as dimensions on which to evaluate the stakeholders. In order to be customer focused there is a need to understand the customer needs. These needs are often referred to as “the voice of the customer” (Griffin & Hauser, 1993). Griffin and Hauser (1993) promote the use of interviews and focus groups with approximately the same outcomes in terms of collected needs. Around 20-30 interviews lead to the capture of 90-95 percent of the needs (Griffin & Hauser, 1993). Figure 6 Principles, techniques and tools according to Hellsten and Klefsjö (2000) 17 3.1.2. Quality Function Deployment The voice of the customer is used as an input to Quality Function Deployment (QFD), a quality management practice (Hellsten & Klefsjö, 2000) for systematically translating the customer needs into product characteristics and further into requirements on what actions need to be taken (Bergman & Klefsjö, 2011). QFD is supported by the House of Quality (HoQ)(Figure 7), a QM technique. In the HoQ the different areas are called rooms (Lager, 2005). According to Raharjo, Brombacher and Xie (2008) there are generally five different inputs to the HoQ; “the customer requirement, the technical attribute, the relationship matrix, the correlation matrix, and the benchmarking information” (Raharjo, Brombacher & Xie, 2008:253). In one of the rooms, the relationship matrix, the “what’s” are matched with the “how’s”. The what’s represent customer needs while the how’s represent quality characteristics (or technical attributes) in the first HoQ (Govers, 2001). Franceschini and Rupil (1999) explain the what’s as goals while the how’s are the means to achieve the goals. The what’s are listed in the rows and given an importance rating. The importance rating could, according to Matzler and Hintlerhuber (1998), be based on the Kano classification of the customer needs. Tan and Shen (2000) presented another framework with the same idea. The how’s are then listed in columns providing the opportunity to fill in the relationship matrix between the what’s and how’s. The relationship can be shown in a number of different ways (Franceschini & Rupil, 1999). According to Akao (1992) the relationship needs to be quantified and provided in a numerical form. An important choice is then whether to have nominal or ordinal scales as rating as well as whether the ordinal scales should be proportional or logarithmic (Franceschini & Rupil, 1999). Examples of the different scales are 1,2,3 (proportional) and 1,3,9 (logarithmic). According to Franceschini and Rossetto (1998) an important and often forgotten issue is that everyone involved in the rating should understand the rating system. If a rating scale will be used for multiplication it will have the implication that a rating of 9 is nine times a high as a rating of 1. In the roof of the HoQ the correlation matrix displays synergies and conflicts between the how’s (Hauser, 1988). The correlation can be positive, negative or non-existing (Magnusson, Kroslid & Bergman, 2000). According to Johnson (2003) the emphasis is on finding conflicts between needs. W h at ’s How’s ”Roof” How much Figure 7 The house of quality (Govers, 2001) 18 The QFD methodology can be explained in two ways (Lager, 2005). One is as a set of four matrices representing four phases in QFD; product planning, product design, process design and production planning (Bergman & Klefsjö, 2011). The other view is a matrix of matrices suggested by Akao (1992) which consists of 16 matrices divided into four areas; quality deployment, technology deployment, cost deployment and reliability deployment (Lager, 2005). Although a simplification, QFD is often represented by the series of houses as illustrated below (Figure 8) Figure 8 The four phases in QFD according to Hauser and Clausing (1988) According to Bergman and Klefsjö (2011), in the first phase the customer attributes are translated into engineering characteristics; in the second phase the engineering characteristics are then translated into parts characteristics; and the third phase includes translating the part characteristics into key process operations which are translated into production requirements in the fourth phase. According to Franceschini (2001) there is a step before the first phase which he calls identifying customer needs. The phases can be divided into the following steps (Franceschini, 2001) (Table 5). Table 5 The stages in QFD according to Franceschini (2001) Customer needs Determine who the customers are Determine customer needs Prioritize customer needs Product planning specifications Identify product design requirements Drawing relationship matrix Planning and deploying expected quality Analyzing correlations between design requirements Part/Subsystem planning specification Identify part characteristics Drawing relationship matrix Planning and deploying product characteristics Analyzing correlations between part characteristics Process planning specification Identify key process operations Drawing relationship matrix Planning and deploying part characteristics Analyzing correlations between key process operations Quality control specification Identify production requirements Drawing relationship matrix Planning and deploying key process operations Analyzing correlations between production requirements 19 Although QFD is fully applicable to service industries there is a need to align the methodology with the intangible products (Akao, 1992; Mazur, 1993). Although Akao (1992) keep the same terminology Mazur (1993) instead divides QFD for services into nine steps with similar content as QFD for products. 3.1.3. The Kano model All customer needs are not the same (Löfgren & Witell, 2005). According to the Kano model customer needs can be divided into basic needs, expected needs and excitement needs (Bergman & Klefsjö, 2011). The relationship between how well these needs are fulfilled (degree of achievement) and customer dissatisfaction/satisfaction is displayed below (Figure 9). According to Bergman and Klefsjö( 2011), the collection of these groups of needs is different. In one hand the basic needs are rarely mentioned in interviews as they are assumed to be present. On the other hand the expected needs are mentioned while the excitement needs are seldom known by the customers themselves (Bergman & Klefsjö, 2011). According to Löfgren and Witell (2005) the nature of a specific customer need is not stable over time. Instead needs travel from being excitement needs, to being expected needs and finally basic needs. Therefore the customer needs have to be constantly updated. 3.1.4. Improvement and management tools Basing decisions on facts is one corner stone of TQM. According to Bergman and Klefsjö (2011) basing decisions on fact is facilitated by the seven improvement tools and the seven management tools. The seven improvement tools are designed to process information while the seven management tools are designed to handle unstructured verbal data (Bergman & Klefsjö, 2011). A summary of the tools are shown below (Figure 10 and 11). In this research different set of tools are used as a support for implementing the practices of QM. For example, during different phases of the study the Affinity Diagram or the Affinity Interrelationship Method (AIM) is used for grouping and clustering reasons since according to Ryan (2011), the AIM is a structured way of organising a brainstorming result that involves grouping and clustering (Ryan, 2011). This technique involves seven steps from generating ideas to discussing the results (George, Figure 9 The Kano model (Matzler and Hinterhuber, 1996) 20 2005). Stratification is another tool that is used in this study since it is a tool that splits up the data based on different criteria (Magnusson, Kroslid & Bergman, 2000). In addition, the control chart is found as a useful tool to meet some of customer needs in this research. Control chart is a visualization of results over time and is based on stochastic variation theory where an upper and lower specification limit is chosen based on the common variation within the process (Du Toit, Steyn & Stumpf, 1986). 3.1.5. Summary A summary of the presented principles, practices and tools can be seen in Figure 12. Principles Practices Techniques Quality Function Deployment Kano model House of Quality Voice of the customer Customer roles Rating scales Stakeholder ranking Affinity Interrelationship Method Data collection Scatter plot Stratification Cause-and- effect diagram Histogram Pareto chart Control chart Seven improvement tools Matrix data analysis Affinity diagram Interrelation diagraph Activity network diagram Process decision program chart Matrix diagram Tree diagram Seven management tools Figure 12 A summary of the principles, practices and techniques of QM Figure 10 The seven improvement tools (Bergman & Klefsjö, 2011) Figure 11 The seven management tools (Bergman & Klefsjö, 2011) 21 3.2. Business analytics business analytics (BA) can be defined as ensuring that the right users get the right information at the right time (Laursen & Thorlund, 2010). This definition is identical to Bogza and Zaharies (2008) definition of Business Intelligence (BI) and according to Saxena and Srinivasan (2013) BI is often used as a synonym for BA although they mean that BI is only a part, and not all, of BA. Loshin (2012) on the other hand means that BI encompasses BA tools which illustrate the similarities of the two concepts. Today the key role of big data and analytics in providing support for the business to achieve the strategic goals is known for many organizations. However, there is still not a best known way of organizing the analytics activities and defining the core processes to support the analytics efforts in the organization (Grossman and Siegel, 2014). According to Saxena and Srinivasan (2013) rational decisions are made in four steps; Idea, Analysis, Decision and Execution. Analytics can support this process to different degrees. They advocate what they call “full lifecycle support” which can be described as an extensive use of analytics to support the process for rational decisions. This support comes from six areas in the analytics domain; decision framing, decision modeling, decision making, decision execution, data stewardship and business intelligence. The first four correspond to a step in the process for rational decisions while the last two supports all of the steps as can be seen in Figure 13 (Saxena & Srinivasan, 2013). The decision framing is the area of defining the decision need. This step starts with mapping the current state of the business and identifying the requirements for decision-making. In addition, understanding both current and future capabilities of the processes is a crucial factor since the organization should be able to execute the decisions. However, the decision frame is not fixed and can be iteratively improved based on the feedback from the decision execution area. As the second step in BA, key variables and relationships are shown through the decision model to give a better understanding of the context. In this area of the framework the important factor is to identify the target variables amongst a mass of available variables and focus on those variables that are related to the decision needs. Therefore, the decision model should be made based on the decision frame. There are several techniques and models to show different types of contexts. For example, the different types of diagrams, the mathematical models and techniques such as control charts, correlation and regression, project management with CPM and PERT, decision trees, etc. The decision modeling step can be broken into other sub steps. Saxena and Srinivasan (2013) define these sub steps as; formulation, data collection, development, testing, evolution and presentation. Decision framing Decision modeling Decision making Decision execution Data stewardship Business intelligence Figure 13 The BA process according to Saxena and Srinivasan (2013) 22 The output from the first two BA steps are then used as the input to the informed and rational decision making as the following step before the last step of business analytics when the decisions need to be executed in a way that lead to an added value for the business (Saxena & Srinivasan, 2013). BI is another part of the BA framework. There is an interaction between this area and other mentioned areas of the framework. In fact, the different databases, systems and tools to support data management, data analysis and decision making are provided by BI. In addition, in order to prevent incorrect and misleading analysis it is necessary to provide usable data for analysis. Therefore, the quality of the data should be measured and its fitness for usage in decision models should be assessed. This requirement can be reached through data stewardship as a part of the BA framework. Another framework related to BA is provided by Fayyad et al. (1996). This framework is called knowledge discovery in databases (KDD) and includes the process of extracting knowledge from data. There are several steps included in this process with the aim of making the data more compact, abstract and useful in order to gain useful knowledge from the data (Fayyad et al., 1996). An overview of the KDD process is provided in Figure 14. Figure 14 The KDD process (Fayyad, 1996) According to Fayyad et al. (1996), the KDD process contains a number of different steps. The process, according to them, starts with identifying customer needs in order to define the goal of the process. Creating a target data set and focusing on the relevant variables, which are selected based on the process goal is the second step. At the preprocessing step, the main sub steps are data cleaning, removing noise from the data and handling the missing data (Fayyad et al., 1996). Further, they mean that in the next step, through the transformation methods, the number of variables is reduced to those that are effective and invariant representations of the data. At the data mining step several processes are performed such as selecting a particular data mining method based on the goals of KDD, exploratory analysis and selection of data mining algorithm to be used in searching for patterns in data (Fayyad et al., 1996). The next step is, according to them, to visualize and interpret the patterns and other information derived from previous steps. The final step is to take the discovered 23 knowledge into action through using it directly or reporting it to the people who are interested or need it (Fayyad et al., 1996). The overview of the KDD process can be seen in Figure 14. Similar to KDD the cross industry standard process for data mining (CRISP-DM)presented by Shearer (2000) comprises of a process model to conduct data mining projects through six phases including business understanding, data understanding, data preparation, modeling, evaluation, and deployment. According to Shearer (2000), the CRISP-DM process can be explained by Figure 15. As it can be seen in Figure 15, in this process the focus of business understanding phase is on defining the problem through assessing the current situation and understanding the business goals (Shearer, 2000). The results of business understanding lead, according to him, to the understanding of which data that need to be analyzed and how. The second phase of the model generally focuses on data collection and data quality verification, which is then the input to the data preparation as the third phase of the model (Shearer, 2000). Shearer (2000) further mean that the data modeling phase will be fed by the final data set provided through previous phase and will be evaluated in the next phase. Finally, the knowledge derived from the created model need to be organized and presented in a proper way to the users that can be achieved through processes included in the deployment phase (Shearer, 2000). The mentioned six phases of the process model by CRISP-DM are simplified by Runkler (2012) through introducing a four phase process model including preparation, preprocessing, analysis and post processing. The framework of this process model together with different sub steps of each phase can be seen in Figure 16. Figure 15 The CRISP-DM process (Shearer, 2000) 24 Three of the six areas suggested by Saxena and Srinivasan (2013) have parallels to the traditional view of analytics. BI is seen as traditional IT, decision making as traditional business and decision modeling as traditional analytics. Similarly, Grossman and Siegel (2014) believe the integration of analytics, business knowledge and IT as an important factor in defining the organizational BA framework. According to them analytics should be integrated to other operations in the organization and therefore it needs to be viewed as a value adding function of the organization. In addition, they believe having deep data analytics knowledge is an important element to create information from data and manage the information and this knowledge would not bring real value to the business unless it is completed with business knowledge. Kiron et al. (2011) also emphasize the importance of a data-oriented culture as it enables the company to act on the data. Furthermore, the knowledge about information technology tools and infrastructure also need to be available for applying the BA functions in the organization (Grossman & Siegel, 2014). See Figure 17 for a visualization of this framework. This indicates that all three of these business environments are included in BA, a statement which is supported by Laursen and Thorlund (2010) that views analytics as a bridge between the business- driven environment and the technically oriented environment (Figure 18). Analytics Business Knowledge Information Technology Knowledge about data and analytics Knowledge about business products, services and operation Knowledge about tools and infrustructure Figure 17 The organizational BA framework ( Grossman & Siegel, 2014) Figure 16 The BA process according to Runkler (2012) 25 Holsapple, Lee-Post and Pakath (2014) present a holistic perspective on BA. They present the Business Analytics Framework (BAF) developed from the many different definitions of BA. BAF consists of six core perspectives; a movement, capability set, transforming process, specific activities, practices & techniques and decisional paradigm. Parallels can be drawn between the BA processes described above and the core perspective a transformation process where “evidence is transformed via some process into insight or action” (Holsapple, Lee-Post & Pakath, 2014:14). This relates well to Davenport et al. (2001:128) definition that “the analytics process makes knowledge from data”. This statement identifies a need to differentiate between data and knowledge as well as a third concept, information, which is frequently mentioned when discussing BA. According to Laursen and Thorlund (2010) data is an information carrier while information is aggregated data. The two concepts are also different in their ability to be understood as data is hard to interpret without any processing which means converting it to information. The ability to interpret the data is important for converting it into knowledge which is the understanding you get from analyzing the data (Laursen & Thorlund, 2010). In addition, Laursen and Thorlund (2010) divide the Information into lead information and lag information depending on the use in the process. Lead information is used as an input to the process and supports decisions on what activities to prioritize while lag information is used to follow up on executed activities. If the activities have been performed before there is a record of lag information, which we can use to create lead information giving us a forecast for future activities (Laursen & Thorlund, 2010). Laursen and Thorlund (2010) further emphasize the importance of understanding the business requirements when conducting an analysis. This is in line with the corner stone models idea of putting the customer in the center (Bergman & Klefsjö, 2011). The authors also identify three areas that the analyst needs to define before analyzing the data. These areas are the overall problem, the delivery and the content. Laursen and Thorlund (2010) finally suggest interviews as a method for collecting these business requirements. Strategy creation Business processes Reporting and analytics Data warehouse Data sources and IT infrastructure Business-driven environment Technologically oriented environment In fo rm atio n req u irem en ts Figure 18 The BA process according to Laursen and Thorlund (2010) 26 3.2.1. Big data The amount of data produced in the world is increasing rapidly (Loshin, 2013), especially digital data (Mayer-Schönberger & Cukier, 2013). This has facilitated the use of new expressions such as big data. The meaning of big data is debated (Loshin, 2013). McKinsey for example define big data as data that is too big to store (Manyika et al., 2011) which would indicate that it is impossible to use big data. Gartner define it as “high-volume, high-velocity and high-variety information assets that demand cost-effective, innovative forms of information processing for enhanced insight and decision making” (Gartner, 2013) while Mayer-Schönberger and Cukier (2013) say that it is dependent on the degree to which the whole data set as opposed to a sample is used. Mayer-Schönberger and Cukier (2013) therefore say that data is abundant today and the need for sampling is reduced with big data. However, according to them the problems that can arise from big data make using it challenging. Some of these challenges according to Helland (2011) are related to data collection e.g. the data might come from different or unclear sources over a period of time. Another part of the challenges are related to data processing where a part of information might be lost during converting or transferring efforts. In addition, there is the risk of changes in data during data transaction and it means while processing the data received from a data source it might have changed right now at the origin source (Helland, 2011). 3.2.2. Data analysis Fayyad et al. (1996) stated that the data analysis method depends on the purpose of extracting knowledge from data. They divided the goals of knowledge extraction into two main categories as verification of the user’s hypotheses, and discovery of patterns in data. The discovery of the patterns is divided into prediction and description. The prediction refers to finding the patterns to predict the future of data patterns, and description is related to present data to the user in an understandable form (Fayyad et al., 1996). Similarly, Kenett and Shmueli (2009), classify the general data analysis goals into causal explanation, prediction and description. In addition, Laursen and Thorlund (2010) classify the analytics methods into hypothesis-driven, which is proper for when wanting to describe correlations of data in pairs, and data-driven, which is preferred when having a large amount of data which is constantly changed or updated and there is limited knowledge about the correlations in data. According to them, in case of using the data-driven method there are different techniques that can be applied depending on the purpose of the analysis. They believe if the purpose is to identify different kinds of patterns in data, one need to reduce the large number of variables to a smaller number without losing the information value and interpret different kind of information to know which factors really mean something. This can be done through the techniques such as data reduction to find the variables that contain information and are relevant to what we need, and cluster analysis that focuses on algorithms to combine observations that are similar (Laursen and Thorlund, 2010). However, if the purpose is to examine the correlation between given variables then data mining techniques can be applied for this reason (Laursen and Thorlund, 2010). Fayyad et al. (1996) mentioned data mining as the core of the process of KDD in order to discover the patterns in data and extraction. According to them, KDD is the overall process of extracting knowledge from data and data mining is a specific step in that process. Knowledge extraction, information discovery and information harvesting are some of the names historically used for data mining (Fayyad et al., 1996). However, they believed using data mining without considering the 27 statistical aspects of the problem can lead to discovering a significant pattern in data which in reality is insignificant. Therefore using a blind data mining can lead to the discovery of invalid or even meaningless patterns in data (Fayyad et al., 1996). In addition, according to Fayyad et al (1996), the patterns that are identified through the process of converting data into knowledge should have four main characteristics. These characteristics are validity, novelty, usefulness, and simplicity. The validity refers to the degree of certainty of the new data. Regarding the novelty the identified patterns need to be novel to the system and preferably to the user. The usefulness refers to containing benefit for the user, and simplicity means that the pattern should be understandable. 3.2.3. Presentation According to Orna (2005) there is a continuous transformation between information and knowledge through the organization since people use the information to create knowledge and in order to transfer the knowledge created in their mind to other users they present it in the shape of information. Communication is the factor that plays a key role in creating knowledge and affects the transformation process between information and knowledge (Orna, 2005). In other words, in order to create knowledge both information and communication are needed. Kenett and Shmueli (2009) mentioned effective communication as a factor that directly affects the quality of the information. In their studies among both research environment and industry, they realized that even if the analysis results have high quality, miscommunication can lead to the risk of misunderstanding of the results by the people. According to Marchese and Banissi (2013), knowledge visualization is a factor that leads to improved communication. Therefore proper knowledge visualization improves the business process in the organization. The focus of knowledge visualization specifically in the context of management is on using interactive graphics in a collaborative way to create, integrate and apply the knowledge (Marchses and Banissi, 2013). According to Few (2005), removing the distractions is a factor that contributes to effective communication. Regarding that, anything that does not lead to any added value and does not essentially contribute to the meaning of a graph is a distraction that negatively affects the communication (Few, 2005). One of the common distractions in graphical presentation such as charts and graphs are misuse of color. Overwhelming the user by using different colors without reason or using a mix of bright colors that visually harm the user are the common examples in misusing the color. Regarding this issue using soft colors which are lowly saturated and exist in nature in the graphs and using bright, dark or highly saturated colors only for making a specific data highlighted are recommended (Few, 2005). Tufte (2009) mentioned the issue of devoting too much of the ink to add unnecessary graphical features such as gridlines and detailed labels that do not contain added value for the viewer. Tufte (2009) further believe that the data graphics should lead the user`s attention to the meaning and substance of data and not to anything else. According to that theory, erasing non-data ink and redundant data-ink, maximizing the data-ink ratio and focusing on showing the data above all else are the principles that Tufte (2009) introduces regarding the data graphics theory related to the design options. 3.3. Synthesis of theoretical framework According to the literature related to the BA, several processes are introduced by different researchers. An overall view of mentioned processes is provided in Figure 19 in order to show the 28 relationship between different phases of them. Considering the overall view, although the first phase in different processes is named differently, the main idea is to identify the users requirements by for example identifying the business objectives, understanding the current status of the business and processes and identifying the decision needs. The preprocessing phase in the process introduced by Runkler (2012) is divided in two sub steps in the CRISP-DM and KDD but all of them follow a similar process. By comparing the data analysis phase in the different processes it can be realized that the main focus of the KDD is on data mining while the other processes emphasize no specific analysis method. The last phase before decision making in the different BA processes is named differently (interpretation, deployment and post processing) but the overall focus of all these phases is on interpretation and evaluation of the output. Figure 19 Comparison between BA processes As suggested in the figure above the processes have considerable overlaps between phases as well as a difference in level of granularity. In order to provide an appropriate level of detail as well as for the sake of clarity one process was chosen, the KDD by Fayyad (1996) (Figure 20). This process is frequently used in literature and the article in which it is presented is referenced 5842 times (Google scholar, 2014). The frequent use combined with the displayed similarities with other models indicates that KDD can be representative for BA processes. Figure 20 The suggested BA process (Fayyad, 1996) Earlier in the theory chapter a framework for displaying QM as a system of principles, practices and techniques was presented. Considering these QM principles, practices and techniques and the BA process presented above a framework for their relationship can be visualized in the following way (Figure 21). Selection Preprocessing Transformation Data Mining Interpretation/ Evaluation 29 Figure 21 Initial framework integrating QM and BA The corner stones presented by Bergman and Klefsjö (2011) should according to the them form the basis for the company culture, which then would require that it should be integrated in all steps of the BA process. Hellsten and Klefsjö (2000) also emphasized that the corner stones should be viewed in conjunction and not separately, the corner stones work together as a system. QFD as a practice is used to collect and translate customer needs into design requirements and on to production requirements (Lager, 2005). This aligns well with the purpose of the selection phase (Fayyad et al., 1996). The obvious phase to use QFD would therefore be the Selection phase. The same applies to the Kano model. Using QFD involves using techniques such as the HoQ, AIM, data collection and rating scales, which would then also be used to support the selection phase. Furthermore, In the first phase the “goal of the KDD process from the customer’s viewpoint” should be established (Fayyad et al., 1996:42). This could be supported by the stakeholder identification and ranking techniques such as customer roles and stakeholder ranking. If the goal should be based on the customers’ viewpoint they also need the opinions of customers which is facilitated by the collection of Voice of the Customer. Since the voice of the customer is qualitative data (Griffin & Hauser, 1993) and the seven management tools are designed to handle the verbal and qualitative information (Bergman & Klefsjö, 2011) the use of these techniques in the selection phase could be beneficial. For example, the affinity diagram that is one of the seven management tools could be used in order to group different customer needs together. The data mining phase consists of data analysis and a search for patterns (Fayyad et al., 1996). The seven improvement tools are used for structuring the numerical data and data analysis (Bergman & Klefsjö, 2011), therefore the use of these tools such as control charts and scatter plot would facilitate data analysis in this phase. However, based on the KDD goal different data analysis methods can be used in this phase (Fayyad et al., 1996). The improvement techniques that are used to support the data analysis can then be selected based on the data analysis method. Kano model Rating scales Stakeholder ranking AIM Seven management tools Selection Preprocessing Transformation Data Mining Interpretation/ Evaluation business analytics process q u ality m an agem e n t Voice of the customer Data collection Principles Practices Techniques Customer roles Quality Function Deployment Seven improvement tools 30 The blank cells in the framework represent no known relationship. The authors have not, through the literature review, found a way for QM to support all phases of BA. Therefore, the framework will be updated with the findings from the case study in section 4.8.6. BA on the other hand has the purpose to provide the right information to the right people at the right time (Laursen & Thorlund, 2010). This facilitates basing decisions on facts, which is one of the corner stones in quality management (Bergman & Klefsjö, 2011). According to Fayyad (1996) the last phase of the BA process (or KDD as he refers to it) is to evaluate and improve the process. This is in line with the quality management principle of continuous improvements (Bergman & Klefsjö, 2011). Grossman and Siegel (2014) as well as Laursen and Thorlund (2010) present BA as a bridge between different organizational functions and emphasize the need to understand the requirements on the BA process. This indicates a focus on customers at the same time as it involves more people and thereby lets more people be committed, both of which are principles in QM. 31 4. Results and analysis This chapter will show the results from the case study as well as analyze the results in order to answer the two research questions. 32 4.1. The case – Volvo The company chosen for this case study is Volvo GTT, a part of the Volvo Group. The study was performed at the Powertrain Engineering department in Gothenburg. The Volvo Group provides transport solutions on a global scale with 115000 employees (Volvo, 2014a) and a turnover of SEK 273 billion during 2013 (Volvo, 2014b). The group services markets in 190 countries through its manufacturing sites in 18 countries (Volvo, 2014a). The Volvo group is divided into 8 business entities; 3 sales & marketing entities, Group Trucks Operations (GTO), Group Trucks Technology (GTT), Construction Equipment, Business Areas and Volvo Financial Services. Group Trucks Technology work with product development while Group Trucks Operations work with manufacturing. Volvo GTT is the product development organization for trucks manufactured all over the world. The business entity employs 10 000 people worldwide (Volvo, 2014c). Sixty percent of R&D is conducted in Sweden (Volvo, 2014d) with the head quarter in Gothenburg. Volvo GTT is divided into seven departments; Product Planning, Project & Range Management, Complete Vehicle, Volvo Group Advanced Technology & Research, Volvo Group Powertrain Engineering, Vehicle Engineering and Volvo Group Purchasing (Volvo, 2014d). Volvo Group Powertrain Engineering is a global organization with 2000 employees in six countries Brazil, France, India, Japan, Sweden and USA. The Sweden main office of Powertrain engineering is located in Gothenburg with the work scope of engineering and design of engines, transmissions and drivelines for Volvo Group customers. The Gothenburg organization is the platform and application center for Heavy Duty engines as well as for Hybrids and Transmissions. The organizational chart of Powertrain Engineering in Sweden can be seen in Figure 22. Figure 22 Organizational structure at Volvo GTT PE Gothenburg 33 4.1.1. The COP and Hot test The product development process at Volvo PE includes a number of tests such as K1, K2 and certification tests. Two of these tests are called Conformance of Production (CoP) and Hot test. Although a part of the development process, these tests are initiated after the development efforts have ended and the tests are performed at the manufacturing sites by GTO. Despite the fact that the engines are manufactured by GTO the product ownership never shifts over. There is still a section within Volvo GTT PE that owns all the engine models. This section is called the maintenance and verification section. Because of this the tests are analyzed by specialists in Volvo GTT PE in order to find and solve issues surrounding the engine. The Hot test is a short test, less than 30 minutes, where mainly performance parameters such as power, torque, temperatures and pressures are measured. The test is performed at the end of the production line in special test rigs. The sampling of the Hot test is conducted so that new engines and engines with major changes are tested to 100% while engines that have been in production for a long time without any issues between 3% and 10% of the engines are tested. The test results from the Hot test therefore have a large sample size compared to the CoP test. The CoP test is a longer test, 15-30 hours, and mainly focused on measuring emission parameters such as NOx, carbon monoxide and soot although the test also measures some performance parameters. The overlap between the different test parameters are sometimes used to verify the Hot test results as the CoP test rigs have a better measurement accuracy. A long test time requires smaller sampling sizes for the CoP test. Just as with the Hot test the sample size depends on production volume, a high volume engine is tested more frequently than a low volume engine. 4.2. QFD as a supportive practice for business analytics As explained in the Theory chapter, QFD involves a number of steps (Franceschini, 2001) although there is a need to adapt the practice to a service such as BA (Mazur, 1993). With the steps suggested by Franceschini (2001) as base the following steps for QFD as a support for BA is suggested (Table 6) Table 6 Suggested stages and steps for QFD when supporting BA The process will be explained and justified in the context of the case study used to develop it. In the following section the case will be presented and each phase explained with examples from the case study. In section 4.6 a methodology is suggested. Determine who the customers are Understand the current situation Determine customer needs Prioritize the customer needs Analyze correlations between customer needs Identify quality attributes Draw a relationship matrix Summarize quality attribute weights Analyze correlations between quality attributes Identify actions Draw a relationship matrix Summarize actions weights Analyze correlations between actions Prioritize actions Assign actions to appropriate BA phase Requirements investigation Outcome planning Process planning Act on findings 34 4.3. Requirements investigation The first stage involves finding and evaluating the customer needs. The stage is divided into five steps; Determine who the customers are, Understanding the current situation, Determine customer needs, Prioritize customer needs and Analyzing correlations. Each step is further explained below. 4.3.1. Determine who the customers are According to the literature, identifying the customer needs, decision needs, and defining the goal of KDD are different expressions of the early phase of all mentioned business analytics process and the overall emphasize is on identifying the needs (Fayyad et al., 1996; Saxena & Srinivasan, 2013; Runkler, 2012). During the case study stakeholders to the test results were identified and ranked. The stakeholder identification and ranking is an important method for ensuring a customer focus in the BA process which is one of the principles of QM (Bergman & Klefsjö, 2011). This phase has the best potentials for fulfilling customer needs if the customers are first identified and their needs collected (Griffin & Hauser, 1993). Collecting the voice of the customer (VoC) enables BA to set up the BA process for greater customer satisfaction. As most customers to a BA process are internal customers the collection of V