Academia.eduAcademia.edu

Figure 33 – uploaded by Ossi Ylijoki

See full PDFdownloadDownload figure
Figure 5 reflects our process model in the theoretical context of our study. We have replaced the generic IT value generation model (Soh and Markus, 1995) with a more specific model that applies to big data value creation processes. As in Figure 1, the overlapping in Figure 5 is indicative. In the preceding sections, we have linked our model to current big data literature. The model aligns well with existing theories created before the era of big data, while providing new insights into the big data value creation processes.

Figure 5 reflects our process model in the theoretical context of our study. We have replaced the generic IT value generation model (Soh and Markus, 1995) with a more specific model that applies to big data value creation processes. As in Figure 1, the overlapping in Figure 5 is indicative. In the preceding sections, we have linked our model to current big data literature. The model aligns well with existing theories created before the era of big data, while providing new insights into the big data value creation processes.

Related Figures (78)

Figure 1. An illustration of the paradigm shift of the business landscape.
Figure 1. An illustration of the paradigm shift of the business landscape.
Figure 2. Research overview. This section clarifies the need for the study, the intention of the study, and the specific issues that this dissertation addresses. The first section provided an overview of the challenges which the current paradigm shift towards data-driven business poses to enterprises. The research problem stems from this phenomenon. After discussing the research problem, including related factors and the research gap, the “Research purpose” -section identifies the major idea of the research as well as defines several key terms. Thus, it is perhaps the most important passage in positioning the research. Finally, narrowing the purpose statement, the specific research questions to be answered in this research are presented. Figure 2 presents an overview of the research: The research problem stems from the paradigm shift caused by digital transformation that leads to (big) data deluge. The purpose of the research is to explore the impact of big data on the context of a single firm. The research questions set the scope of the research by pinpointing the specific topics that this research addresses using a pragmatic approach. The following three sections (“Research Problem”, “Research Purpose” and “Research Questions’) discuss the matters shown in Figure 2 in detail.
Figure 2. Research overview. This section clarifies the need for the study, the intention of the study, and the specific issues that this dissertation addresses. The first section provided an overview of the challenges which the current paradigm shift towards data-driven business poses to enterprises. The research problem stems from this phenomenon. After discussing the research problem, including related factors and the research gap, the “Research purpose” -section identifies the major idea of the research as well as defines several key terms. Thus, it is perhaps the most important passage in positioning the research. Finally, narrowing the purpose statement, the specific research questions to be answered in this research are presented. Figure 2 presents an overview of the research: The research problem stems from the paradigm shift caused by digital transformation that leads to (big) data deluge. The purpose of the research is to explore the impact of big data on the context of a single firm. The research questions set the scope of the research by pinpointing the specific topics that this research addresses using a pragmatic approach. The following three sections (“Research Problem”, “Research Purpose” and “Research Questions’) discuss the matters shown in Figure 2 in detail.
Figure 4. Example of text analysis tools of the KH Coder software. 1 Introduction
Figure 4. Example of text analysis tools of the KH Coder software. 1 Introduction
Figure 5. The number of big data articles by subject area and year.
Figure 5. The number of big data articles by subject area and year.
Figure 6. An exemplary illustration of Internet data volumes’. The resulting data volumes are huge. Figure 6 shows an example. This illustration presents, how Internet users (3.7 billion people in 2017) produced data using some popular services and platforms. The volumes are from 2017 and it should be noted that the numbers in the illustration are per minute.
Figure 6. An exemplary illustration of Internet data volumes’. The resulting data volumes are huge. Figure 6 shows an example. This illustration presents, how Internet users (3.7 billion people in 2017) produced data using some popular services and platforms. The volumes are from 2017 and it should be noted that the numbers in the illustration are per minute.
Figure 7. Global IP traffic growth forecast’. Moreover, the growth shows no signs of slowing down. On the contrary, in many areas (such as IoT), the trend of the growth is exponential. The number of people who have Internet access is growing, the number of smartphones is increasing, and the IoT concept is emerging, just to name a few factors. Thus, the amount of data is growing at an ever- increasing pace. Figure 7 displays a forecast of annual IP traffic globally. Compound annual growth rate (CAGR) is 23 % per year.
Figure 7. Global IP traffic growth forecast’. Moreover, the growth shows no signs of slowing down. On the contrary, in many areas (such as IoT), the trend of the growth is exponential. The number of people who have Internet access is growing, the number of smartphones is increasing, and the IoT concept is emerging, just to name a few factors. Thus, the amount of data is growing at an ever- increasing pace. Figure 7 displays a forecast of annual IP traffic globally. Compound annual growth rate (CAGR) is 23 % per year.
Figure 8. A visualisation of Laney’s (2001) 3V definition of big data. According to the suggestions made in the Publication I, this research distinguishes between data related actions, data usage related actions, and the big data phenomenon. The following definitions are used.
Figure 8. A visualisation of Laney’s (2001) 3V definition of big data. According to the suggestions made in the Publication I, this research distinguishes between data related actions, data usage related actions, and the big data phenomenon. The following definitions are used.
Figure 9. Virtual value creation process, modified from Rayport & Sviokla (1995). 2.2 From Data to Actionable Insights
Figure 9. Virtual value creation process, modified from Rayport & Sviokla (1995). 2.2 From Data to Actionable Insights
This research uses a multi-disciplinary view. A multidisciplinary approach is required to explain how enterprises can adopt a more data-driven approach in their businesses, as big data has profound effects on many aspects of an enterprise. Figure 10 illustrates the approach. Strategic management must be considered, as big data enables new ways of doing business. New ways indicate changes, which in turn require innovation. Thus, innovation research is one discipline related to this research. Big data naturally relates to information technology, and driving value from data requires human actions. This leads to information systems research. The combination of these three disciplines, information systems research, innovation research and strategic management research, is at the core of the theoretical playground relevant to this study. Figure 10. Multidisciplinary approach of the research.
This research uses a multi-disciplinary view. A multidisciplinary approach is required to explain how enterprises can adopt a more data-driven approach in their businesses, as big data has profound effects on many aspects of an enterprise. Figure 10 illustrates the approach. Strategic management must be considered, as big data enables new ways of doing business. New ways indicate changes, which in turn require innovation. Thus, innovation research is one discipline related to this research. Big data naturally relates to information technology, and driving value from data requires human actions. This leads to information systems research. The combination of these three disciplines, information systems research, innovation research and strategic management research, is at the core of the theoretical playground relevant to this study. Figure 10. Multidisciplinary approach of the research.
In order to gain sustainable competitive advantage, the competencies must meet the following criteria (Barney 1991): They must be valuable (i.e. able to create rents), rare, costly to imitate, and non-substitutable. Later, Barney (1995) attuned the last attribute into the form of the question “Is a firm organised to exploit the full competitive potential of its resources and capabilities?” Barney’s model is often referred as the VRIO-tool. Figure 11 illustrates how the key attributes (criteria) of the model can be used as a tool. This tool is useful for assessing both current capabilities and those resulting from the planned changes. Figure 11. The VRIO tool, modified from Barney (1991; 1995).
In order to gain sustainable competitive advantage, the competencies must meet the following criteria (Barney 1991): They must be valuable (i.e. able to create rents), rare, costly to imitate, and non-substitutable. Later, Barney (1995) attuned the last attribute into the form of the question “Is a firm organised to exploit the full competitive potential of its resources and capabilities?” Barney’s model is often referred as the VRIO-tool. Figure 11 illustrates how the key attributes (criteria) of the model can be used as a tool. This tool is useful for assessing both current capabilities and those resulting from the planned changes. Figure 11. The VRIO tool, modified from Barney (1991; 1995).
Figure 12. Business canvas framework (Furr & Dyer 2014). Furr & Dyer (2014) present a slightly modified version of the business canvas framework. Their business model components, shown in Figure 12, include the solution (which is further divided to value proposition and pricing strategy), cost structure (activities and resources), and customer acquisition (relationships and channels). This version focuses especially on the components that are most relevant to innovations, offering an appropriate tool for evaluating the impact of big data. Using a tool like the business model canvas helps to remain focused on the relevant issues and questions. For example, using the tool to compare a proposed business model with existing ones, or a company’s own or a competitor’s, reveals differences and possible weak spots.
Figure 12. Business canvas framework (Furr & Dyer 2014). Furr & Dyer (2014) present a slightly modified version of the business canvas framework. Their business model components, shown in Figure 12, include the solution (which is further divided to value proposition and pricing strategy), cost structure (activities and resources), and customer acquisition (relationships and channels). This version focuses especially on the components that are most relevant to innovations, offering an appropriate tool for evaluating the impact of big data. Using a tool like the business model canvas helps to remain focused on the relevant issues and questions. For example, using the tool to compare a proposed business model with existing ones, or a company’s own or a competitor’s, reveals differences and possible weak spots.
on firm’s resources and capabilities. Implementing evolutionary changes, such as an isolated, specific-purpose piece of software or integrating a firm’s systems may be valuable, e.g. in terms of reduced costs. However, reaching the maximal benefits requires more revolutionary approaches, namely redesigning the processes, redefining the business network, or coming up with a new business scope. In the turbulent era of digitalisation, firms should innovatively seek business model innovations and then leverage IT to create the required capabilities. Figure 13. IT-enabled business transformation framework (Venkatraman 1994),
on firm’s resources and capabilities. Implementing evolutionary changes, such as an isolated, specific-purpose piece of software or integrating a firm’s systems may be valuable, e.g. in terms of reduced costs. However, reaching the maximal benefits requires more revolutionary approaches, namely redesigning the processes, redefining the business network, or coming up with a new business scope. In the turbulent era of digitalisation, firms should innovatively seek business model innovations and then leverage IT to create the required capabilities. Figure 13. IT-enabled business transformation framework (Venkatraman 1994),
A short example clarifies the terms. Explicit knowledge about a smart phone refers to an understanding of the specifications and physical attributes of the device. Manufacturing processes contain a great deal of know-how, i.e. tacit-embodied knowledge. Self- transcending knowledge is the process of thoughts that led to the idea of the smart phone in the first place. In the business environment, recognising emerging opportunities are of course important. New ideas, seeds of innovations, often emerge sub-consciously, when our brain associates things with one another. Figure 14. Three forms of knowledge (Scharmer 2001).
A short example clarifies the terms. Explicit knowledge about a smart phone refers to an understanding of the specifications and physical attributes of the device. Manufacturing processes contain a great deal of know-how, i.e. tacit-embodied knowledge. Self- transcending knowledge is the process of thoughts that led to the idea of the smart phone in the first place. In the business environment, recognising emerging opportunities are of course important. New ideas, seeds of innovations, often emerge sub-consciously, when our brain associates things with one another. Figure 14. Three forms of knowledge (Scharmer 2001).
Table 1. Information systems resources, modified from Wade & Hulland (2004). The RBV explains how tangible or intangible resources and core competencies create a firm’s competitive advantage. Wade & Hulland (2004) have studied the combination of the RBV and information systems. They conclude that the RBV is a useful paradigm for information systems research, because it provides a valid model for evaluating the value of information systems. Based on a review of information systems research papers they proposed a set of key information systems resources, followed by a description of those resources using traditional resource attributes used in the RBV literature. This research establishes the connection between information systems resources and the RBV by building on their ideas as discussed below. Information system capabilities are summarised in the Table 1.
Table 1. Information systems resources, modified from Wade & Hulland (2004). The RBV explains how tangible or intangible resources and core competencies create a firm’s competitive advantage. Wade & Hulland (2004) have studied the combination of the RBV and information systems. They conclude that the RBV is a useful paradigm for information systems research, because it provides a valid model for evaluating the value of information systems. Based on a review of information systems research papers they proposed a set of key information systems resources, followed by a description of those resources using traditional resource attributes used in the RBV literature. This research establishes the connection between information systems resources and the RBV by building on their ideas as discussed below. Information system capabilities are summarised in the Table 1.
Table 2. A tool for evaluating IS resources with the RBV resource attributes, modified from Wade & Hulland (2004) and Barney (1995). Information systems research considers both humans and technology as important factors for technology adoption. This research focuses on big data value creation at the enterprise level. In an enterprise humans make the decisions that drive the adoption of big data and related technology. Defining the IS resources and connecting them with the RBV provides a vehicle to evaluate the potential competitive advantage of IS in general and big data in particular. Therefore, information systems research meets the needs of this research as one of the theoretical building blocks.
Table 2. A tool for evaluating IS resources with the RBV resource attributes, modified from Wade & Hulland (2004) and Barney (1995). Information systems research considers both humans and technology as important factors for technology adoption. This research focuses on big data value creation at the enterprise level. In an enterprise humans make the decisions that drive the adoption of big data and related technology. Defining the IS resources and connecting them with the RBV provides a vehicle to evaluate the potential competitive advantage of IS in general and big data in particular. Therefore, information systems research meets the needs of this research as one of the theoretical building blocks.
Numerous definitions of big data can be found in the literature, as well as among practitioners. Technology vendors, the public sector, private companies, consumers, and policy makers, among others, have interests in the field. Therefore, big data definitions were searched broadly in major reference databases (Scopus, ProQuest, and Web of Science) as well as Internet sources. At the end of the search process phase, 62 papers were identified to contain a definition of big data. The year-wise distribution of these papers is presented in Figure 15. Although the first definition was presented more than 10 years ago, the discussion on the definition of big data started only a few years back. Figure 15. Year-wise distribution of the found papers.
Numerous definitions of big data can be found in the literature, as well as among practitioners. Technology vendors, the public sector, private companies, consumers, and policy makers, among others, have interests in the field. Therefore, big data definitions were searched broadly in major reference databases (Scopus, ProQuest, and Web of Science) as well as Internet sources. At the end of the search process phase, 62 papers were identified to contain a definition of big data. The year-wise distribution of these papers is presented in Figure 15. Although the first definition was presented more than 10 years ago, the discussion on the definition of big data started only a few years back. Figure 15. Year-wise distribution of the found papers.
Figure 16. Evolution of the definition of big data. The included 62 papers were arranged by their publishing date, and each paper was inspected against previously published definitions. If the paper contained a new definition or added some new elements to the existing definitions, it was considered to be a new definition. This analysis resulted in 17 different definitions. These 17 definitions have similarities in the sense that many of them aim to widen the 3V definition to cover technical and especially business aspects. The rest of the papers (45) contained definitions which were essentially covered in earlier papers. The fishbone diagram in Figure 16 gives an overview of the evolution. The bones show the essential additions of all 17 different definitions, i.e. new aspects or components that each definition adds.
Figure 16. Evolution of the definition of big data. The included 62 papers were arranged by their publishing date, and each paper was inspected against previously published definitions. If the paper contained a new definition or added some new elements to the existing definitions, it was considered to be a new definition. This analysis resulted in 17 different definitions. These 17 definitions have similarities in the sense that many of them aim to widen the 3V definition to cover technical and especially business aspects. The rest of the papers (45) contained definitions which were essentially covered in earlier papers. The fishbone diagram in Figure 16 gives an overview of the evolution. The bones show the essential additions of all 17 different definitions, i.e. new aspects or components that each definition adds.
(HO: no difference in means, 2-sided test) Table 3. T-test results — moderator effects on behavioural intentions (INT).
(HO: no difference in means, 2-sided test) Table 3. T-test results — moderator effects on behavioural intentions (INT).
Figure 17. The model explaining the big data intentions of executives. On the basis of the regression analysis, t-tests and mean analysis described above, we drew the model shown in Figure 17. This statistically valid model explains 48.4 % of the variance in the behavioral intentions of the executives.
Figure 17. The model explaining the big data intentions of executives. On the basis of the regression analysis, t-tests and mean analysis described above, we drew the model shown in Figure 17. This statistically valid model explains 48.4 % of the variance in the behavioral intentions of the executives.
Figure 18. Respondents’ perceptions regarding big data by experience.
Figure 18. Respondents’ perceptions regarding big data by experience.
Content analysis is an established methodology for investigating textual data (Berelson 1952; Holsti 1969; Krippendorf 1989). Weber (1990) defines content analysis as a repeatable, systematic procedure that reduces the many words of a text to much fewer content categories. Novel applications of computerised content analysis have received the attention of scholars recently, e.g. (Han Hu et al. 2014; Lewis et al. 2013; Yu et al. 2014), as researchers wish to utilise new big data sources. In our case, manual coding of the texts of the 33 articles would have been a time-consuming job, and therefore we considered computerised content analysis to be a proper method for revealing common big data concepts and lessons learnt in the articles. We visualised the results with KH Coder software using co-occurrence maps. Co-occurrence maps build on the idea that words are related to the concepts they are connected to (Ryan & Bernard 2003). Osgood (1959) was among the first scholars to use co-occurrence matrices to reveal connected concepts in textual data. Figure 19. Co-occurrence map of the terms in the big data case study articles.
Content analysis is an established methodology for investigating textual data (Berelson 1952; Holsti 1969; Krippendorf 1989). Weber (1990) defines content analysis as a repeatable, systematic procedure that reduces the many words of a text to much fewer content categories. Novel applications of computerised content analysis have received the attention of scholars recently, e.g. (Han Hu et al. 2014; Lewis et al. 2013; Yu et al. 2014), as researchers wish to utilise new big data sources. In our case, manual coding of the texts of the 33 articles would have been a time-consuming job, and therefore we considered computerised content analysis to be a proper method for revealing common big data concepts and lessons learnt in the articles. We visualised the results with KH Coder software using co-occurrence maps. Co-occurrence maps build on the idea that words are related to the concepts they are connected to (Ryan & Bernard 2003). Osgood (1959) was among the first scholars to use co-occurrence matrices to reveal connected concepts in textual data. Figure 19. Co-occurrence map of the terms in the big data case study articles.
Several studies, e.g. (Davenport 2014; Manyika et al. 2011; Mayer-Schénberger & Cukier 2013) have made claims that big data causes pervasive changes, which will affect almost every sector of life. The cases con applications are emerging in various firmed the claims, at least partly. Clearly, big data walks of life. The studies recognised positive results and opportunities. However, several studies also reported challenges such as data inconsistencies and poor data quali strategies, lack of leadership, lack o new analytics and technology skills. y, security and/or privacy issues, missing analytics f data-driven organisational culture, and the need for These challenges reflect the disruptive nature of big data. They are indications of the major shifts required; changes that affect not only technical platforms and skills, but they also — and more importantly — influence the organisational culture, decision-making processes and management functions. 4 Research Contribution
Several studies, e.g. (Davenport 2014; Manyika et al. 2011; Mayer-Schénberger & Cukier 2013) have made claims that big data causes pervasive changes, which will affect almost every sector of life. The cases con applications are emerging in various firmed the claims, at least partly. Clearly, big data walks of life. The studies recognised positive results and opportunities. However, several studies also reported challenges such as data inconsistencies and poor data quali strategies, lack of leadership, lack o new analytics and technology skills. y, security and/or privacy issues, missing analytics f data-driven organisational culture, and the need for These challenges reflect the disruptive nature of big data. They are indications of the major shifts required; changes that affect not only technical platforms and skills, but they also — and more importantly — influence the organisational culture, decision-making processes and management functions. 4 Research Contribution
Figure 20. Research method of Publication IV, applied from Peffers et al. (2007). The study followed the design science research method (DSRM) framework (Peffers et al. 2007), which provides a nominal process for conducting and evaluating design science research in information systems. Figure 20 shows how the DSRM framework was applied in the study.
Figure 20. Research method of Publication IV, applied from Peffers et al. (2007). The study followed the design science research method (DSRM) framework (Peffers et al. 2007), which provides a nominal process for conducting and evaluating design science research in information systems. Figure 20 shows how the DSRM framework was applied in the study.
Figure 21. Innovation as a mediator framework. 4 Research Contribution
Figure 21. Innovation as a mediator framework. 4 Research Contribution
Figure 22. Algorithm follow-up principle. The accuracy of the algorithm was verified every month by comparing predicted outcomes with actual outcomes from the previous four months. Based on the results, the model may be re-trained. Figure 22 illustrates the principle. Field evidence shows that the model’s accuracy is 65-67%, when we look at the pipeline as a whole. However, the accuracy in the early phases of sales, i.e. in the bid qualification and planning stages, is significantly higher, and was 78-86%.
Figure 22. Algorithm follow-up principle. The accuracy of the algorithm was verified every month by comparing predicted outcomes with actual outcomes from the previous four months. Based on the results, the model may be re-trained. Figure 22 illustrates the principle. Field evidence shows that the model’s accuracy is 65-67%, when we look at the pipeline as a whole. However, the accuracy in the early phases of sales, i.e. in the bid qualification and planning stages, is significantly higher, and was 78-86%.
As a theoretical context Publication VI builds applies the IT value creation model by Soh & Markus (1995) and two value creation theories: the data, information, knowledge, wisdom (DIK W) hierarchy (Ackoff 1989; Zeleny 1987) and the virtual value creation (VVC) framework (Rayport & Sviokla 1995). The IT value creation model provides a starting point for the processes and the value creation models explain how the value cumulates. Current big data literature relevant to the study context is categorised topically (see Table 6). Moreover, the contingency factors that may either promote or demote the value creation are identified from the literature (+/- column).
As a theoretical context Publication VI builds applies the IT value creation model by Soh & Markus (1995) and two value creation theories: the data, information, knowledge, wisdom (DIK W) hierarchy (Ackoff 1989; Zeleny 1987) and the virtual value creation (VVC) framework (Rayport & Sviokla 1995). The IT value creation model provides a starting point for the processes and the value creation models explain how the value cumulates. Current big data literature relevant to the study context is categorised topically (see Table 6). Moreover, the contingency factors that may either promote or demote the value creation are identified from the literature (+/- column).
4.6 Publication VI — A Recipe for Your Big Data Cook Book Next, the literature is directed towards the IT value creation process. Organisations must make investments in order to achieve big data assets (asset creation process). We adopt an idea from Soh & Markus (1995), who presented the IT conversion process for turning IT expenditures into IT assets. Expenditures in this context are economic investments and spending that aims to, or supports, big data asset creation.
4.6 Publication VI — A Recipe for Your Big Data Cook Book Next, the literature is directed towards the IT value creation process. Organisations must make investments in order to achieve big data assets (asset creation process). We adopt an idea from Soh & Markus (1995), who presented the IT conversion process for turning IT expenditures into IT assets. Expenditures in this context are economic investments and spending that aims to, or supports, big data asset creation.
Figure 23 presents the model, consisting of four sequential processes. Big data assets connect the asset creation process to the capability creation process. Big data capabilities connect the capability creation process to the transformation process, which in turn links to the competition process via the impacts of big data. Using process theory terminology, the inputs and outputs are necessary but not sufficient conditions with regard to big data value creation. Figure 23. AC/TC process — creating economic value with big data.
Figure 23 presents the model, consisting of four sequential processes. Big data assets connect the asset creation process to the capability creation process. Big data capabilities connect the capability creation process to the transformation process, which in turn links to the competition process via the impacts of big data. Using process theory terminology, the inputs and outputs are necessary but not sufficient conditions with regard to big data value creation. Figure 23. AC/TC process — creating economic value with big data.
The results of this research answer the primary research question (RQ) — understanding and utilising big data in the transformation process towards big data driven business — by explaining the phenomenon as well as the value creation processes as described in Table 7, and by connecting theoretical aspects to practice as shown in Figure 24. Section 5.1 which describes the theoretical and practical implications discuss the results in more detail. Figure 24. Research contribution from theory to practice viewpoint.
The results of this research answer the primary research question (RQ) — understanding and utilising big data in the transformation process towards big data driven business — by explaining the phenomenon as well as the value creation processes as described in Table 7, and by connecting theoretical aspects to practice as shown in Figure 24. Section 5.1 which describes the theoretical and practical implications discuss the results in more detail. Figure 24. Research contribution from theory to practice viewpoint.
presented in various contexts in the articles. Therefore, to give an overview to the reader, Table 8 presents a condensed summary of the tools. For details, refer to articles. The level of abstraction of the tools is quite high, thus their usage in practical situations needs facilitating. In addition to the tools listed in Table 8, the whole dissertation can be seen as a high-level framework for big data adoption.
presented in various contexts in the articles. Therefore, to give an overview to the reader, Table 8 presents a condensed summary of the tools. For details, refer to articles. The level of abstraction of the tools is quite high, thus their usage in practical situations needs facilitating. In addition to the tools listed in Table 8, the whole dissertation can be seen as a high-level framework for big data adoption.
After removing the duplicates, we read the abstracts, and where necessary, the whole text of each of the resulting papers. We categorized the papers by using the following inclusion/exclusion criteria: If the paper contains a definition of big data, include it, otherwise reject it. Due to the loose search criteria, a number of papers defining other Our initial search covered three major reference databases: Scopus, ProQuest, anc Web of Science. We considered this as a good starting point, as these databases index a broad range of papers, covering both technical and business fields. Figure 1 gives ar overview of the search process. In addition to wide research questions, Kitchenhan (2007) suggests that mapping studies should use rather loose search criteria. We searched the databases (title, keywords, abstract) by using (“big data” anc “definition”) as a search string. All papers written in English and indexed up to 02. Sep-2015 were included in the initial result set. No additional limitations were set. A total of 479 papers were identified. Next, we removed duplicate articles (117).
After removing the duplicates, we read the abstracts, and where necessary, the whole text of each of the resulting papers. We categorized the papers by using the following inclusion/exclusion criteria: If the paper contains a definition of big data, include it, otherwise reject it. Due to the loose search criteria, a number of papers defining other Our initial search covered three major reference databases: Scopus, ProQuest, anc Web of Science. We considered this as a good starting point, as these databases index a broad range of papers, covering both technical and business fields. Figure 1 gives ar overview of the search process. In addition to wide research questions, Kitchenhan (2007) suggests that mapping studies should use rather loose search criteria. We searched the databases (title, keywords, abstract) by using (“big data” anc “definition”) as a search string. All papers written in English and indexed up to 02. Sep-2015 were included in the initial result set. No additional limitations were set. A total of 479 papers were identified. Next, we removed duplicate articles (117).
Fig. 2. Year-wise distribution of found papers. At the end of the search process phase, we had identified 62 papers that contained a definition of big data. The year-wise distribution of these papers is presented in Figure 2. It seems that although the first definition was presented more than 10 years ago, the discussion of the definition of big data started only a few years back. These papers and their definitions were examined further.
Fig. 2. Year-wise distribution of found papers. At the end of the search process phase, we had identified 62 papers that contained a definition of big data. The year-wise distribution of these papers is presented in Figure 2. It seems that although the first definition was presented more than 10 years ago, the discussion of the definition of big data started only a few years back. These papers and their definitions were examined further.
Fig. 3. Most common big data characteristics. During the following decade, trailblazers like Google and Amazon developed practical big data solutions. These solutions have proved to add value to their businesses. In fact, the trailblazers build their business models on big data solutions. An article published in 2008 in the Wired magazine (Anderson, 2008) aroused public interest in the use of big data and its effects in science. The next significant milestone was 2011, when McKinsey Global Institute and IDC published reports (Gantz and Reinsel, 2011; Manyika et al., 2011) that drew wide public attention to the potential value of big data. Since then a number of newspaper articles, scientific big data papers and books have been published.
Fig. 3. Most common big data characteristics. During the following decade, trailblazers like Google and Amazon developed practical big data solutions. These solutions have proved to add value to their businesses. In fact, the trailblazers build their business models on big data solutions. An article published in 2008 in the Wired magazine (Anderson, 2008) aroused public interest in the use of big data and its effects in science. The next significant milestone was 2011, when McKinsey Global Institute and IDC published reports (Gantz and Reinsel, 2011; Manyika et al., 2011) that drew wide public attention to the potential value of big data. Since then a number of newspaper articles, scientific big data papers and books have been published.
Fig. 4. Evolution of the definition of big data.
Fig. 4. Evolution of the definition of big data.
Table 1. Mapping the new perspectives of the big data definitions to the value chain.
Table 1. Mapping the new perspectives of the big data definitions to the value chain.
Fig. 5. Virtual value creation process.
Fig. 5. Virtual value creation process.
Definitions of big data sorted by date. The Definition column contains either a direct quotation from the paper or essential parts of the given/referenced definition. The New perspective column indicates what new component or aspect the definition has added to the previous ones.
Definitions of big data sorted by date. The Definition column contains either a direct quotation from the paper or essential parts of the given/referenced definition. The New perspective column indicates what new component or aspect the definition has added to the previous ones.
Figure 1 Research model (see online version for colours) Building on existing big data research, we applied the UTAUT constructs (Figure 1) to the context of our study as follows:
Figure 1 Research model (see online version for colours) Building on existing big data research, we applied the UTAUT constructs (Figure 1) to the context of our study as follows:
With regard to the sample, we are confident that we have a simple random sample, which represents the population. Finland Chamber of Commerce has studied the demographics of the management groups of Finnish companies (Linnainmaa and Turunen, 2014). Our respondents’ gender and age distributions shown in Figure 3 aligned well with their statistics. Moreover, the respondent distribution by industry aligns with the population distribution (Figure 2). ible 1 Cronbach’s alpha values
With regard to the sample, we are confident that we have a simple random sample, which represents the population. Finland Chamber of Commerce has studied the demographics of the management groups of Finnish companies (Linnainmaa and Turunen, 2014). Our respondents’ gender and age distributions shown in Figure 3 aligned well with their statistics. Moreover, the respondent distribution by industry aligns with the population distribution (Figure 2). ible 1 Cronbach’s alpha values
Figure 2 Companies and industries of the respondents The respondents were members of the management group of their companies (86.8%), IT managers (7.5%), or line-of-business executives (5.7%). Three of the respondents did not
Figure 2 Companies and industries of the respondents The respondents were members of the management group of their companies (86.8%), IT managers (7.5%), or line-of-business executives (5.7%). Three of the respondents did not
Figure 3 Characteristics of the respondents (see online version for colours)
Figure 3 Characteristics of the respondents (see online version for colours)
‘able 2 Results of regression analysis — significant constructs Regression analysis has some assumptions that must be met for the results to be reliable: linear relation between the response variable and predictors, normality, homoscedasticity, independence of errors, sufficient number of observations and absence of outliers. As the scatterplots of the residuals were linear and evenly varied, the linearity and
‘able 2 Results of regression analysis — significant constructs Regression analysis has some assumptions that must be met for the results to be reliable: linear relation between the response variable and predictors, normality, homoscedasticity, independence of errors, sufficient number of observations and absence of outliers. As the scatterplots of the residuals were linear and evenly varied, the linearity and
Note: HO: no difference in means, two-sided test. ‘able 3 T-test results — moderator effects on behavioural intentions (INT)
Note: HO: no difference in means, two-sided test. ‘able 3 T-test results — moderator effects on behavioural intentions (INT)
Table 4 Responses by construct and experience On the basis of the regression analysis, t-tests and mean analysis described above, we drew the model shown in Figure 4. This statistically valid model explains 48.4% of the variance in the behavioural intentions of the executives. In the next section we look more
Table 4 Responses by construct and experience On the basis of the regression analysis, t-tests and mean analysis described above, we drew the model shown in Figure 4. This statistically valid model explains 48.4% of the variance in the behavioural intentions of the executives. In the next section we look more
figure 4 The model explaining the big data intentions of executives (see online version for colours) Figure 5 Respondents’ perceptions regarding big data by experience (see online version for colours)
figure 4 The model explaining the big data intentions of executives (see online version for colours) Figure 5 Respondents’ perceptions regarding big data by experience (see online version for colours)
3.3 Factors explaining intentions
3.3 Factors explaining intentions
Figure 6 Respondents’ perceptions by statement (see online version for colours) Our effort expectancy scale (EEnn) consisted of four statements that measured the personal effort required to utilise big data whereas in the social influence scale (SInn) five statements probed the impact of peers. See the Appendix for the verbal representation of each statement. Almost six out of ten respondents with no big data experience considered that utilising big data in their area of responsibility is difficult (EE01), whereas more than eight out of ten of the experienced ones disagreed with this. The same tendency applied to the perceived difficulty of measuring the impact of big data (EE02). Another noteworthy difference could be seen in the role of peers. The social influence of other management group members and IT management was high among the
Figure 6 Respondents’ perceptions by statement (see online version for colours) Our effort expectancy scale (EEnn) consisted of four statements that measured the personal effort required to utilise big data whereas in the social influence scale (SInn) five statements probed the impact of peers. See the Appendix for the verbal representation of each statement. Almost six out of ten respondents with no big data experience considered that utilising big data in their area of responsibility is difficult (EE01), whereas more than eight out of ten of the experienced ones disagreed with this. The same tendency applied to the perceived difficulty of measuring the impact of big data (EE02). Another noteworthy difference could be seen in the role of peers. The social influence of other management group members and IT management was high among the
Figure 7 Respondents’ intentions to promote big data by experience (see online version for colours) Almost all respondents in both groups agree that big data was an important factor in their area of responsibility (INT01). However, there was a clear difference between the groups. Four out of ten non-experienced respondents were not actively looking for information about big data whereas almost all of the experienced ones were. The experienced respondents indicate strong agreement towards actions that would promote the utilisation of big data in their organisation: all respondents either agreed or strongly agreed with the statement “I make decisions that drive the utilisation of big data in my company” (INTOS).
Figure 7 Respondents’ intentions to promote big data by experience (see online version for colours) Almost all respondents in both groups agree that big data was an important factor in their area of responsibility (INT01). However, there was a clear difference between the groups. Four out of ten non-experienced respondents were not actively looking for information about big data whereas almost all of the experienced ones were. The experienced respondents indicate strong agreement towards actions that would promote the utilisation of big data in their organisation: all respondents either agreed or strongly agreed with the statement “I make decisions that drive the utilisation of big data in my company” (INTOS).
Table Al All respondents (N = 109) Table A2 presents the differences between respondents who had big data experience and those who did not. Mean values and their difference for each of the statements as well as for each of the constructs are given. Statements SI02 and EE04 were discarded based on Cronbach’s alpha tests.
Table Al All respondents (N = 109) Table A2 presents the differences between respondents who had big data experience and those who did not. Mean values and their difference for each of the statements as well as for each of the constructs are given. Statements SI02 and EE04 were discarded based on Cronbach’s alpha tests.
Big data is a multidisciplinary phenomenon. Unlike some other subject areas, big-data-related articles cannot be found only in certain highly focused forums. Although there are some new journals that Figure 1. Search process for big data case studies
Big data is a multidisciplinary phenomenon. Unlike some other subject areas, big-data-related articles cannot be found only in certain highly focused forums. Although there are some new journals that Figure 1. Search process for big data case studies
Table I. Big data case studies by application area
Table I. Big data case studies by application area
Figure 2. Number of big data case study articles by the publishing year
Figure 2. Number of big data case study articles by the publishing year
Figure 3. Geographical distribution of the big data cases
Figure 3. Geographical distribution of the big data cases
Figure 4. Co-occurrence map of the terms in big data case study articles
Figure 4. Co-occurrence map of the terms in big data case study articles
Table II. Guidelines for big data utilization Copyright © 2016 John Wiley & Sons, Ltd.
Table II. Guidelines for big data utilization Copyright © 2016 John Wiley & Sons, Ltd.
A.1 Big data case study articles
A.1 Big data case study articles
Figure 1. Research method, applied from Peffers et al. (2007).
Figure 1. Research method, applied from Peffers et al. (2007).
Figure 2. Innovation capabilities as a mediator framework. The data deluge and changing business models are challenging incum- bents to develop new capabilities. Approaches using resource-based theory have been popular among big data researchers recently, (Braganza, Brooks, Nepelski, Ali, & Moro, 2017; Gupta & George, 2016; Mikalef et al., 2017). We make no exception to the approach. However, we wish to add to the conversation the role of innovation capabilities in shaping the business model, as shown in Figure 2. In the big data context, we distinguish between “data- and human-driven” innovations.
Figure 2. Innovation capabilities as a mediator framework. The data deluge and changing business models are challenging incum- bents to develop new capabilities. Approaches using resource-based theory have been popular among big data researchers recently, (Braganza, Brooks, Nepelski, Ali, & Moro, 2017; Gupta & George, 2016; Mikalef et al., 2017). We make no exception to the approach. However, we wish to add to the conversation the role of innovation capabilities in shaping the business model, as shown in Figure 2. In the big data context, we distinguish between “data- and human-driven” innovations.
Guidelines for assessing the value of a predictive algorithm: a case study Operationalization unacceptable to SWCo. Thus, the probability of loss is significant, and cancelled bids can be treated as losses.
Guidelines for assessing the value of a predictive algorithm: a case study Operationalization unacceptable to SWCo. Thus, the probability of loss is significant, and cancelled bids can be treated as losses.
Fig. 1 SWCo’s bid prediction concept The accuracy of the algorithm is verified every month by comparing predicted outcomes with actual outcomes from
Fig. 1 SWCo’s bid prediction concept The accuracy of the algorithm is verified every month by comparing predicted outcomes with actual outcomes from
Zadrozny and Elkan 2001). Table 3 presents a binary classification matrix for possible outcomes of our algo- rithm. If the algorithm predicted a no-win, and the actual outcome was also a no-win, the prediction was correct (true negative). Correspondingly, a true positive represents a correct prediction. False negatives and false positives are misclassified bids. Ntn, Nfn, Nfp, and Ntp represent the number of bids in each class. The total number of bids Ntot = Ntn + Nfn + Nfp + Ntp. Guidelines for assessing the value of a predictive algorithm: a case study One business consideration is the risk of lost opportu- nities due to misclassification as no-win (false negatives). A classification matrix with actual, historical data provide data for estimation. The percentage of actual false nega- tives (Nfn% = Nfn/Ntot * 100) against predictions tells us the risk level. If the Nfn% has been e.g. 5-6% in the past, we can assume that it will hold also in the near future. In other words, for each 100 opportunities, we will most probably miss five or six due to incorrect prediction. This is data-based, hard evidence, at least when compared to personal opinions and valuations that may change over time, and case by case. Thus, Nfn% offers a coherent basis for decision making. Moreover, personal judgement, or expertise may always override predictions, as long as the elimination of bids is not fully automated.
Zadrozny and Elkan 2001). Table 3 presents a binary classification matrix for possible outcomes of our algo- rithm. If the algorithm predicted a no-win, and the actual outcome was also a no-win, the prediction was correct (true negative). Correspondingly, a true positive represents a correct prediction. False negatives and false positives are misclassified bids. Ntn, Nfn, Nfp, and Ntp represent the number of bids in each class. The total number of bids Ntot = Ntn + Nfn + Nfp + Ntp. Guidelines for assessing the value of a predictive algorithm: a case study One business consideration is the risk of lost opportu- nities due to misclassification as no-win (false negatives). A classification matrix with actual, historical data provide data for estimation. The percentage of actual false nega- tives (Nfn% = Nfn/Ntot * 100) against predictions tells us the risk level. If the Nfn% has been e.g. 5-6% in the past, we can assume that it will hold also in the near future. In other words, for each 100 opportunities, we will most probably miss five or six due to incorrect prediction. This is data-based, hard evidence, at least when compared to personal opinions and valuations that may change over time, and case by case. Thus, Nfn% offers a coherent basis for decision making. Moreover, personal judgement, or expertise may always override predictions, as long as the elimination of bids is not fully automated.
Table 3 A classification matrix for bid qualification
Table 3 A classification matrix for bid qualification
Guidelines for assessing the value of a predictive algorithm: a case study algorithms live on data. Data often reside in silos, meaning that gathering, harmonizing, and processing the data repre- sents a significant effort. Moreover, a great deal of poten- tially useful data are unstructured, like sales presentations and documents. One potential improvement for producing high quality data for sales predictions is to first use a sys- tematic approach for metrics definition (e.g. Monat 2011; Neely et al. 1997), and then record the data during the sales process. Another approach would be to utilize the metadata generated by the ever-increasing B2B sales platforms. separation of algorithm implementation and management accounting, which eases the division of work. Management accounting expertise is often available in-house, whereas specialists capable of building predictive analytics appli- cations are a scarce resource. Loose coupling between the algorithm and costing components eases change manage- ment, e.g. sales cost allocation rules may be changed without touching the algorithm. Moreover, the separation allows applying the method to existing solutions.
Guidelines for assessing the value of a predictive algorithm: a case study algorithms live on data. Data often reside in silos, meaning that gathering, harmonizing, and processing the data repre- sents a significant effort. Moreover, a great deal of poten- tially useful data are unstructured, like sales presentations and documents. One potential improvement for producing high quality data for sales predictions is to first use a sys- tematic approach for metrics definition (e.g. Monat 2011; Neely et al. 1997), and then record the data during the sales process. Another approach would be to utilize the metadata generated by the ever-increasing B2B sales platforms. separation of algorithm implementation and management accounting, which eases the division of work. Management accounting expertise is often available in-house, whereas specialists capable of building predictive analytics appli- cations are a scarce resource. Loose coupling between the algorithm and costing components eases change manage- ment, e.g. sales cost allocation rules may be changed without touching the algorithm. Moreover, the separation allows applying the method to existing solutions.
en nnn EN NO EN LOI IOI FE PINOLE EOI IOI Process theories differ from variance theories in several ways. The differences are summarized in Table I, adapted from Mohr (1982) and Soh and Markus (1995). Markus and Robey (1988)
en nnn EN NO EN LOI IOI FE PINOLE EOI IOI Process theories differ from variance theories in several ways. The differences are summarized in Table I, adapted from Mohr (1982) and Soh and Markus (1995). Markus and Robey (1988)
ee ee dl eee ee i Fi igure 1 summarizes the theoretical background of our research. All three models that we build on are sequential in nature. The models overlap or limit each other. However, the overlapping shown in Figure 1 is indicative. The models shed light on the value creation processes from three different perspectives. The information systems view, ie. VVC framework, provides building blocks for software implementations, the DIKW perspective explains the knowledge creation process, and the IT value creation theory enlightens the related processes.
ee ee dl eee ee i Fi igure 1 summarizes the theoretical background of our research. All three models that we build on are sequential in nature. The models overlap or limit each other. However, the overlapping shown in Figure 1 is indicative. The models shed light on the value creation processes from three different perspectives. The information systems view, ie. VVC framework, provides building blocks for software implementations, the DIKW perspective explains the knowledge creation process, and the IT value creation theory enlightens the related processes.
Taken together, the four processes together with affecting +/— factors (discussed in Table III, see also Table II) and success factors combine a recipe for big data value creation. The model explains how big data and related technologies can contribute to firm performance. It takes into account the necessary conditions, such as big data assets and capabilities, while it also recognizes they are not sufficient on their own: each of the processes is probabilistic, Le. it may fail to produce desired outcomes due to additional conditions that are unpredictable, and often difficult to manage in variance-theory-based approaches. A process-theory-based model allows us to build a holistic end-to-end view. Numerous variance-theory-based studies that cover certain parts of the whole support the view. We named the model AC/TC, taking first letters from each process and adding a slash to indicate the technology/performance conversion.
Taken together, the four processes together with affecting +/— factors (discussed in Table III, see also Table II) and success factors combine a recipe for big data value creation. The model explains how big data and related technologies can contribute to firm performance. It takes into account the necessary conditions, such as big data assets and capabilities, while it also recognizes they are not sufficient on their own: each of the processes is probabilistic, Le. it may fail to produce desired outcomes due to additional conditions that are unpredictable, and often difficult to manage in variance-theory-based approaches. A process-theory-based model allows us to build a holistic end-to-end view. Numerous variance-theory-based studies that cover certain parts of the whole support the view. We named the model AC/TC, taking first letters from each process and adding a slash to indicate the technology/performance conversion.
SR a SAK SRR ee RT ES eeTIS ae SNM ma N SEE OnE TaT SeER Our research acknowledges well-established and widely used theories created by other scholars. The elements our model builds on — DIKW hierarchy, VVC framework and process
SR a SAK SRR ee RT ES eeTIS ae SNM ma N SEE OnE TaT SeER Our research acknowledges well-established and widely used theories created by other scholars. The elements our model builds on — DIKW hierarchy, VVC framework and process
Ossi Ylijoki can be contacted at: ossiylijoki@phnet.fi
Ossi Ylijoki can be contacted at: ossiylijoki@phnet.fi
Related topics:
Business
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved