A study on the relationship between big data analytics capability, business process agility, and firm performance

(1)

Master Thesis

A Study on the Relationship Between Big Data Analytics Capability, Business

Process Agility, and Firm Performance

Thesis MSc. Business Administration

Strategy Track

Name Student number Supervisor Roy Schoonderbeek 11383054 dr. Andreas Alexiou

(2)

Statement of originality

This document is written by Student Roy Schoonderbeek who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

Abstract

This paper examines the relationship between Big Data Analytics Capability (BDAC), Business Process Agility (BPA), and Firm Performance (FPER). It tries to find an answer to the following research question: What is the key mechanism that drives the relationship

between Big Data Analytics Capability (BDAC) and Firm Performance (FPER)? It tries to

identify Business Process Agility as mediator in the established relationship between Big Data Analytics Capability and Firm Performance. To test the proposed conceptual model, an online survey was distributed among Big Data Managers at European firms. The findings confirm a direct effect of Big Data Analytics Capability on Firm Performance and of Big Data Analytics Capability on Business Process Agility. The study failed to find a direct effect of Business Process Agility on Firm Performance, so a mediating role of Business Process Agility could not be confirmed. In the end, implications for practice and future research are discussed.

Keywords: Big Data Analytics, Big Data Analytics Capability, Business Process Agility, Firm Performance

(4)

1. Introduction

Data is continuously collected, stored, and shared around the world by different actors and institutions. Governments, businesses, and people are doing it. Every second

incommensurable amounts of data are being generated and shared across the globe (Agarwal & Dhar, 2014). Big data is booming, and companies have developed new technologies that create the possibility to collect more data than ever before. Big data has the potential to transform the entire business process, and Gobble (2013) even identifies big data as the next big thing in innovation. It has caught the attention of almost every company in every industry, and every manager around the world is looking for guidance and understanding of big data and the role it could play in strategizing (Mazzei & Noble, 2017). Wamba, Gunasekaran, Akter, Ren, Dubey and Childe (2017) see the emergence of big data as a result of the increasing adoption and usage of Big Data Analytics (BDA) technologies, tools, and

infrastructure. Part of that is the rise of social media, mobile devices, Internet of Things (IoT), and cloud-based platforms which are used by organizations to gain a sustainable competitive advantage over their competitors (Wamba et al., 2017).

MIT Sloan Management Review partnered up with IBM Institute for Business Value to

create new insights into the importance and emergence of big data and BDA. One of their key findings was that top-performing organizations use analytics five times more than lower performing organizations ( LaValle, Lesser, Shockley, Hopkins & Kruschwitz, 2011). This clearly shows that analytics offer new value to the business. A growing number of managers want to run their businesses based on data-driven decisions. Though, the biggest challenge for organizations at the moment, according to LaValle et al. (2011), is not about the collection of data, but about the managerial and cultural aspects of data. Organizations can invest in big data, but will still face many challenges regarding the management of their data, since there is no substantial amount of empirical evidence on the relationship and the nature of the

(7)

relationship between big data analytics capability (BDAC) and firm performance (FPER). Analytics-driven insights must be closely linked to the business strategy of a company, it must be embedded in the organizational processes, and it is also important to be easy to understand for end-users so they can use it to enhance the functioning of the organization (LaValle et al., 2011). As Akter, Wamba, Gunasekaran, Dubey, and Childe (2016) propose, further research is necessary to find out how exactly FPER and BDAC are connected, what other variables play a role in this relationship, and how organizations can use BDAC to create valuable insights for their business. They also suggest that the mediating role of business process agility (BPA) can be a possible topic to further look into.

However, even though firms are paying increasing attention to the role of BPA and BDAC, not enough is known about how both variables influence FPER (Akter & Wamba, 2016). In existing research, Akter and Wamba (2016) concluded that BDA provides

increasing value to firms with businesses in e-commerce (Akter & Wamba, 2016). Wamba et al. (2017) looked at BDA, FPER, and the effect of dynamic capabilities and found a direct impact from BDAC on FPER. Chen, Wang, Nevo, Jin, Wang, and Chow (2014) looked at the mediating effect of BPA on the relationship between IT capability and organizational

performance. They have concluded that BPA fully mediates the relationship between IT capability and organizational performance (Chen et al., 2014).

This study builds on the findings of Wamba et al. (2017), who established a positive effect of BDAC on FPER. It investigates a possible mechanism that explains this relationship. In other words, the research question of this study is:

What is the key mechanism that drives the relationship between Big Data Analytics Capability (BDAC) and Firm Performance (FPER)?

(8)

The expectation is that BPA mediates the relationship between BDAC and FPER. That means that BDAC will influence BPA, which is already stated in prior research. A recent study by Côrte-Real, Oliveira, and Ruivo (2017) demonstrated that BDA applications which are based on effective knowledge management could help firms to create organizational agility. In the paper of Lu and Ramamurthy (2011), the authors found that IT capability enhances market capitalizing agility, as well as operational adjustment agility. Raschke (2010) concluded that IT functions as a platform for agility. In the light of these reports, it is conceivable that the deployment of BDAC leads to BPA (Côrte-Real et al., 2017; Lu & Ramamurthy, 2011; Raschke, 2010).

Previous studies have emphasized the relationship between BPA and FPER as well. Chen et al. (2014) have found that BPA plays a mediating role in the relationship between IT capability and performance. Ray, Barney, and Muhanna (2004) also have suggested that the possession of BPA as a capability forms a basis unit for competitive advantage, and thus for performance. Lastly, Raschke (2010) has provided evidence that BPA leads to more

efficiency and outcomes of better quality. These statements strongly suggest that the

possession of BPA can lead to a better FPER. In other words, these prior studies suggest that BPA explains why the deployment of BDAC enhances FPER. The goal of this study is to investigate this relationship between BDAC and FPER further and look for a possible mediation effect of BPA. If such a mechanism will be uncovered in this study, this will contribute to the existing literature on BDAC, BPA, and FPER. It will add to the already uncovered mechanisms that drive the relationship between BDAC and FPER and will provide insights into the changing nature of emerging IT capabilities, like BDAC.

The structure of the remainder of this paper is as follows: in the first part, different definitions and terminologies of BDA, BDAC, BPA, and FPER are given. Besides that,

(9)

several conducted studies on big data, BDAC, agility, and performance and the relationships between these concepts will be discussed. This part will also discuss the development of the hypotheses and the presentation of the research model. The next part will contain an overview of the used research design and method of data collection and analysis. The section after that presents the analysis of the data, findings, discussion, and implications for practice and further research. In the final section, a short conclusion will be drawn about the findings and

(10)

2. Literature Review and Hypothesis Development

In this section, prior literature on the topics of big data, big data analytics, business process agility, and firm performance is presented. It provides an overview of existing studies, and it forms the basis for this research. This fundament is used to derive hypotheses from, by which will be tried to close the earlier mentioned research gap.

2.1 Big Data

The hype around big data can be attributed to the campaigns by IBM and other leading tech companies to promote the market of BDA (Gandomi & Haider, 2015). To get more understanding of the topic of research, it is essential to have some clear definitions to work with. There is not much consensus about the concept, definition, and characteristics of big data. Because of the rapid evolvement of the concept and definition of big data, some confusion among scholars and managers has emerged. Beyer and Laney (2012) describe big data as massive, complex, and real-time data, that needs analysis, processing, and

management to derive value from it. The definition of Schroeck, Shockley, Smart, Romero-Morales, and Tufano (2012) has a broader scope on information, which includes real-time information, non-traditional forms of media data, new technology-driven data, the large volume of data, the latest buzz-word, and social media data. Laney (2001) introduced Volume,

Variety, and Velocity as The Three Vs, the three dimensions of challenges related to the

management of data. The Three Vs have become an accepted and standard way to describe big data (Chen, Chiang & Storey, 2012; Gandomi & Haider, 2015; McAfee & Brynjolfsson, 2012).

The Three Vs will be described next. The first V is Volume, which refers to the vast amount of data. According to a survey by IBM in 2012, more than half of the respondents considered datasets that exceed one terabyte as being big data (Schroeck et al., 2012). In fact,

(11)

it is difficult to define a specific threshold for big data volumes, since the volumes are relative and vary by different factors (f.i. time and type of data) and among various industries

(Gandomi & Haider, 2015).

The second V is Variety, which refers to the heterogeneity within a dataset. Data can be structured, semi-structured, or unstructured. The term variety also emphasizes the various sources and formats where data is generated (Russom, 2011). Examples of forms of data are messages, images, GPS signals, videos and more. It includes both structured and unstructured data.

The third V is Velocity, referring to the rate of creation of new data and the speed at which it should be analyzed and transferred (Gandomi & Haider, 2015; Hashem, Yaqoob, Anuar, Mokhtar, Gani & Khan, 2015; Russom, 2011). Data have been created faster and faster with the increasing use of new devices like smartphones and tablets, social media platforms, and the emergence of Internet of Things (IoT).

Some researchers and scholars also add another two Vs to define big data; Value and

Veracity. Forrester (2012) is talking about a fourth V, which stands for the value of big data.

This points out the importance of the data and the value and economic benefits it can give companies and society. The density of the value of big data is seen as low, the value relative to its volume is low. Nevertheless, a high value can be created if large volumes of data are being analyzed.

Lastly, White (2012) and IBM added a fifth V to this typology, which is Veracity. This fifth dimension highlights the importance of data of good quality and the needed level of trust in the vast variety of data sources. Even uncertain data can contain valuable information, so dealing with imprecise and uncertain data is also a part of big data. Several tools are

(12)

Manyika, Chui, Brown, Bughin, Dobbs, Roxburgh and Byers (2011) identified five dimensions on the creation of business value via big data solutions. These dimensions are as follows: data policies, technology and techniques, organizational change and talents, access to data, and industry structure (Manyika et al., 2011). They also emphasize the requirements for storage and analysis of big data, while other authors like Davenport, Barth, and Bean (2012) target the variety of sources big data comes from. This is a clear example of the different aspects of big data that are pointed out by various scholars. The final definition Wamba, Akter, Edwards, Chopin, and Gnanzou (2015) give for big data is: "a holistic approach to manage, process and analyze 5 Vs to create actionable insights for sustained value delivery, measuring performance and establishing competitive advantages" (p. 235).

2.2 Big Data Analytics Capability

To derive sustainable value from big data, proper analysis is key. Since big data is a booming topic, BDA is a relatively new and hot field among scholars as well. It is also seen as a requirement to manage BDA and BDA adoption, according to Verma and Bhattacharyya (2017). As Gandomi and Haider (2015) state, the demand for new technologies and

development in analytical methods has increased, because of the new emerging challenges regarding management and value creation of data. There are different definitions of BDA, used by different authors and researchers. One definition of BDA is: a holistic approach to managing, processing, and analyzing the 5 V data-related dimensions (i.e., volume, variety, velocity, veracity, and value) to create actionable ideas for delivering sustained value,

measuring performance and establishing competitive advantage (Wamba et al., 2015; Wamba et al., 2017). According to the existing literature on big data, big data analytics capability (BDAC) has three fundamental building blocks; organizational, physical, and human. The critical challenges for BDAC are identified by McAfee and Brynjolfsson (2012) as being

(13)

leadership, talent management, IT infrastructure, decision-making capability, and the company culture.

Côrte-Real, Oliveira, and Ruivo (2017) talk about BDA as being a new generation of technology and architecture, which is developed and designed to extract value and insights from data with large volumes and a wide variety. Chen, Chiang and Storey (2012) and Russom (2011) see BDA as an aspect of business intelligence and analytics since it can be technologies which support the creation of insightful trends in business intelligence. That means that it is time to act on the growing importance of big data technology and start focusing on other resources which are crucial for building firm-specific BDAC (Gupta & George, 2016). Gupta and George (2016) classify important resources for BDAC as tangible, human, and intangible. As necessary tangible resources, they name data, technology, and basic resources (e.g., time and investments). They divide human resources into two type of skills; managerial skills and technical skills. Finally, two essential intangible resources for building this capability, are a data-driven culture and an intense form of organizational learning. Firms need a specific combination of these tangible, human, and intangible assets to successfully build up BDAC (Gupta & George, 2016).

2.3 Business Process Agility

Agility has received much attention from scholars as well. The Economist Intelligence Unit conducted a survey among executives (Glenn, 2009) and found out that 88% of the executives identify agility as the key to global success. According to Raschke (2010), agility consists of two essential competencies; it is about adapting to change in a responsive manner. The increasing attention for agility may be caused by the growing belief that agility may provide firms with the ability to improve their business and business processes efficiently and

(14)

2014). Business process agility (BPA) is an important form of organizational agility, which is of particular relevance to Information Systems (IS) research. It is defined as follows by Tallon (2008); the extent to which firms can easily and quickly retool their business processes to adapt to the market environment. The concept of BPA highlights the ease and speed with which companies engage in key business activities (Tallon, 2008). It is seen as an important mechanism through which firms interact with the market environment and can explain inter-firm performance variance over time (Raschke, 2010). Agile business processes can also form the basis for firms to exploit opportunities for innovation and competition (Sambamurthy, Bharadwaj & Grover, 2003). Raschke (2010) emphasizes that BPA has four main

components; reconfigurability, responsiveness, employee adaptability, and a process-centric view. She considers reconfigurability and responsiveness as the fundament of agility, while employee adaptability and a process-centric view embody the human aspect related to the processes. BPA is seen as a dynamic capability, Teece (2007) categorizes agility as a higher-order dynamic capability that emerges over time. Firms can use it to quickly respond to environmental opportunities and threats.

Despite the increasing attention for BPA, there is not much knowledge about how to become more agile as an organization (Sambamurthy et al., 2003). Therefore, it could be stated that BPA is a rare capability to possess (Chen et al., 2014). As Raschke (2010) proposes, BPA gives firms the opportunity to redesign business processes and develop new processes, to have the ability to take advantage of unclear market conditions. This is part of the organizational routines of a firm, which makes it difficult for competitors to distinguish the valuable parts of the business processes. As Chen et al. (2014) state, BPA is inimitable and non-substitutable, so it could be seen as a strategic organizational capability.

(15)

2.4 Big Data Analytics Capability and Firm Performance

As already found out by scholars, BDA is also expected to have an enormous impact across various industries. For example, the healthcare industry has decreased the operational costs and enhanced the quality of life by using BDA (Liu, 2014). In the retail sector,

companies are using BDA to improve the customer experience (Tweney, 2013). Existing literature on BDA has discovered and described a positive relationship between the analysis of customers and firm performance (Germann, Lilien, Fiedler & Kraus, 2014). BDA

technologies can provide firms with the ability to create a competitive advantage over their competitors (Chen, Chiang & Storey, 2012). Using and adopting BDA can provide a

competitive advantage for businesses in several ways. Examples of these types of competitive advantages are increased operational efficiency, better customer service, identifying new customer and markets, and development of new products and services (Chen et al., 2012). According to the paper of Wamba et al. (2017), BDA has started to become a critical factor that can potentially change the game for managers, because it enables improved business efficiency and effectiveness (Atyeh, Jaradat & Arabeyyat, 2017). It can add value for the firm and even lead to higher performance. So, it seems logical that big data technology will grow as an essential asset for companies.

Brands (2014) found out that BDA enables the possibility for firms to analyze and make strategic decisions by looking through a data lens. Gupta and George (2016) stated that the deployment of BDAC leads to superior firm performance. Côrte-Real, Oliveira, and Ruivo (2017) conclude their research with a statement that proposes that organizations need an integrated view on their BDA activities, to optimally employ these capabilities and achieve a competitive advantage over their opposition. According to them, BDA can be a useful tool to survive in competitive, changing environments. European firms tend to value external insights

(16)

contributed by BDAC more than insights from internal knowledge management systems (Côrte-Real et al., 2017).

So, from the previous paragraphs can be concluded that several studies on the topic of BDAC have identified a positive relationship between the use of BDA by a firm and the enhancement of abilities that are positively related to the performance of firms (Chen et al., 2014; Germann et al., 2014; Gupta & George, 2016; Liu, 2014; Tweney, 2013; Wamba et al., 2017). The deployment of BDA can provide firms with a competitive advantage over their competitors, so, therefore, could be hypothesized that:

H1: The deployment of BDAC will have a positive effect on FPER.

2.5 Big Data Analytics Capability and Business Process Agility

Previous studies found relationships between BDAC and BPA as well. Companies in the manufacturing industry are using BDA to enhance the processing of business processes and assets (Davenport, Barth & Bean, 2012), and to facilitate smoother business

transformation (Gardner, 2013). BPA is also linked to the operational flexibility of IT systems and organizational processes which can support (un)structured change (Chen et al., 2014). Agility is vital for companies to survive in a rapidly changing environment and BDA can help them to do that (Côrte-Real et al., 2017). The deployment of BDAC will provide firms with large amounts of data. Analyzing this growing amount of data provides real-time information, which can result in a more agile organization than competitors (McAfee & Brynjolfsson, 2012). As Lu and Ramamurthy (2011) conclude their paper, more investments in IT do not necessarily support agility, but investments to enhance IT capabilities will. Those investments can support organizations to become agile eventually. This is also confirmed by Tallon (2008) since he states in his paper that IT capabilities have a positive impact on the agility of an

(17)

organization. BDAC is an example of an IT capability, so investments in BDAC could help to build agile organizations. On the basis of these prior studies, the following hypothesis is developed:

H2: The deployment of BDAC will have a positive effect on BPA.

Côrte-Real et al. (2017) also conclude that agility directly leads to a better

performance on process level and regarding competitive advantage. They place a remark on that; several paths could lead to competitive advantage, so managers should not focus on agility only. Côrte-Real et al. (2017) advise organizations to invest in BDAC tools, to support agility within the firm, which could eventually lead to the achievement of sustainable

competitive advantage. Sambamurthy et al. (2003) have emphasized that BPA is seen as beneficial for firms because it helps to achieve superior financial performance. Firms with a high level of BPA can act proactively in customer retention, responding to customer needs, improving operational flexibility, and in the end increase revenues while, at the same time, reducing costs (Tallon, 2008).

Furthermore, Prahalad and Hamel (1990) and Teece, Pisano, and Shuen (1997) state that BPA can contribute to the superior performance of a firm. Chen et al. (2014) propose that enhanced BPA provides an opportunity for organizations to create high revenues,

profitability, return on investment (ROI), and sales and market growth. On the basis of these arguments, the following hypothesis is developed:

(18)

2.6 The mediating role of Business Process Agility

As mentioned before, the strategic capability of BPA depends on the ability of a firm to carry out different IT resources (Bharadwaj, 2000). If a firm deploys BDAC in the right way, it could enable robust business processes. These processes could enhance the operational processes of the firm. As Wamba et al. (2017) established, BDAC positivelyaffects FPER. Previous research by Côrte-Real et al. (2017) showed that BDA applications could lead to a higher grade of organizational agility. Chen et al. (2014) have emphasized in their study that BPA functions as a mediating variable in the relationship between IT capability and

performance. BDAC can be seen as an IT capability, but IT capability is a broad subject which encompasses a lot of different aspects of IT systems and capabilities. Gupta and

George (2016) stress in their paper the difference between BDAC and IT capability. They see IT capability as the facilitator of day-to-day activities and a capability that includes

connectivity, information, and communication technologies. BDAC on the other hand, enables organizations to derive insights from data from different sources, and base their decisions upon these insights. Besides that, BDA professionals need other skills, abilities, and responsibilities than IT professionals (Gupta & George, 2016). Scholars already researched IT capability from the early 1990s onwards (Gupta & George, 2016), while studies on BDAC date back to the second decade of the 21st century, from 2010 and on. Big data has emerged as a hot and recent topic since organizations in all industries try to keep up with the rapid

emergence of big data. Therefore, it is interesting to see if the effects of a specific IT

capability like BDAC are different from the influences of IT capability in total and if big data can fill in the role of IT, which was seen as a competitive weapon in the 1980s (Gupta & George, 2016).

(19)

In the light of the previous findings on IT capability and BDA applications, it is conceivable that the deployment of BDAC could lead to the possession of BPA. The relationship between BDAC and FPER is investigated by other studies, with several moderators and mediators, but the possible role of BPA in this relationship is not yet examined. If a firm possesses a certain amount of BPA, this could create opportunities for firms to perform better regarding financial performance (f.i. higher profitability, return on investment, sales, and market share) and market performance. If a firm does not possess that certain amount of BPA, this could influence the performance negatively. IT applications are increasingly embedded in business processes. As a result of this, the extent to which a firm can quickly redesign and modify its processes seems to be massively dependent upon its ability to implement and use IT capability (Chen et al., 2014). BDAC is considered as a particular type of IT capability.

If the findings of these prior studies are summarized, it could be assumed that BPA has a mediating role in the relationship between the deployment of BDAC and FPER. Thus, the following hypothesis is developed:

(20)

The following conceptual model was developed, to give a visual representation of the variables and hypothesis:

(21)

3. Methodology

The next chapter of this thesis is about the research approach and the research design. It discusses the research design, sampling, and data collection. After that, the variables and measurements procedures are discussed. Finally, the statistical procedure will be discussed.

3.1 Research Design

The underlying philosophy of this research is positivism since it is the intention to work with observable and measurable variables, testing for causality. According to this measurement, a theory can be proposed. This theory is tested and refined until it is as close as possible to reality (Saunders & Lewis, 2012). The research is conducted following a

deductive approach. The basis is a grounded theory from prior literature, which is tested by the development and testing of hypotheses (Saunders, Lewis & Thornhill, 2009). The nature of this research in quantitative and the survey design will be online cross-sectional since this is most common in this field of research. According to Saunders and Lewis (2012), a cross-sectional design is highly suitable for the statistical testing of hypotheses. The limited amount of time plays a role as well since the timeframe for a master thesis is not sufficient for a longitudinal study. The data will be gathered by the distribution of a survey, aimed at Big Data Analytics managers and people who oversee Big Data Analytics projects at European firms. Since four students are sharing the same supervisor who study Big Data Analytics, the data collection will be shared by the four of them. This increases the chance of collecting as many respondents as possible. The dependent and independent variable (BDAC and FPER) and their constructs are identical in all four studies, but every student is studying a different variable of possible influence. In this particular case that is BPA. The scales of the different variables were all validated in prior literature.

(22)

3.2 Sampling and Data Collection

The sampling population is, as earlier mentioned, a selection of Big Data Analytics Managers and people who oversee Big Data Analytics projects at firms in European countries. The respondents were found through LinkedIn. This is an example of convenience sampling (Saunders, Lewis & Thornhill, 2009), a specific type of non-probability sampling. This type of sampling relies on the collection of data from a population which is conveniently available to engage in this study. On LinkedIn, the search query ‘Big Data Analytics Manager' was entered and different countries were filtered out. The countries with more than one thousand results were skipped since LinkedIn only displays the first one thousand results. By using the software RocketReach, email addresses from managers were retrieved from the results. The countries which were targeted are Belgium, Denmark, Finland, France, Germany, Ireland, Norway, The Netherlands, Spain, Sweden, and Switzerland. The total sample size was 1050 since the purchased RocketReach package only gives the option to retrieve a maximum of 1050 email addresses. The results were checked on companies which were on the list twice or more, to avoid the approach of the same company multiple times. The targeted Big Data Managers were contacted via email and requested to take part in this joint research project. In the next weeks, two reminders were sent to the managers that did not participate yet.

Eventually, this resulted in 163 responses, which corresponds to a response rate of 15,5%.

3.3 Measurement

The survey will contain questions about the size of the firm, the industry in which the firm operates, and the period the firm has been investing in Big Data Analytics. Therefore, the control variables are the size of the company, industry wherein the company is active, and previous BDAC experience. The other constructs and scales are adopted from existing literature. The survey contains a total of seventy questions. In the section below the variables

(23)

are discussed. A complete list of the used items and their references is included in Appendix A.

3.3.1 Independent variables

The construct of Big Data Analytics Capability is developed by Gupta and George (2016). It consists of 32 items which are divided into first-order constructs and second-order constructs. The first-order constructs are Data, Technology, Basic Resources, Technical Skills, Managerial Skills, Data-driven Culture, and Intensity of Organizational Learning. Basic Resources, Data, and Technology form the second-order construct Tangibles. Technical Skill and Managerial Skill form the second-order construct Human Skills. Intensity of

Organizational Learning and Data-driven Culture form the second-order construct Intangibles. The three second-order constructs together form the third-order construct of BDAC (see Appendix B). The entire construct has a Cronbach's Alpha of 0.87, and the article is cited 40 times. An example item is: "We have access to very large, unstructured, or fast-moving data for analysis". The construct is measured through a validated 7-point Likert scale (strongly disagree – strongly agree), measured at an interval level.

The construct for the other independent variable (BPA) is developed by Chen, Wang, Nevo, Jin, Wang, and Chow (2014) and consists of 8 items. The Cronbach's Alpha for this construct is 0.91, and the article has 136 citations. An example item is: "Respond to changes in aggregate consumer demand". The BPA construct is measured through the same 7-point Likert scale as BDAC.

(24)

George, 2016; Wu, Straub & Liang, 2015). Several studies on IS suggest that operating performance and market-based performance are considered to be two important aspects of FPER (Wang, Liang, Zhong, Xue & Xiao, 2012). Gupta and George (2016) classify these two dimensions as operational performance and market performance. As previous studies over the years have shown, when discussing FPER and IT capabilities, it is common to assess a firm's organizational capabilities while making inter-firm comparisons (Bharadwaj, 2000). Also, Wu, Straub, and Liang (2015) emphasized that FPER can be measured best relative to competition.

To measure the dependent variable (Firm Performance) a construct of Gupta and George (2016) will be used as well. This construct consists of 8 items about market performance and operational performance. The scale consists of a validated 7-point Likert scale (strongly disagree – strongly agree), measured at an interval level. The Cronbach’s Alpha is 0.89. An example item is: “We have entered new markets more quickly than our competitors”.

3.4 Statistical Procedure

After the collection of data, the raw data needs to be cleaned and prepared for analysis. This is done to lower the level of biases as much as possible. After that, the dataset was checked for errors and missing values, and incomplete answers were filtered out and removed from the dataset. Also, double companies were deleted. New variables were computed from the different items, to test the hypotheses. To check if all measurements are consistent, a reliability analysis was run. To test the hypotheses, variance and regression analysis was conducted. All these analyses are performed by using the statistical software IBM SPSS (version 24) and SmartPLS (version 3).

(25)

SmartPLS is a software tool to perform structural equation modeling (SEM). Kaplan (2000) gives the following definition of SEM: "a class of methodologies that seeks to

represent hypotheses about the means, variances and covariances of observed data regarding a smaller number of ‘structural' parameters defined by a hypothesized underlying model"(p. 1). SEM offers excellent potential for the development of theories and the validation of different constructs (Anderson & Gerbing, 1988). Besides that, it allows for the analysis of latent variables and possible relationships between them. It offers the possibility to analyze

constructs and its dependencies, without any measurement errors (Nachtigall, Kroehne, Funke & Steyer, 2003). The fact that it a suitable method for analyzing latent variables and their relationships, is a reason to use software for SEM in this case, since the BDAC construct of Gupta and George (2016) that will be used, is a latent variable with formative and reflective relationships. SmartPLS is a simple software tool to analyze models like this. The details of the analysis and the outcomes are presented in the ‘Findings’ section.

(26)

4. Findings

This section will present the findings of this study. First, the descriptive statistics of the variables will be elaborated on, which also includes the analysis of the dataset before the hypotheses were tested. After that, a regression analysis is run using the software SmartPLS, a software for structural equation modeling. Testing the hypotheses will provide insights into the significance of the relationship between the different variables.

4.1 Pre-analysis and Descriptive Statistics

First of all, the complete dataset of 163 responses was analyzed and adjusted for errors and mistakes, to prepare the dataset for the testing of the hypotheses. The responses were checked on three criteria; completeness, duration, and engagement. First, the responses which did not incorporate all three variables examined in this study were deleted. Since the shared nature of the data collection, not all items included in the questionnaire are of interest for this particular study. After that, the responses with a mentally and physically too short duration were deleted (cut-off point at four minutes). Lastly, the responses were checked for

unengaged participants. This was done by calculating the standard deviation of all answers. If the standard deviation of a particular case was very low, the response was deleted. After this analysis, there was a remainder of 62 responses. The scale means were calculated, and new variables were created to test the hypotheses. In table 1, the characteristics of the remaining sample are shown. These are also the answers to the questions about the control variables.

(27)

Table 1: Sample

Characteristics

Industry

Computer/Software 23%

Manufacturing 10%

Finance, Insurance, Real

Estate 15% Retail, Wholesale 3% Services 11% Healthcare 6% Others 32%

Previous BDAC Experience

Less than 3 years 29%

3-6 years 40%

More than 6 years 31%

Company Size Fewer than 100 31% Between 101 and 250 6% Between 251 and 500 8% Between 501 and 1000 5% Between 1001 and 2500 6% Between 2501 and 5000 2% More than 5000 42%

Next, the reliabilities of the scales were tested via the Cronbach’s Alpha coefficient. BDAC scores 0.93, BPA scores 0.87, and FPER scores 0.90. George and Mallery (2003) reported a rule of thumb on the Cronbach’s Alpha coefficient; “_ > .9 – Excellent, _ > .8 – Good, _ > .7 – Acceptable, _ > .6 – Questionable, _ > .5 – Poor, and _ < .5 – Unacceptable” (p. 231). Since all scales scored above 0.80, the constructs are considered to be reliable. Also, all items for every variable score a corrected item-total which is above 0.30. From this fact could be concluded that each item has a good correlation with the total score of the relevant scale. The items were also checked on the Cronbach’s Alpha if the item was deleted. The only item where deletion will lead to a higher Cronbach’s Alpha is BPA8, but the increase is, with

(28)

only 0.001, not substantially. Therefore, there was decided to keep using this item within the scale.

After that, the scale means were calculated, and new variables were computed to test the hypotheses. The hypotheses were tested through correlation and regression analyses. The direct relationships (H1, H2, and H3) and mediation (H4) were tested through linear

regression analyses. The variables industry, previous BDAC experience, and company size were used to control the analyses.

A descriptive statistics analysis was run on the residual 62 responses to research the distribution of the variables. Below, the mean and standard deviation are described. After that, skewness, kurtosis, and normality is tested, to check whether the dependent variable is

normally distributed. The results are displayed in table 2. As all correlations are below 0.70, all variables can be used to perform the regression analyses.

Table 2: Means, Standard Deviations,

Correlations

Variables M SD 1 2 3 4 5 6

1. BDAC Age 2.02 0.78 x

2. Firm Size 4.63 1.54 -0.08 x

3. Industry Type 3.73 2.82 0.06 0.09 x

4. Big Data Analytics

Capability 5.13 0.83 0.02 -0.18 -0.07 (0.93)

5. Business Process Agility 4.79 1.10 0.12 -0.07 -0.05 0.65** (0.87) 6. Firm Performance 4.58 1.05 -0.20 0.02 -0.13 0.49** 0.4** (0.90) **. Correlation is significant at the 0.01

level (2-tailed).

A quick look at table 2 shows that the companies in the sample score relatively high on all three variables; BDAC (M = 5.13, SD = 0.83), BPA (M = 4.79, SD = 1.10), and FPER (M = 4.58, SD = 1.05). When looking at the correlations, it stands out that there is a positive, significant correlation between BDAC and BPA (r = 0.65, p = 0.01), BDAC and FPER (r =

(29)

0.49, p = 0.01), and BPA and FPER (r = 0.40, p = 0.01). Therefore, it seems that BDAC has a positive influence on both BPA and FPER, and that BPA has a positive influence on FPER. The depended variable (FPER) was tested for normality. This test resulted in skewness of 0.24 and kurtosis of -0.26. According to Field (2009), a variable is normally distributed if the skewness and kurtosis interval is close to -1,0 – 1,0. In this case, FPER is normally distributed, since the values for skewness and kurtosis fall within this interval.

Since Gupta and George (2016) already have performed an exploratory factor analysis (EFA) on the BDAC construct, this study assumes dimensionality of the scale items of the BDAC construct. Therefore, a confirmatory factor analysis (CFA) was performed to test if the different items correspond to the right dimensions. The results of these CFA's can be found in Appendix C (table 3, 4, and 5).

There is not much consensus about the reliability threshold of factor loadings.

Guadagnoli and Velicer (1988) propose that a factor is reliable if it has at least four loadings of 0.6 or more, but hey do not take into account the sample size. MacCallum, Widaman, Zhang, and Hong (1999) do take into account the sample size. They advocate that a small sample size should have loadings of at least 0.60 or an average of 0.70. Two constructs have a loading below 0.60 (TS and DD). Average loadings are respectively 0.69 and 0.70. That means that TS is not reliable, but only by 0.01. Therefore, was decided to leave TS in and consider it as reliable. All other loadings are on average above 0.70.

(30)

4.2 Hypothesis Testing 4.2.1 Direct Effects

The following three hypotheses investigate a direct relationship:

H1: The deployment of BDAC will have a positive effect on FPER. H2: The deployment of BDAC will have a positive effect on BPA. H3: BPA will have a positive effect on FPER.

To test the direct effects of H1, H2, and H3, two linear regressions were performed. The first regression regards the independent variables BDAC and BPA and dependent variable FPER (H1 and H3), and the second involves the independent variable BDAC and dependent variable BPA (H2). The control variables previous BDAC experience, firm size, and industry were added to the model. Since industry type is a nominal measure, dummy variables were computed to check for significant influence of any of the different industry types.

First, the possible impact of the control variables was tested. Previous BDAC

experience (p = 0.08), firm size (p = 0.85), and the different types of industry (all p-values > 0.05) were all found to be not significant. Therefore could be concluded that these variables do not have a significant impact on the dependent variable FPER.

As can be seen in table 6, there is an insignificant effect of BPA on FPER (b = 0.21, t = 1.40, p = 0.17). Next to that, there is a significant effect of BDAC on FPER (b = 0.49, t = 2.39, p = 0.02). Therefore, there could be concluded that there is support for H1, but no support for H3.

(31)

Table 6: Regression Direct Effect on Firm Performance

Unstd. Coeff. Coeff. Std.

Model B Std. Error Beta t Sig.

1 (Constant) 5.18 0.69 7.52 0.00 BDAC Experience -0.37 0.21 -0.28 -1.80 0.08 Firm Size 0.02 0.09 0.03 0.19 0.85 Industry: Computer/Software 0.22 0.98 0.09 0.59 0.56 Industry: Manufacturing 0.60 0.52 0.17 1.16 0.25 Industry: Finance/Insurance -0.03 0.47 -0.01 -0.07 0.95 Industry: Retail/Wholesale 0.27 0.81 0.05 0.34 0.74 Industry: Services -0.37 0.48 -0.11 -0.77 0.44 Industry: Healthcare 0.03 0.60 0.00 0.05 0.96 2 (Constant) 1.49 1.04 1.43 0.16 BDAC Experience -0.37 0.18 -0.28 -2.08 0.04 Firm Size 0.06 0.08 0.08 0.68 0.50 Industry: Computer/Software 0.04 0.33 0.01 0.11 0.91 Industry: Manufacturing 0.65 0.45 0.19 1.47 0.15 Industry: Finance/Insurance 0.09 0.41 0.03 0.22 0.83 Industry: Retail/Wholesale 0.54 0.69 0.09 0.78 0.44 Industry: Services -0.05 0.43 -0.02 -0.13 0.90 Industry: Healthcare -0.25 0.52 -0.06 -0.49 0.63

Business Process Agility 0.21 0.15 0.22 1.40 0.17

Big Data Analytics Capability 0.49 0.21 0.38 2.39 0.02

a. Dependent Variable: Firm Performance

After testing H1 and H3, a second regression was performed. In this case, BPA is the dependent variable and BDAC is the independent variable. As can be seen in table 7, there is a significant effect of BDAC on BPA (b = 0.91, t = 6.37, p = 0.00). That means that there was found support for H2.

(32)

Table 7: Regression Direct Effect on Business Process Agility

Unstd. Coeff. Coeff. Std.

Model B Error Std. Beta t Sig.

1 (Constant) 4.65 0.72 6.44 0.00 BDAC Experience 0.12 0.22 0.08 0.53 0.60 Firm Size -0.02 0.10 -0.03 -0.22 0.83 Industry: Computer/Software 0.41 0.40 0.16 1.05 0.30 Industry: Manufacturing -0.27 0.54 -0.07 -0.50 0.62 Industry: Finance/Insurance -0.39 0.50 -0.13 -0.78 0.44 Industry: Retail/Wholesale -0.50 0.85 -0.08 -0.59 0.56 Industry: Services -0.09 0.51 -0.03 -0.18 0.86 Industry: Healthcare 0.23 0.63 0.05 0.36 0.72 2 (Constant) -0.40 0.96 -0.41 0.68 BDAC Experience 0.16 0.17 0.12 0.99 0.33 Firm Size 0.04 0.08 0.06 0.54 0.59 Industry: Computer/Software 0.23 0.30 0.09 0.77 0.45 Industry: Manufacturing -0.28 0.41 -0.08 -0.68 0.50 Industry: Finance/Insurance -0.31 0.38 -0.10 -0.84 0.41 Industry: Retail/Wholesale -0.20 0.64 -0.03 -0.31 0.76 Industry: Services 0.46 0.39 0.14 1.18 0.24 Industry: Healthcare -0.21 0.48 -0.05 -0.44 0.66

Big Data Analytics Capability 0.91 0.14 0.68 6.37 0.00

a. Dependent Variable: Business Process Agility

4.2.2 Mediation

The following hypothesis investigates an indirect effect through a mediator:

H4: BPA will mediate the relationship between the deployment of BDAC and FPER.

(33)

To test the mediation effect of BPA (H4) on the relationship between BDAC and FPER, a bootstrap analysis was performed using SEM software SmartPLS. Since BDAC is a multidimensional construct, SmartPLS is the best software to deal with this

multidimensionality.

The direct effects, as well as the indirect effects, were tested and the model was analyzed for mediation. Mediation is the case when four different conditions are met,

according to Baron and Kenny (1986). The first condition they mention is that there must be a significant relationship between the independent and dependent variable. As the second condition, they state that the mediator should relate significantly to the independent variable, and as the third condition it should relate significantly to the dependent variable. The fourth and last condition is that the primary effect between the independent and dependent variable should be weakened or made insignificant by adding the mediator.

Table 8a: Bootstrap Analysis Consequent Business Process Agility (M) Firm Performance (Y)

Antecedent Coeff. SE p Coeff. SE p

Big Data Analytics

Capability (X) a 0.79 0.04 <.001 c' 0.58 0.18 <.001 Business Process Agility (M) --- --- --- b 0.12 0.20 0.57 Constant i1 -0.17 0.90 0.85 i2 1.65 0.95 0.09 R2 = 0.62 R2 = 0.46 F(4,57) = 10.90, p<0.001 F(5.56) = 5.35, p<0.001

In table 8a, the results of the mediation analysis are shown. As seen at c’, the first condition is fulfilled, since there is a significant relationship between the independent and dependent variable. This was also concluded in table 7, where the direct effects were tested.

(34)

hence condition two is not met. Also, there was found an indirect effect (ab) of 0,17, as can be seen in table 8b. This means that if two firms which differ one unit in BPA, are expected to differ by 0.17 units in their FPER. The confidence interval of the bootstrap is -0.25 to 0.42, so the interval spans zero. This confirms that the indirect effect is not significant in this case. The indirect effect of BDAC on FPER through mediator BPA is also not significant (p = 0.60). This confirms the findings that there is no statistical evidence that BPA functions as a mechanism that drives the relationship between the deployment of BDAC and the performance of a firm.

Since not all four conditions of Baron and Kenny (1986) are met, and the indirect

effect of BDAC on FPER through mediator BPA is not significant, there is no support for H4.

Table 8b: Bootstrap Analysis

Effect SE p LLCI ULCI

Direct effect c' 0.58 0.18 <.001 0.33 1.01

Total effect c 0.68 0.06 <.001 0.63 0.85

Boot

SE Boot LLCI Boot ULCI

(35)

5. Discussion

In this section, the findings from the previous chapter will be discussed. Besides that, the research contributions, theoretical and practical implications, limitations, and possible directions for future research will be elaborated on.

5.1 Discussion

As stated before, Big Data and BDAC are booming, but companies still struggle with the implications and value creation of BDAC. The purpose of this study was to dive deeper into the relationship between BDAC, FPER, and BPA and find out how BPA can influence the relationship between BDAC and FPER.

However, the proposed conceptual model is not fully supported by the results of the study. The first hypothesis (H1) tested the role the possession of BDAC plays in the

achievement of high performance. H1 stated that the possession of BDAC would positively influence FPER. The testing of this hypothesis leads to statistical evidence that the possession of BDAC has a significant, positive influence on FPER. In addition, the results in the

correlation table between BDAC and FPER show a significant correlation (0.65) as well. Therefore, from these results can be concluded that the possession of BDAC enhances FPER.

Looking at the second hypothesis (H2), it tested the role of it states that the deployment of BDAC will have a positive effect on the possession of agility in business processes (BPA). There was found statistical evidence to support this claim. That means that, according to the results, BDAC can enhance BPA within a firm. The correlation table shows a significant (0.40) correlation between BDAC and BPA.

The third hypothesis (H3) tested the role of agile business processes in the realization of high performance. It states that the possession of BPA will lead to a higher FPER. There

(36)

effect were insignificant. However, the correlation table shows a positive, significant

correlation between BPA and FPER. Nevertheless, H3 was not supported by the outcomes of this study.

The fourth and last hypothesis (H4) sums up the complete conceptual model. It tested the role of BPA as a mediator in the relationship between the deployment of BDAC and FPER. As a logical consequence of the rejection of H3, the mediation hypothesis (H4) should also be rejected. There was found no statistical evidence that the agility of business processes strengthens or weakens the relationship between the deployment of BDAC and a firms’ achievement of high performance. The bootstrap analysis in SmartPLS gave no significant indirect effect, and the mediation effect was also insignificant. From this double check can be concluded that there is no support for H4, and therefore, no mediation of BPA in the

relationship between BDAC and FPER.

5.2 Theoretical Implications

As stated in section 5.1, there was found no significant evidence to fully support the conceptual model of this research. There was found no mediation of BPA, so BPA does not influence the relationship between BDAC and FPER, according to the results of this study. That could mean that BPA has not a significant impact on that relationship. Although not all four hypotheses were rejected entirely, there were found some direct effects.

Drawing upon prior literature about the topics of BDAC and FPER, the stated conceptual model assumed that the deployment of BDAC has a positive influence on the performance of a firm. That relationship is positively influenced by the agility of business processes, which acts as a mediator. Existing literature which was researched, clearly showed a positive relationship between BDAC and FPER Several authors claim that in almost all cases, BDAC has a positive influence on performance (Chen et al., 2012; Gupta & George,

(37)

2016; Wamba et al., 2017). This is in line with the results of this study, which found statistical evidence for a positive relationship between BDAC and FPER, and therefore support for H1.

The second hypothesis stated that the deployment of BDAC positively influences the agility of business processes. As the definition of BPA by Tallon (2008) states; it is about the extent to which organizations can simply and fast reassemble their business processes to adapt to the changes in their market environment. BPA is seen as a rare dynamic capability (Chen et al., 2014; Teece, 2007), which is vital for firms to keep a sustainable business in the rapidly changing business environment (Côrte-Real et al., 2017). Côrte-Real et al. (2017) also state that BDAC applications can support organizations in creating agile business processes. As Raschke (2010) states; IT is a platform for agility, in other words, a dynamic capability like BDAC can form a basis for BPA. The research on H2 comes to the same conclusion; it shows significant statistical evidence that the deployment of BDAC positively influences the

possession of BPA. The outcomes of the research support the theoretical basis for this hypothesis.

The third hypothesis proposed that the agility of business processes positively affect FPER. However, no significant statistical evidence was found for this claim. That means that, according to this study, more BPA does not necessarily lead to a better performance. That is not in line with prior literature, which proposed convincing evidence for the existence of this relationship. As Raschke (2010) states, BPA is regarded as an important mechanism which firms use to interact with their market environment, and which can explain variance in performance between firms over a more extended period. She also proposes that with the possession of BPA, business processes can be redesigned and new processes can be developed, to take advantage of uncertain market conditions (Raschke, 2010). As

(38)

Consequently, this can lead to a better performance. According to this existing literature, a direct positive effect from BPA on FPER would be plausible, but in the findings of this study, there was found no connection as previous scholars established.

The fourth hypothesis sums up the first three and proposed a role for BPA as a mediator in the relationship between the deployment of BDAC and FPER. Regarding the existing literature on the first three hypotheses, it seems logical that BPA affects this relation. Nonetheless, no statistical evidence for mediation was found in this study. So, there seem to be signals which indicate mediation, but the statistical results of this research do not confirm that. Therefore, H4 could not be supported.

In summary, the outcomes of this study show that BDAC positively influences both FPER and BPA. On the other hand, more BPA does not lead to a higher FPER, and there is no statistical evidence that it acts as a mediator in the relationship between BDAC and FPER. Therefore, the proposed conceptual model is only proved to be significant in parts of the relationship, but not in all. Accordingly, the answer to the research question is as follows:

There is a significant relationship found between Big Data Analytics Capability (BDAC), Business Process Agility (BPA), and Firm Performance (FPER), but BPA does not act as a mediating mechanism between BDAC and FPER, neither does it influence FPER significantly.

5.3 Contributions and Practical Implications

Although the conceptual model of this study could not be supported entirely, there still are some statements that could be confirmed by the results of this research. So far, the

literature on BPA as a mediator has only looked at IT capability and FPER. This study has tried to test this for a specific IT capability, namely BDAC. Unfortunately, there was found no

(39)

statistical evidence for BPA as a mediator. Nevertheless, this study highlights and confirms the relationship between BDAC and FPER. It contributes to the existing literature on this topic of research. The results that are significant are a confirmation of the direct positive effect of BDAC on the performance of a firm. Also, they confirm that the deployment of BDAC will lead to increased agility of business processes. So, the results contribute to a stronger theoretical basis for these relationships.

Data-oriented and data-driven organizations are emerging, so there are important implications for practice as well. Since BDAC enables a better FPER, BDAC should be an essential main topic in the determination of the strategy of organizations. As Akter et al. (2016) state, the development of BDAC can have practical implications for various industries. The improvement of BDAC can lead to an improvement in serving customer's needs, higher sales and revenues, the development of new products and services, and the expansion into new markets (Akter et al., 2016). Therefore, it is vital to address the development and

improvement of BDAC as an essential topic for managers. Also, it is crucial to not only focus on the technology and data collection, but also on other skills. Just collecting data with superior technology is not adding any business value. As Gupta and George (2016) state, it is also about the availability of managerial and technical skills, a good climate for organizational learning, and supportive organizational culture. A combination of these resources will

eventually provide a competitive advantage. So, it is essential for managers to focus on a combination of those skills and technologies.

5.4 Limitations

Like any other study, this research paper has its limitations. First of all, the retrieved data is not very extensive, and the respondents were targeted via a convenience sampling

(40)

be highly vulnerable to selection bias and other influences that do not lie within the control of the researcher(s). It also has a high level of sampling error. These two disadvantages give studies which use convenience sampling lower credibility than studies with other sampling methods. Convenience sampling could also lead to a self-selection bias. Eventually, that will lead to a sample that is unlikely to be representative for the full studied population.

Also, the sample size of 62 participants can be considered as (too) small. As Saunders, Lewis, and Thornhill (2009) state, a larger sample size reduces the likelihood of errors in generalizing the outcomes to the total target population. When choosing the actual sample size, four things have to be taken into account. The first thing is the confidence in the data, the second is the margin of error which is permissible, the third thing is the different types of analyses which are planned to run, and finally the size of the total target population. Keeping these things in mind, it is a matter of judgment as well as calculation to decide the final sample size (Saunders, Lewis & Thornhill, 2009). Because of the limited timeframe for this master thesis, and the problems with finding adequate respondents, eventually only 163 respondents took the time to (partially) fill in the questionnaire. After checking these responses, as earlier mentioned, on a few criteria, only 62 complete, usable responses

remained. This can be considered as too small when keeping the four criteria from Saunders, Lewis, and Thornhill (2009) in mind.

Another possible limitation could be the proposed BDAC framework, developed by Gupta and George (2016). The framework could not be applied universally in other cases, as they state in their article (Gupta & George, 2016). Since the understanding of BD and BDAC is still not very mature, it is difficult to present a list of resources that will lead to the

development of BDAC. The evolvement of BDAC will continue and change in the future, so it is possible that the framework cannot merely be replicated anymore.

(41)

The cross-sectional time horizon of the data could also be a limitation of this study. It limits the reliability of the findings because causal relationships between the different

constructs need further research from longitudinal or experimental studies. In that way, the question of causality could be answered in a justified way. Also, the use of an online questionnaire has always the possibilities of errors and biases. To reduce the chance of, for instance, social desirable answers, in the introduction letter is asked for honesty and not skipping questions which deemed difficult. This might have reduced the risk of errors and biases.

Furthermore, the possibility of self-report bias should be addressed. Self-reported studies are always prone to a few types of biases, which will affect the quality of the measurement (Dodd-McCue & Tartaglia, 2010). Examples are social desirability and the acquiescence bias. Social desirability occurs when the respondent tends to present his or her opinion according to the current cultural standards (Dodd-McCue & Tartaglia, 2010). An acquiescence bias comes from the tendency of respondents to give positive answers without regarding the content. It can be introduced by using a Likert scale (Chin, Johnson & Schwarz, 2008). Since this study is self-reported and using a Likert scale-based questionnaire, it is prone to biases like these. Therefore, the final results could be influenced by this.

Moreover, the respondents were targeted via LinkedIn. By using RocketReach email addresses linked to LinkedIn profiles were retrieved. Since RocketReach was not able to retrieve all email addresses, not all search results were approached to fill in the questionnaire. Also, LinkedIn is not yet validated by many researchers as a conventional method for data collection. Besides that, the targeting via LinkedIn also creates another limitation, which is the fact that LinkedIn only shows the first one thousand results from the search query. Since the sampling is not supposed to be random, countries with more than one thousand results

(42)

Lastly, although the control variable industry type did not significantly affect BDAC, it could be that new industry types in the future do influence BDAC. New data-driven companies as Airbnb, Foodora, Picnic, and Uber have disrupted the traditional industries. Companies like these are basing their decisions on the data they leverage. The BDAC framework could be too limited to gain insights into the way these new types of companies develop analytical capabilities.

5.5 Directions for Future Research

As earlier mentioned, this study was not able to corroborate the proposed conceptual model. Therefore, it would be interesting to do future research projects in this field of interest. By taking the limitations as mentioned above into account, different outcomes could be possible. This study confirmed the direct effects of BDAC on FPER and BPA, so more

research should be done to find the presence of a possible mediating or moderating role in one of those relationships since prior literature does point in that direction.

Also, large countries with over one thousand search results and countries outside of Europe were excluded from the sampling because of the imposed limit by LinkedIn. It could be interesting to look into larger countries within Europe, as well as outside Europe, to see if they differ significantly from smaller countries. The trend towards big data is visible

worldwide, so it is interesting to study if differences between countries affect the relationship between BDAC and FPER Another data collection method should be used then since

LinkedIn is not suitable for this. Besides that, it is an additional benefit that data from other, non-LinkedIn, sources could validate the possible results even further.

Moreover, the cross-sectional nature of this study limits the possibilities to find

causalities between the different variables. So, it could be valuable to reproduce the study, but as a longitudinal or experimental research. In that way, more evidence could be obtained to

(43)

test the hypotheses. Also, the reliability of the results would be improved if more respondents were used.

Another point of interest is the development of the adopted BDAC construct, developed by Gupta and George (2016). They propose that there could be other types of resources incorporated within the construct, to enhance the results about the capabilities of organizations to analyze their big data.

As earlier mentioned in the ‘Limitations' section, new company and industry types with a new, different view on data, are emerging rapidly. It could be interesting to look further into these new, data-driven companies and compare them with traditional companies, regarding the adoption and development of capabilities for BDA.

(44)

6. Conclusion

This study was developed and set out to find an answer to the proposed research question: What is the key mechanism that drives the relationship between Big Data Analytics

Capability (BDAC) and Firm Performance (FPER)? Therefore, an online survey was

distributed among Big Data Managers from European firms. This survey aimed to collect data on the relationship between BDAC and FPER and a possible mediation effect of BPA on this relationship and test the developed hypotheses.

Unfortunately, no significant direct effect of BPA on FPER was revealed by the analysis. As a result, BPA could not be established as the key mechanism that drives the relationship between BDAC and FPER. Nevertheless, other direct effects were found and confirmed by this study. A significant, direct effect of BDAC on FPER was found. So, if a firm deploys BDAC that has a direct, positive influence on the performance of that

organization. Also, a significant, direct effect of BDAC on BPA was found. That means that the deployment of BDAC by an organization positively influences the agility of its business processes. It provides the firm with the ability to adapt and change their business processes more efficiently.

Since prior research already established the relationship between BPA and FPER, it still is likely that this effect is present and that BPA can act as a mediator in the proven relationship between BDAC and FPER. Since this study could not contribute to this aspect of the conceptual model, it would be interesting to address this in further research, to try to find statistical evidence for this mechanism. The study does contribute to the growing attention and importance of BDAC. Since it shows that BDAC positively influences FPER and BPA, it highlights the relevance and importance of the development and deployment of BDAC for organizations.

(45)

The main conclusion of this study is that there has not been found any evidence for BPA as a key mechanism which drives the relationship between the deployment of BDAC and FPER. Although prior literature points in that direction, this study did not succeed in finding statistical evidence. Therefore, it could not draw conclusions, but it could point in directions of effects that, in all probability, do exist, but lack statistical evidence from this study.

(46)

7. References

Agarwal, R., & Dhar, V. (2014). Big Data, Data Science, and Analytics: The Opportunity and Challenge for IS Research. Information Systems Research, 25, 443-448.

Akter, S., & Wamba, S.F. (2016). Big data analytics in E-commerce: a systematic review and agenda for future research. Electronic Markets, 26(2), 173-194.

Akter, S., Wamba, S. F., Gunasekaran, A., Dubey, R., & Childe, S. J. (2016). How to improve firm performance using big data analytics capability and business strategy

alignment?. International Journal of Production Economics, 182, 113-131.

Anderson, J. C., & Gerbing, D. W. (1988). Structural equation modeling in practice: A review and recommended two-step approach. Psychological Bulletin, 103(3), 411.

Atyeh, A. J., Jaradat, M. I. R. M., & Arabeyyat, O. S. (2017). Big Data Analytics Evaluation, Selection and Adoption: A Developing Country Perspective. IJCSNS, 17(9), 159. Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable distinction in social

psychological research: Conceptual, strategic, and statistical considerations. Journal of

personality and social psychology, 51(6), 1173.

Beyer, M. A., & Laney, D. (2012). The importance of ‘big data’: a definition. Stamford, CT:

Gartner, 2014-2018.

Bharadwaj, A. S. (2000). A resource-based perspective on information technology capability and firm performance: an empirical investigation. MIS Quarterly, 169-196.

Brands, K. C. M. A. (2014). Big data and business intelligence for management accountants.

Strategic Finance, 95,64–65.

Chen, H., Chiang, R. H., & Storey, V. C. (2012). Business intelligence and analytics: from big data to big impact. MIS Quarterly, 1165-1188.

A study on the relationship between big data analytics capability, business process agility, and firm performance

Master Thesis