How can firms utilize Big Data and what is the impact on their business performance in the Netherlands and in Hungary

(1)

Master Thesis

How can firms utilize Big Data and what is the

impact on their business performance in the

Netherlands and in Hungary

Author: Tivadar Bartók Student Number: 11186542 Date of submission: 26.01.2017. Version: Final

Course: MSc Business Administration Track: International Management Institution: University of Amsterdam Supervisor: Dr. Markus Paukku

(2)

2

Statement of originality

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

I declare that the text and the work presented in this document is 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)

3

Abstract

The goal of this study is to answer the research question: “How can firms utilize Big Data and what is its impact on their business performance in the Netherlands and in Hungary?” The theoretical foundation was based on many different scholars, in regards of Big Data, business performance and the economic environments of different European countries. To widen the available range of International Management literature in the topic and to fill the research gap, this study observes the influence of different Big Data applications in two countries, which have not been compared in this context yet, in relation to their business performance. The research studied 116 firms from the Netherlands and Hungary via a web-questionnaire and the quantitative statistical analysis revealed some significant relationship between the Big Data usage of firms and their business performance. Furthermore, my research has concluded that the business performance of firms operating in Hungary is somewhat lower than that of Dutch companies, suggesting that Eastern European economic environment has a negative impact on business performance. However, some positive expectations regarding Big Data were rejected because the relation between these variables is not fully significant. These results indicate that targeted Big Data usage may be the most efficient tool for firms regarding their business performance.

Implications of these findings are discussed.

Key words: Big Data, business performance, European economic environment, decision

(6)

6

1. Introduction

In this day and age, technological advancement has reached a point where data is more available and usable than ever before, and firms of today have the opportunity to gain comparative advantage based on their capabilities of making use of it. This enormous amount of unstructured data is very complex and quite hard to utilize solely. This huge database is called Big Data, and to use it properly we need a data storage that is large enough, a management specialised for this purpose and precise processing of the data (Manyika, 2011). Others define Big Data as a pile of data, which cannot be processed, managed nor stored with current technology. (Gobble, 2013) Currently most data related to Big Data is not generated by humans but rather comes from the Internet of Things. Sensors and RFID (Radio Frequency Identification) deal with the collection of transition of data automatically (Gobble, 2013). In the following years, the aforementioned technological advancements will likely become part of our everyday life and will have a huge impact on the IT sector. For this change to be successful and safe, we have to make sure that privacy violation and data storage are guaranteed (Bollier et. al., 2010). Taking these options into consideration, Big Data usage provides an enormous range of possibilities for evolution in many industrial sectors and even for national governments. Authors reviewed by me (Young et al., 2001) fragmented the ways Big Data is used and observed the effects caused by these on the business performance of firms. They studied the management of Big Data utilization and also considered how managers’ previous experience with newer technology had an impact on the firms performance (Young et al., 2001). In my research, I used three aspects of Brynjolfsson and McElheran (2016) to analyze the relationship of Big Data and business performance. These aspects are the following: 1) who decides upon the type of data to be used in the firm, 2) how

(7)

7 big is their availability, 3) what amount of data is used to assist decision-making (Brynjolfsson & McElheran, 2016).

Measuring the business performance of firms is not easy, as there are many aspects to consider among the attributes. Author and scholar, Neely (2002) claimed that firm effectiveness implies to the measure to which stakeholder requisites are achieved, while efficiency is an extension of how economically the enterprise’s means are used in case of providing a given level of stakeholder satisfaction.

Furthermore, previous research also analyzed the effect different economic environments have on business performance of firms. Baldwyn and Wyplosz (2009) dealt with Eastern European economic environment and its difference from Western environment in their research.

Research Goal and Research Question

As of the things mentioned before, the impact of Big Data on business performance in different economic environments is a very interesting topic of research and the examination in the Netherlands and in Hungary do provide a research gap for this thesis. As the relation between the variables has already been studied by many scholars before, I only chose the ones that had the most significance to my research topic. The choice of the Netherlands and Hungary contributes to the originality of my research: I observed two countries that have never been compared and studied in this context before. The aim of this research is to contribute to the literature used by examining the relationship between the utilization of Big Data and business performance of firms in the Netherlands and in Hungary. Therefore, the research question is stated as following:

(8)

8 “How can firms utilize Big Data and what is the impact on their business performance in the Netherlands and in Hungary.”

More precisely, this study examines the impact of methods and routines of data usage which I have mentioned before, in correlation with Young et al. and Brynjolfsson and McElheran (Young et al., 2001), (Brynjolfsson & McElheran, 2016).

This research applies quantitative data collection via a web-questionnaire targeted at Dutch and Hungarian firms that have validated measures regarding Big Data usage and self-reported and perceived business performance.

This thesis is structured the following way: The next section is an extensive literature review about Big Data, decision making, business performance and European economic environment. Also, the current state of literature is discussed along with the conceptual model of the research. The third part presents the data collection and research method applied to investigate the research question. Following this, the result of the data analysis and hypothesis testing is shown. This would be followed by the discussing the results, where the research question is finally answered and the limitations of research are presented. Lastly, conclusion will be drawn from the prime findings and a proposal is presented for future researches on this topic to be concluded.

(9)

9

2. Literature review

In my literature review I analyze related material to my thesis, all of which explore research and effect of Big Data regarding business performance of firms. My research ought to reveal why, when and how firms can use Big Data, how it affects the firm itself and how much they know about the relation between this and business performance. I would also like to investigate the effects of Eastern- and Western-European economic environments on the business performance of firms in order to reveal the differences and make contribution to the International Management researches.

2.1. Big Data

Digital data is present everywhere nowadays in each sector, economy and organization. At the moment the potential privacy obtrusion creates scepticism across the globe about the use of Big Data, the potential benefits are eminent (Bollier et al., 2010).

It is not possible to adequately answer related questions about its relation to business performance of firms without a systematic approach to the concept of Big Data. In the last years the term Big Data was often used by scholars and researchers, and has created a large number of definitions and descriptions. Some authors define Big Data as huge amount of complex data, which almost always exceeds a multitude of terabytes, or sometimes even Exabyte's or Zettabytes, and is highly unstructured (Aiden and Michel, 2013). Other people define Big Data as an enormous quantity of data, which requires an advanced storage, management and analysis to be able to use it (Manyika et. al., 2011). According to MacAffee and Brynjolfsson (2012) the volume of data created each day in 2012 was 2.5 Exabyte per day and this number is expected to double in size every 40 days. Other scholars refer to Big Data as to datasets which size is beyond the capability of current software tools to capture, store,

(10)

10 manage and analyze it (Gobble, 2013). A few years ago, the description of Big Data was firstly determined by its size but more researches now agree that a more important parameter of what Big Data actually is, ought to be how smart it is, or the insights that the volume of data can provide. The fine grind nature of the data is important (George, Haas & Pentland, 2014). In the same research they also describe five key sources of high volume (big) data. These are (1) public data, (2) private data, (3) data exhaust, (4) community data, and (5) self-quantification data.

A few scholars defined Big Data by the velocity of it. What this means, is that there is not only a big sum of data created, but it also comes to us fast and also changes rapidly (Lohr, 2012). One of the best known definitions of Big Data comes from McAffee and Brynjolfsson (2012) they define Big Data, using the three "V's" - Volume, Variety and Velocity. Thus, for the goal of this research we describe Big Data as an extraordinarily huge quantity (Volume) of unstructured (Variety), rapidly moving (Velocity) data, which requires very specific advanced storage, management and analysis tools in order to use it for customer value proposition creation.

Davenport and Harris (2010) define analytics as the extensive use of data, statistical and quantitative analysis, explanatory and predictive models, and fact-based management to drive decisions and actions. Moreover, analytics are considered a part of Business Intelligence and this is specified as a set of technologies and processes that utilize data to comprehend and analyze business performance. For this research we refer to Big Data and BI as one attribute.

2.1.1. The use of Big Data

The utilization of Big Data has huge potential value for, not only for commercial enterprises but also for firms of other character. Firms make better business decisions based on data compared to firms who make decisions on experience and gut feeling (Brynjolfsson, 2012).

(11)

11 Firms are more innovative than their peers that do not use Big Data to differentiate in the market place (LaValle et al., 2011). The use of Big Data has a downside as well. According to Aiden and Michel (2013) Big Data will become more insightful and also predictive if the data sets used are more longitudinal. So the longer companies collect data over a longer period of time, the predictions, which companies would get from the data earned, will be more precise. The authors also mention that part of the job of working with Big Data is to know your data very closely to be able to reverse engineer its anomalies. The main reason firms must be aware of the data's oddities is because Big Data is very unstructured and so big, that you might see connections between data discoveries, which in the end could turn out to be unrelated (Aiden & Michel, 2013). Even the most impactful decision at firms used to be made based on experience or subjective perception. As we are able to store and manipulate data easily nowadays, decision can be made based upon real time data analysis. Decision-making based upon data has been proved to be more effective and successful as well. This is also proven in the research of Chen et al. (2008) where the most impactful result was that when data was used within firms, the improved performance is the accurate estimate of customer's future value, which in turn can be used to select these customers who promise the highest value in the future. Big Data analysts might also miss certain relations between data discovery, as at first sight the given relation may not seem obvious because of its unstructured nature. We must put the data in context to be able to properly use it. Large quantities of data without the context are useless (Crawford, 2011).

Author Kraska T. (2013) in his findings examines the relevance of Big Data and comes to the conclusion that it is ever increasing, so newer and more systems are created to “tackle” the challenge. On the other hand leading researchers claim there is nothing new or interesting about Big Data. The author finds that Big Data is forming the landscape of how we manage data and impacts other things beyond technology. I tried here to get to know how the research

(12)

12 community and how they perceive it. It is only a short preview two of the most famous types of systems, key value stores and Map Reduce systems. I find this article to be closely related to my thesis because I would like to find out how great the need is for firms to start using Big Data and helps analyze the effects of it. This article helps the reader understand what Big Data can be used for and what kind of fields can use it generally.

Volume is one aspect of Big Data but not the only one; other attributes are variety, velocity, value, and complexity. Data storage and transport are technology issues which can be temporarily solved but they pose newer problems which need to be solved in a new paradigm. These issues and obstacles are deeply analyzed while Kaisler, S., Armour, F., Espinosa, J. A., and Money, W. (2013) start a research program on Big Data analysis and design. Big data is just the beginning of the problem. Technological singularity is going to cause more available data in one year than all existing data ever before. (Kaisler et al., 2013) Our knowledge is widely distributed and widely accessible. The data will be spread out, it won’t be just one format and there will not be just one or a few cross- linkages among different data elements. It is an example of some of the issues they address.

Solving the issues and challenges mentioned in their findings will require an in-depth research effort, which are planned to evolve over the years. Authors start a collective research to begin examining Big Data issues and challenges to investigate what is the real use of Big Data. The authors found some of the biggest issues in Big Data storage, management, and processing which are all crucial elements of further utilization. They also identified these as some of the major obstacles that must be talked about within the next decade. Their future research will concentrate on developing a more complete understanding of the issues associated with Big Data, Big Data analysis and design methodology (Kaisler et al., 2013).

(13)

13

2.1.2 Decision making and Big Data

Big Data Analytic Systems make firms able to use organisational and customer data to improve decision-making and organisational processes (Watson & Wixom, 2007). Several case studies show the impact of Big Data used across the firm in regards to marketing, finance, firm and manufacturing (Davenport & Harris 2007). Research has also shown, that successful analytical companies have a data driven and analytical culture. These findings are consistent with scholars like Brynjolfsson who claims that data driven decisions lead to better firm performance (Brynjolfsson et al., 2011 and 2012). According to these scholars, firms that base their business decisions on data, have a higher output and productivity of 5-6% compared to the competition. These researches also show that data driven decisions have favourable impacts on different measures of performance, like asset utilization, return on equity and market value. The best performing organisations use analytics five times more than lower performers and research by MIT Sloan showed that analytics transmits value for organisations (LaValle et al., 2011). If a firm is being enabled by this information, in the competition, it tends to be a winner. These companies can use the information very efficiently, making the right decisions and products very quickly. Employees need to have proper decision rights for the firm’s strategy to be properly executed. Employees in strong organisations usually have a clear or good idea of the decisions and actions they are responsible for (Neilson et. al., 2008). This is true for firms who have implemented a Big Data or BI practice too. The successful execution of their Big Data or BI strategy is impossible without the proper decision rights and authority.

In the article of Provost F. et al. (2013) data science and its relationship to Big Data and data-drives decision making it is claimed there are good reasons why it has been hard to define data science. One reason is that data science is very closely connected to other important concepts which are growing fast, such as Big Data and data-driven decision making. Though

(14)

14 we can academically discuss this, for business, data science needs to be intertwined and understood with other relevant and important concepts, and we have to explore of fundamentals of data science. Once we do this, we can much better understand and explain exactly what data science has to offer. Also, only after we embrace should call it data science. In this article, we present a point of view that touches upon all these concepts. The authors finish with offering a partial list of fundamental principles underlying data science. Data-driven decision making can be improved drastically by data-Data-driven decision making, 3 Big Data technologies, 4 and data-science techniques based on Big Data.9,10 Data science supports data-driven decision making and frequently allows making decisions automatically at a grand scale. This depends on technologies for ‘‘Big Data’’ storage and engineering (Provost et al., 2013).

The article, “Critical questions for Big Data: Provocations for a cultural, technological, and scholarly phenomenon” by Boyd et al. (2012) is still investigating the possible uses of Big Data. Its core conception is about many groups claim that potential benefits and thus the cost of analyzing genetic sequences or social media interactions are emerging. It is a question whether Big Data research will help us develop better services and goods or create a new wave of invasive marketing. Will it help us understand collectives or will it be utilized to abuse civil right? Given the significance of Big Data as a socio-technical phenom, we argue that it is crucial to critically view its assumptions. In this article, the authors offer six thoughts to start conversations about the issues of Big Data: a cultural, technological, and scholarly phenomenon that is based upon the interplay of technology, analysis, and mythology. This article is questioning the use of Big Data or at least questioning the general help of it. This paper suggests to specify fields to use Big Data and decision making could be one of the most plausible choice (Boyd et al., 2012).

(15)

15 The next author is also starting with provocations for Big Data and I find it important to include it in my thesis I am writing about the essential use of Big Data by firms. Crawford K. (2011) discusses six cases about Big Data. In the current phases of Big Data it is needed to be examined as a phenomenon and methodological persuasion. The author finds a deep government and industrial drive to exploit maximal value from data, that can be information which will lead to targeted advertising and decision making (Crawford, 2011).

Author Tien J.M. claims that in order to profit from Big Data, one must accept risk and constant change; it must be the core of any entrepreneurship and in his article about Big Data saw as the unleashing information (Tien, 2013). By overcoming the problems around data quality and quantity, restriction of data access with on-demand cloud computing, analytics and applications, we can take action with the given information. Newer and newer additions are developed to make Big Data more modern and more effective in decision making. (Tien, 2013) The author is also talking about the future of Big Data and how firms can prepare themselves to change their strategic business tactics to rely their business decisions entirely on Big Data. This paper suggests that the future is all about Big Data because the new technologies are using so big amount of data that it is going to be Big Data very soon (Tien, 2013).

Some scholars suggest that challenges and opportunities with Big Data can be used by firms in the field of decision making. While Big Data is very promising there is no clear consensus on what it is exactly. The article of Labrindis et al. (2012) provides some further aspects for my paper. Firstly, to identify if or why Big Data is different from past huge data bases in relation with business performance and what are the most challenging aspects of Big Data. Secondly, to determine how can the data management industry solve Big Data challenges in

(16)

16 decision making. Finally, to consider the role of the data management community within the Big Data solutions sphere (Labrinidis et al., 2012).

In my thesis I am also looking for the most suitable areas of Big Data usage and the findings of these authors show that the social network and decision making related to it is one of the most suitable areas for Big Data utilization. This interesting venue is about the social networks because they are perhaps the largest Big Data producers, though this fact could help to solve Big Data processing challenges. Tan, W., Blake, M. B., Saleh, I., & Dustdar, S. (2013) in their article: Social-network-sourced Big Data analytics and decision making, the authors utilize personal random clouds consisting of random individuals from different social networks. As social science, mathematics, physics, and now computer science touch on this subject, social interactions among humans have been deeply explored. But since groups of people on the internet produce a large sum of data, that data showcases an interesting side of collective human intelligence. It can be summarized as the overlap of social networks for Big Data analysis and decision making. This is the intersection of Big Data research and social networks which is a vast pool of opportunities for scientists (Tan et al., 2013).

The research of Manyika et al. (2011) questioning whether companies are ready for Big Data or not, also discusses the field of decision making, therefore I find it relevant for the development of my research question. Brown, Chui and Manyika, (2011) consider radical customization, constant experimentation, and novel business models to be the new venues of competition as companies collect and work up huge volumes of data in order to support decision making. The paper studies how the important ways of Big Data may change competition by transforming processes, modifying corporate ecosystems, and facilitating innovation (Manyika et al., 2011).

(17)

17

2.2 Business performance

After absolving the literature of Big Data driven decision-making, this study aims to dwell deep into certain elements of business performance. Decision makers steer the company, but being a successful decision maker does not necessarily mean targets of the firms or the business figures are good. Success in firm is defined as creating a competent firm in the long term (Bouchiki, 1993) and is considered as a dynamic process. Although, Sarasvathy (2008) states that there must be a distinction between the success of the decision maker and the success of the firm, because the vast majority of studies link the success of the decision maker to the success of the firm. Simply, because success is a dynamic and abstract term, which is hard to measure; and because business performance is mostly the result of the decisions or actions of the decision maker (Sarasvathy, 2008). On the other hand, every decision maker has the goal to establish a sustainable business, which generates profit in order to further develop his or her vision. Studies that analyze firms with a trait approach, which means they are more focused on the personality aspects of the decision maker, use business performance as a metric to measure entrepreneurial success (Sarasvathy, 2004). This thesis also aims to analyze the connection between the decision making pattern, regarding to Big Data, of decision makers and the business performance of the underlying firm.

2.2.1 Firm performance

Cameron describes performance as multidimensional (Cameron, 1978). The two most significant attributes of performance are: effectiveness and efficiency. As theorized by Neely (2002), effectiveness implies to the measure to which stakeholder requisites are achieved, while efficiency is an extension of how economically the enterprise’s means are used in case of providing a certain level of stakeholder satisfaction. The definition of Moullin (2003) specifies an organization’s performance as how properly the firm is lead and the value the

(18)

18 firm provides for customers and various stakeholders. Neely states, that to achieve higher performance, the firm has to obtain its expected target with superior efficiency and effectiveness compared to its competitors (Neely, 1998). Despite the fact that several researches have already been executed on the theme of performance measurement, further research is still required. Neely defines performance measurement as the method of quantifying the efficiency and effectiveness of previous operations by procurement, coordination, classification, analysis, interpretation and dissemination of corresponding data. (Neely, 1998) Firms have to be first-rate in cost, quality, access, consumer choice for customer demands to be met while controlling the costs. (Shortell, 2000) Moullin (2003) claims the performance of firms is complex therefore multiple sides must be approached. In order to reach the requirements of measurement process the firm has to be examined by both monetary and non-monetary methods. The most usual monetary measurement items to determine the performance are the sales growth, revenue growth, growth in the number of employees, net profit margin, product/service innovation, process innovation, adoption of new technology, product/service quality, product/service variety (Wiklund and Shepherd, 2003). My questionnaire mostly contains monetary measurement items, because the results of Big Data usage are measured most efficiently with this method.

2.3 Big Data and business performance

Authors Young et al. (2001) in the findings of their paper report data from a study that combined two theoretical perspectives—top manager and network/institutional—to examine the factors influencing organizations and their business performance to adopt innovative management practices. Study results indicate that both top manager and network/institutional factors are important determinants of whether organizations adopt innovations. On the other hand, as predicted, the importance of these two sets of factors appears to change as an

(19)

19 innovation becomes more widely diffused (Young et al., 2001). Now knowing the results of the sources, I claim that if the decision maker had already met a given technological solution before, that can have a positive impact on the whole firm, so my first hypothesis is:

H1: Previous experience with Big Data has positive impact on business performance.

Furthermore the authors Brynjolfsson and McElheran (2016) studied in their research what relation does companies performance have with generic and specific data usage.

The authors provide a systematic empirical study of the dispersion and adoption patterns of data-driven decision making. Using huge amount of data collected by the Census Bureau for a large representative sample of manufacturing plants, they find that data-driven decision making rates nearly tripled between 2005 and 2010. This fast diffusion, along with results from a companion paper, are consistent with case-based results that data-driven decision making tends to be productivity-enhancing. Of course certain plants are significantly more likely to adopt than others. The most important correlates of adoption are size, presence of potential complements such as information technology and educated workers, and firm learning (Brynjolfsson & McElheran, 2016).

One of the main research points of the authors is to state, if in any given firm which employer in which department decides upon what type of data is used during the decision making process (Brynjolfsson & McElheran, 2016). Analysing the results of their research, I got to the conclusion that if the type of Big Data used is not decided on headquarter, but on local department levels, it has a positive effect on their business performance. The second hypothesis of my thesis is thus the following:

(20)

20 In addition, Brnyjolfsson et al. (2016) also analyzed what effect the availability of data has on firms, in other words if the need data is accessible in the right moment. Based on their results, I have found that if a bigger database is available related to Big Data, then it might have a positive effect on the operations of the firm, so my third hypothesis is this:

H3: Availability of Big Data has positive impact on business performance.

My next hypothesis is also based on the findings of these authors as they published their results regarding what effect database availability has on their decision-making process. The authors have found a positive relationship between the variables, thus my fourth hypothesis states:

H4: Amount of Big Data used at decision making has positive impact on business performance.

2.4 Western- and Eastern-European economic environment

Essentially, this study aims to identify the differences in the patterns of business performance between Dutch and Hungarian firms related to the differences in their Big Data usage. The motives behind this investigation consist of several aspects. First of all, I am a Hungarian native, currently studying in the University of Amsterdam in the Netherlands. This factor makes me familiar with the cultural and economic differences between the two countries, which aroused my interest in the realized differences between the business practices of the two areas. Secondly, this unique case has allowed me to get to know a number of firms both in Hungary and in the Netherlands, which was quite instructive about the differences of attitudes, priorities and the perception of standard in business approach. It makes me motivated to investigate the relation of Big Data usage and business performance. Additionally, in spite of the different aspects of the two countries, they are both part of the

(21)

21 European community with entangled historic past, which fact may create similar attitudes towards very fundamental aspects, like society or business.

After the millennium, many Central and Eastern European countries joined the EU, thus making the composition of the whole union somewhat different. This means that not only the newer members changed, but the EU as a whole. Among other factors, such as cultural, institutional and historical differences, economic distance was an obstacle that had to be tackled. The ‘transitional economies’ were and still are pretty far from the average living standard in the West and entering the EU single market does not guarantee an easier future. Different theorists make contradicting predictions about the final outcomes from these scenarios, in which countries with weaker economies engage in free trade with countries with relatively stronger ones. (Baldwin and Wyplosz 2009) In addition, sources I have analysed conclude that Central and Eastern European economic environments have a negative impact on firms based there, compared to those in Western Europe (Baldwin and Wyplosz 2009) .

These results conclude this thesis to hypothesize the following: H5: Eastern European economic environment has negative impact on business performance.

The listed hypotheses are structured in the conceptual model below:

(22)

22

3. Methodology

The next chapter discusses the research approach and the research design of this thesis. First of all, the research instrument of the questionnaire is discussed, then the data collection procedure and the targeted sample is presented. After this, the variables and measurements are discussed, followed by the part about the statistical method.

3.1 Research design

Positivism is the philosophy of the research conducted for this thesis, because it intends to work with observable and measurable variables in the principle of causality. A theory can be proposed according to this, which can be tested and refined until it accurately predicts reality (Saunders and Lewis, 2012). Furthermore, deductive approach is applied in this research, as its basis is a grounded theory in the business literature, and the goal is to test hypotheses derived from it (Saunders et al., 2009). A questionnaire is used as the research technique, and it is aimed at cross-sectional analysis of the relationship of Big Data usage of different firms and their business performance. The limited timeframe available for this paper allows collecting and analyzing data at only one period in time (cross-sectional) (Saunders & Lewis, 2012).

The scales used in the questionnaire were all validated in certain given literature. The original scales are in English, which were translated to Dutch, as approximately half of the participants are form the Netherlands. The Hungarian translation was not conducted as the questions I used in my survey were straight-forward enough to be easily understood in English too. The survey was created together with a fellow Dutch student, who made the Dutch translation. Furthermore, a pre-test of the survey was done with the help of other fellow students to receive feed-back and perfect the final version.

(23)

23 The questionnaire was distributed online and applied multi-channel approaches like company websites and e-mails. To maximize response rates and increase visual appeal to potential respondents, the electronic survey of Qualtrics was used (Qualtrics, 2016).

To collect as many responses as possible, a joint survey with a fellow student was created. The dependant variable and the unit of analysis of both constructs are identical, so a common survey made it possible to unite our network and reach as many entrepreneurs as possible. The survey is divided up into four blocks: The first part is connected to generic information about the firms, the second details the top manager of the firm (excluded from this research), the third is about data usage and the final block is about business performance. Naturally the survey became longer than average and it took approx. 10-15 minutes to complete on average. This resulted in a higher than average drop-out rate though the large target group, which was made possible as the joint-survey compensated for it.

3.2 Sample and Data Collection

Because the research targeted firms that are utilizing Big Data, which is quite challenging to look for, it was not possible to achieve a fully completed sampling frame. To create a representative sample of the population, non-probable self-selective sampling (Saunders & Lewis, 2012) was used and targeted at 610 firms. The research was conducted at the University of Amsterdam in the Netherlands by a Hungarian student, so Dutch and Hungarian firms were the most approachable; therefore the majority of the sample is based on them. After the web-questionnaire was distributed to the possible respondents, they could decide whether they want to take part in the research or not (self-selective sampling). This made it sure that respondents were not forced to take part and anonymity was also guaranteed to all firms. The joint survey made collaboration possible with another student from a different network and background, which ensured a larger diversity of respondents, by covering more

(24)

24 markets and industries. The bigger and more diverse the target group in terms of country, industry and markets, the more likely differences will become apparent and comparable (Dew et al., 2009).

The survey was distributed online to the firms, so there was a possibility that a person without relevant knowledge of the business would respond. As this paper aims to analyze the relationship of Big Data and the business performance of any given firm, it was not expendable to have answers from the most competent person at the company. Many control measures were implemented to reduce the possibility of bias here. The letter sent to the firms, the official cover letter and even the introduction section of the survey contains a statement: forward the questionnaire to a relevant person if the current recipient is not the most competent person. This ensured to filter irrelevant respondents even if they were reluctant to read the instructions and started to fill it out at first glance.

3.2.1 Sample Characteristics

The research targeted 610 firms and received 116 complete responses, which resulted in a 19,02% response rate. The reason for the relatively low participation rate is because of the pretty low benevolence of the respondents to participate in the study.

The survey targeted firms from different countries, so the homogeneity of nationalities is the following: 46% of the samples were Dutch and 54% of them were Hungarian.

3.3 Variables and measures

This research is studying the relationship between Big Data usage and business performance. Several studies are here as a theoretical basis, as they were empirically tested and further developed by different scholars. All applied measures and scales were adopted from original

(25)

25 scales of authors from peer-reviewed journals in the field of International Management. The variables are discussed below and the actually used items are listed in

3.3.1 Independent variables

The literature review section gives a detailed explanation about the different types of Big Data utilization of firms that this research handles. The first independent variable, based on Young’s source, aims to investigate whether the manager has previous experience with using Big Data or not. This variable shows up in the questionnaire as a yes or no option (Young et al., 2001).

Further three independent variables, based on the writings of Brynjolfsson and McElheran, refer to the following: who decides upon which type of Big Data to use, how great is the availability of Big Data and what amount of Big Data is used by the relevant personnel when a decision is made. These questions can be answered with four or five different options in the questionnaire (Brynjolfsson & McElheran, 2016).

The last independent variable studies, what effect does Eastern and Western European economic environment has on the entrepreneurship? In this case, the people who filled out the questionnaire gave the country where their firms operate (Baldwin & Wyplosz, 2009).

3.3.2 Dependent variables

The dependent variable of the conceptual model is the business performance of firms. It is assumed by the research model, that there is a connection between how firms use information gained from Big Data and how successful their businesses are. International Management literature generally assesses business performance by measuring employee and sales growth, also, the change in profitability of firms in the first three years of operating (Sambasivan et al., 2009; Blackburn et al., 2013; Haber & Reichel, 2007; Wiklund & Shepherd, 2005; Rauch

(26)

26 et al., 2009). The three metrics are measured with seven possible interval-categories, which were designed for each measure to offer the appropriate possibilities that could cover the various and wide-ranging performance levels of their firms. All these measures are validated and approved by peer reviewed journals: employee-growth is based on Terpstra and Olson (1993), Bruno and Tyebjee (1985) and Kazanjian (1988); the clusters of sales growth is based on Bruno and Tyebjee (1985) and Kazanjian (1988) and the categories of profitability is based on Huggins and Johnston (2009).

3.4 Statistical procedure

Firstly, the raw data of the research needs to be prepared and cleaned to keep the level of bias at the lowest possibility. The dataset was thoroughly checked for errors and the incomplete answers were filtered out and excluded from the research list-wise. To test the hypotheses, new variables had to be created from the certain items. A reliability analysis was conducted on these measures to make sure that all of them are consistent. In addition, obliquity and kurtosis of the data had to be checked in order to test normality. Regression and Variance Analysis was carried out to test the underlying hypotheses. The statistical software, SPSS (v.22) was used for the analysis. All details of the analysis and the presentation of the results are detailed later in the ‘Results’ section.

(27)

27

4. Results

This section is presenting the results of data analysis, which is testing the hypotheses of the research. The results are structured in three steps, first the method of generating the variables is shown which is followed by the relations between the variables. Finally, a regression is presented where all the variables are tested in one model.

4.1 Method of configuring variables

In order to use business performance number I made another variable from the four questions regarding to business performance and this variable became an index. Concerning the questions in the questionnaire, a sequence of answers was created, where a higher number indicates better performance. The result is an index, more specifically a latent variable, which can be used to define a correlation structure between numerous variables with the help of a few latent variables, so called factors. These factors have no physical meaning, they cannot be observed directly, cannot be measured and their existence is only imaginable based on the original variables. Both seven-term and five-term answers were defined symmetrical to zero, so seven-terms were placed between minus 3 and 3 and five-terms were placed between minus 2 and 2.

Weighted average was used to calculate the importance between the questions. To achieve this, in the generated indicator the following measures were used: 40% for the profitability of the company, 30% for the change in sales volume, 20% for the change in the number of employees and 10% for opportunity and decision making were presented in the generated index.

(28)

28 Anova’s criterion is that the continuous variable is normally distributed, that is why a normality test should be run on the business performance indicator. The Shapiro-Wilk test accepts the null hypothesis of normality, so the distribution is normal as the P value is high (Sig. column). (Table 1.) After all this, the Anova tables can be created safely.

Table 1: Business Performance Normality Test I test the normality of business performance with skewness and curtosis as well

Skewness is a positive value, so negative business performance occurs more often than positive one, thus the model leans rightwards.

Kurtosis shows the frequency of occurrence of extreme values. In my case, this number is negative, which means extreme values are rather rare. (Table 2.)

Table 2: Skewness and Curtosis

4.2 Relationships between variables

4.2.1 Anova

Hypothesis 1: Previous experience with Big Data has positive impact on business

performance.

Variable Skewness Kurtosis

(29)

29 Anova chart was used, which includes a continuous and a category variable. This method deconstructs the continuous variable by the categories. Here, I calculated whether it had previous experience or not with Big Data and that disrupts the business performance index, which is continuous for these two categories. This calculates the average in both categories, and checks whether there is any significant difference between the two averages. If there is a difference, than that is a good descriptive variable which indicates a significant relation between the two variables. (Table 3.)

The result is: not significant.

Table 3: Anova P.Exp. Bp.

Four variables were considered for business performance but actually, only the BP14 weighted is used. In the last column of the table the P value is 64% and I am working on a 5% of significance level so the value is bigger than 5%. Accordingly the null hypothesis is accepted of independence so there is no connection between the two variables. This test does not claim anything about the direction of the connection but when it is graphed I can see the connection is positive.

Therefore I can say in firms, where decision makers had previous experience with Big Data, the business performance is also bigger but the change is so little that it cannot be significant. Statistically I can state the connection of variables is not significant. (Table 4.)

(30)

30 Hypothesis 2: Local choice about the type of Big Data has positive impact on business

performance.

Table 4: Anova Type.BD. BP.

At this hypothesis I used the same Anova method and the null hypothesis is accepted again so I can say there is no significant relation between the two variables. It means that it does not matter if the headquarters or the local department is the place where the decision makers do their choices about what kind of Big Data should be used in the life of the firm in the relation of business performance.

Hypothesis 3: Availability of Big Data has positive impact on business performance.

(31)

31 Anova table is used again to reveal the relationship between the variables in this section as well. The calculated number is 5,3% which is really close to the 5% border but under this condition I have to state there is no connection between the variables. On the other hand on the graph it can b seen clearly as the availability of Big Data is increasing the business performance is getting better too. If the choice would be made on 6% I could say the relationship between the variables is significant so this number was really close to the approval if this hypothesis. Therefore in further investigations this hypothesis should be tested again because this result could generate in this way only accidentally. Furthermore, after the figure is complete, the direction of the relation of variables is moving according to our expectations, as the availability of Big Data is growing, the business performance index of the firm is increasing as well, although I still cannot claim the connection is significant. (Table 5.)

Hypothesis 4: Amount of Big Data used at decision making has positive impact on business

performance.

Table 6: Anova Use. BP.

The Anova test shows, the connection between these variables are significant, because the P value is 0. It means the more a firm uses Big Data during the decision-making process, the faster is the growth of its business performance. (Table 6.)

(32)

32 In this situation the strength of association is also should be measured because the connection is significant. This indicator is going to be between 0 and 1 and it shows the strength of relationship between two variables. In this situation it is 43% which indicates a moderate strength between the variables.

Hypothesis 5: Eastern European economic environment has negative impact on business

performance.

Table 7: Anova Ned. BP.

In this Anova test the indicator is 1 if the variable is Dutch and 0 if it is Hungarian.

The connection between the variables is significant but not so strong as in the previous hypothesis. According to my expectations the average of business performance is much higher in The Netherlands than in Hungary.

The strength of association is 34% so the relationship between the variables is moderate. (Table 7.)

4.2.2 Cross Tables

At the cross tables, I analyse what relation do independent variables have. In this case, there is no relation if the significance indicator in the Chi-square row Asymp. sig column is greater than 5%.

(33)

33 This test only works properly if every field in the cross table has an expected frequency is bigger than five, in any other case, it cannot be trusted.

If I find a relation based on the conditions mentioned before, I analyse the symmetric measure table, as the strength of association of the relation is indicated there. I have found a relation in four cases between the variables, but because of the condition of expected frequency I had to exclude two of them: the relation between the type of Big Data and Use, and the relation between Availability and Use.

Between availability and NED (Netherlands origin firm), I have found a relation, as they show a 0.2% strength of association in the proper field and the condition was thus met. In this case now, the cramer strength of association was tested too, which showed a 35% value, which means that the relation is weak but significant. (Table 8.)

(34)

34 Furthermore, there is a 1,4% valued relation between Use and NED, so like in the case before, I test the Cramer indicator here too. In this case, the strength of association is only 29%, but that also means the relation is weak but significant. (Table 9.)

Table 9: Cross Tab. Use. Ned.

These were the revealing indicators showing what is related to what. As there is no relation between most variables, and even if there is, it is weak, multicolinearity will likely not be a problem in the regression model.

4.3 Regression

Now I will test what would happen if I put all variables in one regression. Business performance will be the dependent variable and the other five dummy variables will be there for me to explain the dependent variable. Basically, this method allows me to test all hypotheses at once

(35)

35 With the dummy variables, we have to choose a category of reference to which we want to compare. With the previous experience for example, the reference category will show that he or she does not have previous experience. At the type of Big Data, we exclude number one, and we can compare numbers two, three and four to number one. At availability the same method is applied, we exclude number one as it was not checked in the questionnaire, so we choose number two as the reference category and put in three, four and five in the model. In the next one it is almost the same as no one chose number one, we exclude number two and put all others in the model. In the last one, whether it is Dutch or not is also chosen as a dummy variable.

We could use Enter methodology to run a regression, which includes all variables in the model, but we have to filter out what is not significant, so this method is not good for this case.

That is why I use stepwise methodology, which filters out not significant variables with iteration method, so that the least significant variable is excluded and restarts the model. After that, if any non significant variable remains, it restarts the model again, and does this as many times as there are only significant variables in the model

Although this method does not take into consideration if there are cognate values like in the case of dummies, where it is recommended to include non significant variables too, for later interpretation. If the possible outcomes of a variable may contain even one which is significant, I leave the rest there too for an easier comparison to be made.

With the stepwise method, I concluded that it is significant whether it is Dutch or Hungarian and whether firms use Big Data in their decision-making process or not. These are essentially categories number four and five, but I included number three for easier comparison.

(36)

36 With this model, many possible analyses could be made, including testing the betas. Testing the betas shows marginal effect, so if the independent variable increases by a given amount, what given amount does the dependent variable increases. Dummy variables are a different case. The indicator does not increase by given amount in this case, but what effect do dummy variables have on the dependent variable compared to the reference category. So for instance, if a „NED” dummy variable, which shows if the firm is Dutch or not, has a beta of 0.389, that means that the business performance index of that Dutch firms is 0,389 points higher than of a Hungarian one, not taking into consideration any other circumstances. (Table 10.)

Table 10: Regression Coefficient

A given firm that only relies on Big Data during its decision-making process, shows a 0,983 points higher business performance index than a firm that does not use Big Data, or uses it less. The more a firms uses it, the bigger the numbers are in column B, so the difference is greater. This means, that the more they use this technique, the greater performance they achieve. (Table 10.)

Betas give the answer to what happens if the independent variables increase by one unit, more precisely these are marginal variables. All variables are dummies here, so we have to always compare them to something.

(37)

37 In H4, where a firm uses solely Big Data during its decision-making process, the business performance index doubles, compared to only partly relying on Big Data.

It is necessary to check the adequacy of the linkage of the model, to which an Anova table is connected, which is essentially a global F test, that checks if the model has any significant variables. This shows that the model is relevant, as p value is close to zero, so there is at least one significant independent variable in the model.

R square can be found in the Regression model summary table, which shows descriptive ability. This value is 23,4% which means it describes the variance of the dependent and independent variables by 23,4%. (Table 11.)

Table 11: Regression Model Summary

The model can be measured partially with the stepwise method, which analyses variables one by one. This way each variable is significant, as the p value related to it is very small, excluding one, the USE3, which I put back for easier comprehension.

There is an important assumption in the linear regression, which says that independent variables are not related to each other, as only this way is it possible to comprehend ceteris paribus properly. So we can only move a variable safely if it is independent from the others. At the end of the table, if the VIF indicator is close to one, then there is no multicolinearity, and such is the case in my model.

(38)

38 The direction of relations can be specified in the regression, and the directions are in the model as expected. In USE, the higher category we are in, the bigger the business performance is and the relation is positive between the Dutch firm and business performance. There are many variables in the model at once.

This is the normality of the random component of the regression, as the random component has to follow a normal distribution and the figure shows that the rectangles are mainly below the bell curve.

Regression Anova test: this is the test of the whole model, its visible in the last column, as the P value is almost exactly zero. So the model has a descriptive ability because the variable is significant. (Table 12.)

Table 12: Regression Anova

I do not conduct a Cronbach alpha test during my statistical analysis, as the variables I am testing during the processing of my database, are not continuous. It is possible to conduct a test like this related to business performance index, but in this case it is meaningless as of the questions asked. In the case of further variables, there are only peer reviewed questions related to the hypotheses, so the variables are not continuous.

The case with the correlation matrix table is the same as with the Cronbach alpha method, namely it is only worth to conduct if I worked with continuous variables.

(39)

39

5. Discussion

In this part of my thesis I would like to answer my research question: “How can firms utilize Big Data and what is the impact on their business performance in the Netherlands and in Hungary.” In addition, the most significant results of the research are reflected to already existing practices and theories. Furthermore, I present limitations for the study, which is necessary to be able to assess the implications.

5.1 Theory

In the research, firms of two countries were investigated from the point of view of Big Data usage to observe its affect on business performance. Namely, how does business performance change if the management has experience using Big Data, if the departments can have the type of Big Data they want, if all data is accessible and they use it accordingly and how all of this impacts overall business performance in Hungary and the Netherlands. Drawing upon the existing literature, the conceptual model assumed the greater usage of Big Data has a generally positive effect on the business performance of firms, but the Eastern European environment has a negative impact on it. In the researched literature, the authors claim that in almost all cases, Big Data has a useful aspect, but opinions are divided on how this usefulness can be utilized in different sectors. In the article of Boyd on the critical questions for Big Data, he asserts some criticism of that, as many people only use it to follow the current trends, without even knowing exactly what to use it for (Boyd et al., 2012). After testing out my hypothesis I can assert my claim that the authors criticising Big Data are somewhat right, as not in all cases did Big Data usage result in a positive outcome, in other words, with a growth in business performance. Despite this, I have to say that targeted Big Data usage had a

(40)

40 positive impact on business performance, the higher utilisation, the better. Thus it is very important to assert the fact, that firms must specify in detail the goals that they acquired with Big Data.

Young et. al. authors had already discussed previous impacts and experience with data usage, and with the given sources I was expecting that managers who already used Big Data achieved a better business performance with their firms (Young et al, 2001). In spite of this, my research concluded that this was not significant at all. I accept the zero hypothesis of the independence, so there is no relation between the two factors. The test does not indicate anything about the direction of the relation, but if I depict it and the sum of the average is constantly growing, it is clearly visible that the relation is positive, although the test itself does not state this. So those who had previous experience have a greater business performance but the change is so little that it is not significant. Statistically, this does not impact anything. Regarding other aspects of data usage I use the sources of authors Brynjolfsson and McElheran (2016) as their publications are more than relevant for my thesis. A systematic empirical study of the adoption and diffusion of patterns of data driven decision making is provided. According to their research, the aforementioned method tends to be productivity-enhancing. Although, some factors could be more influential than others. Key correlates of adoption are, respectively: amount of usage of data, firm learning availability, the presence of potential complements and size (Brynjolfsson & McElheran, 2016). Analysing the database, many question are related to this topic, such as: who choose the type of used Big Data, what is the level of availability of Big Data and the amount of usage of Big Data in decision making. According to the authors, we can expect a positive impact on business performance in all three cases, thus I have marked a positive relation among the mentioned variables in my hypothesis.

(41)

41 After the Anova tests, only the fourth and fifth hypothesis turned out to be correct. The fourth claims that a higher amount of Big Data used to support decision making has positive impact on business performance. The relations between the variables are very significant as they have a value of 0 P, which shows a dominant improvement and increasing value. The strength of association is at 43% which implies a moderate strong relation. This literally means the more they use Big Data in their decision making the faster business performance grow. (Table 6.)

The second and third hypothesis related to these authors did not turn out to be correct however. The second hypothesis was debunked, as it claimed that if the managers of the local department were to decide what type of Big Data to use, that would make a positive impact on business performance. Since there is no relation between the two variables at all, it does not matter who decides the type of used Big Data. According to authors Brynjolfsson and McElheran (2016) these variables should have had a positive relation with each other but in my findings there was no connection similar to the previous ones.

The third hypothesis was debunked too but interestingly, the index number was very close to the 5% limit as we got a value of 5,3%. It is clearly visible on the picture that as the availability of Big Data grows so does business performance. (Table 5.)

These findings might be worth analysing with a different or bigger sample, as it is possible that in these cases the result is accidental. So we can assert that my results reflect what was stated in the literary sources although in certain cases we cannot claim for sure that there is a relation between the variables.

The fifth hypothesis was about the effect of Eastern-European economic environment and according to the literature a negative impact was expected regarding to the business performance. The connection between the variables turned out to be significant, but not so strong as at the third hypothesis. This means I accepted this assumption and the authors

(42)

42 findings were right about the following thought, in the Western-European economic environment a firm can perform better. (Baldwin and Wyplosz 2009)

All in all, the results of the conducted study show that the higher amount of Big Data used in decision making and the Western-European economic environment enhances the business performance of firms, but the other three independent variables have no significant relationship with it. Therefore, the conceptual model proved to be significant only in some aspects, hence the research question is answered as following:

There is some significant relationship found between the Big Data usage of firms and their

business performance.

5.2. Discussion of further analysis

Additional statistical analyses were carried out on the dataset in order to unveil any relationship between Big Data usage of firms and their business performance.

I did a regression test in order to test all of my hypotheses at once. After the test, variables related to hypotheses four and five remained significant and what that means, is whether the firms is Dutch or Hungarian, and what amount of Big Data they use during their decision-making process. Furthermore, we can analyse marginal effects with testing betas. The “ned” (Dutch origin of firms) dummy variable, which concludes if the firm is Dutch or not, has a beta of 0,389 and that means that the business performance index of a Dutch firm is 0.389 points higher than of a Hungarian firm, not taking into consideration any other circumstance (ceteris paribus). This can be a real business life lesson to take into consideration to firms that are planning to go abroad to make FDI because the economical distance between the mentioned countries is relevant. (Table 10.)

(43)

43 Besides this, in the case of a firm who solely relies on Big Data during its decision making process, the indicator number of BP index is higher by 0,983 than of those who use Big Data less frequently during their decision making. The more the firm uses Big Data the bigger the numbers are in column B, so, the more the firm uses this technique the bigger its performance is. This finding suggests firms to use as many Big Data as they can to support decision making processes but as I have already mentioned the application of this technique should be precisely targeted in order to gain significant competitive advantage. (Table 10.)

I analyzed whether the model is relevant or not with a global F test and the result was that the model is relevant as the p value is close to zero which means that there is at least one significant variable in it. If the model was not relevant I could not use my results properly during my research. We can determine the direction of the relations in the regression and all directions went as expected. In USE (use of Big Data) the higher category we are in, the higher the business performance is, also, the relation between the Dutch firm and its business performance is positive. This finding supports previous assumptions according to authors Baldwyn and Wyplosz (2009), namely, firms in the Western European economic regions perform better.

Besides this I found it necessary to perform a Skewness and Kurtosis test in order to comply a full scale statistical analysis. With this test I checked the normality of Business Performance. Skewness takes up a positive value which means that negative business performance occurs more often than positive performance, so the model leans rightwards. Kurtosis shows how often do extreme values occur in the model. In my case, this number is negative, which means extreme values occur very rarely. (Table 2.)

Furthermore I have analysed cross tables, which measured the relationship between independent variables. I have found four cases of connection between the variables but

How can firms utilize Big Data and what is the impact on their business performance in the Netherlands and in Hungary

Master Thesis