The adoption of data-driven decision making in the Netherlands : organizational, network and managerial determinants of DDD adoption

(1)

The Adoption of Data-Driven Decision

Making in the Netherlands

Organizational, network and managerial determinants of DDD adoption

Tijn H.P Gerards January 27, 2017

10202528 Thesis: Final

(2)

Statement of Originality

This document is written by Tijn H.P. Gerards 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)

Acknowledgements

This thesis would never have been possible without the support of student organization Integrand Nederland who did a major contribution by providing access to their customer relations database. Furthermore, I would like to thank my supervisor for his support and critical feedback.

(4)

Abstract

Data and in particular ‘big data’ is starting to play a more important role in all industries. For the U.S. market it has been empirically proven that implementation of decision making based on ever more broadly and widely available data from within and outside of an organization is beneficial to its performance. Signals are in place that suggest differences among countries in the use of management practices and the adoption of DDD. Therefore, this study examines the adoption process of data-driven decision making (DDD) in the Netherlands and formulates policy recommendations for both business and official purposes. The model in this paper is based on a broad base of literature and adapted to the Dutch context. Support is found for a positive influence of top manager experience, organizational learning and structured management on DDD adoption within Dutch businesses. Also an organization’s structure complexity (multi-unit and internationality) is found to be of significance to DDD adoption. These findings are of relevance to business executives, owners and other policymakers in terms of their informative function and ability to be directly translated into an organization’s strategy and/or policy with regard to emphasized use of data within their decision making processes.

(7)

1. Introduction

To what extent are Dutch companies using data to its full potential? This research will create insight into specific company characteristics that drive the adoption of data related innovations within an organization’s management. In particular, the research will focus on big data and its applications within Dutch organizations. To be able to perform research on big data and its applications within Dutch organizations it is needed to identify themes that will prove to be of value. The required context will emerge from a focus on data-driven decision making (DDD) as an innovative management practice with an important technological core, stemming from advances in the field of ‘big data’. This research will be performed to allow for an answer to the question of what the determinants of the adoption of DDD are and, to discover what the current state of DDD adoption in the Netherlands is.

Exploratory research provided interesting insights in practical applications for big data. Especially, within management and decision-making big data is playing a more and more important role. The roles data could play in the management of organizations seem to be innumerous. As already introduced, central to this thesis will be data-driven decision making (DDD). A type of decision making based on data gathered from within and outside of an organization (Brynjolfsson & McElheran, 2016). In the last few years more and more attention has been given to this concept. However, complete consensus on the definition of what this exactly entails and how the term is made ‘research proof’ does not yet exist. The term could be interpreted as making use of a yearly balance sheet and simple profit expectations to make investment and other managerial decisions. This is not the way forward. Information sources as the two mentioned before are about ‘looking back’, which nowadays is just not enough anymore. A shift is noticeable from a retrospective to a perspective that involves both internal and external data. In this thesis the term DDD will be used in such a

(8)

way that a firm has adopted the practice of DDD at the moment that it ‘emphasizes decision making based on data and business analytics’ (specific criteria will follow in the next chapter) (Brynjolfsson et al., 2011). To emphasize: attention will be given to the shift from intuitive decision-making to decision-making based on data and predictive analytics. This should not be interpreted as solely being ‘rocket science’, but as building a decision making framework that is based on data, which can be gathered from several sources. These sources could among others be an enterprise resource planner (ERP), CRM software or the production process (which is common when a company has implemented other management practices such as Total Quality Management: TQM) (Young et al., 2001). A few examples of applications of DDD will follow. Currently, many organizations are changing their HR policy. Prospective employees need to make online assessments and on the basis of certain parameters a computer decides whether the results are desirable in such a way that an interview is proposed. If the results are not desirable the applicant will be refused. In this example a clear shift can be seen from an ‘all-human’ decision to building assurance walls to make a better, more rational decision on a basis of data. It must however be noted that the criteria for decision making will still be made by humans. Another example is combining data from different departments to support decision making. For example, between an organization’s sales and production departments. The sales department’s data sensors have noticed a sudden rise in sales in a certain period and relates this to weather conditions. With such information at hand a decision about the production quantity can be made by a system that is driven by data collected from internal sales numbers and external variables. So, we see DDD can be operationalized in a broad variety of situations. From strategy and HR to production and sales. This research will not emphasize one application at a certain department over the other, but it will focus on the bigger picture and decision making within organizations in general.

(9)

1.2 Research Gap

Careful consideration has gone into the way in which the theory on this research topic was going to be presented. I have discovered, and many others with me, that big data is quite an abstract topic. In order to get more of a feeling on what big data is and especially, how to present it in a way that is more concrete and attractive for all, a search started for a practical application of the possibilities big data has to offer. An important condition is being able to relate the abstract topic to more familiar research areas in business that could benefit from the understanding of such a relationship. Support for the decision within this consideration was found both in scholarly literature but as well when visiting the ‘Small Big Data Conference 2016’ at the Amsterdam ArenA. Attending were representatives from a broad variety of Dutch organizations. From municipalities to start-ups to corporates. The attendees were fuzzy about the hype that is currently generated about the use of data within their organization, however, the real concrete and beneficial application still seemed to be a question. This lack of clarity inspired to wanting to get to know more about the possible applications within management and the suitability for different organizational types. Also, it is important to find out if this lack of clarity if only the case in the context of Dutch business or if other countries are experiencing the same problems.

In the U.S. research has been performed on the adoption of DDD within its manufacturing industry (Brynjolfsson & McElheran, 2016). This research delivered interesting insights in the DDD adoption process. A research performed by Agarwal et al. (2014), shows that the types of used management practices do differ among countries. For that reason, research on the U.S. market can hardly be taken for true in the Dutch context as well. Therefore, interest is piqued on the situation in the Netherlands.

(10)

1.3 Relevance

To this day many researchers have focused on innovations, either managerial or technological, and much attention has been given to diffusion of particular practices across firms and over time and space (Mol & Birkinshaw, 2009). However, much less attention has been given to research that emphasized the firm as the level of analysis. Remarkable, since it is important to understand the causes and consequences of adoption and implementation of new practices. What characteristics make an organization receptive and capable of implementing an innovative management practice? This study examines several organizational and individual factors that influence a firm’s likelihood of DDD adoption. Knowledge about these factors should be recognized as being valuable. Such information is likely to be of interest to managers, investors and policy-makers also in terms of creating a more durable competitive advantage (Agarwal et al., 2014). Furthermore, if we observe differences between the U.S. and the Netherlands, for the same stakeholders it would be of interest to get to know where these differences originate from, to being able to act accordingly.

1.4 Structure

The objective of this thesis is twofold. Firstly, it will create insight in the current state of DDD adoption in the Netherlands is. Secondly, it will present what important determinants for the adoption of DDD are within Dutch organizations. Finally, it will formulate recommendations for managers and other business executives, owners and policy-makers. The conclusions and recommendations will come of good use when initiating a process of adopting enhanced and better decision making through data-driven decision making within their organizations. Establishing these answers and recommendations will happen according to the following structure.

A contextual framework will be built for understanding all that is at play when talking data-driven decision making (DDD). Because the desired outcome is a deep understanding of

(11)

the determinants of DDD adoption, the developed framework will be comprehensive and highly contextual. Research on innovative management practices will be discussed jointly with developments in the field of big data. Developments in these fields are seen as the main concepts relevant to DDD adoption. The main concept through which DDD will be approached is on theories on management innovations and their adoption. No clear distinction will be made between either reasons that an organization decides to adopt DDD or determinants that show the likelihood an organization has adopted DDD. However, both will be approached from a perspective of new institutionalist theories, the resource-based view and the knowledge-based view. Because of a clear technological origin of DDD, the management innovations theories will be further complemented by information about big data and technological innovations. The theories and relevant factors that follow from these sections will result in the formation of multiple frameworks through which this research tries to explain the determinants of DDD adoption. To verify empirically the hypotheses that have follow from the literature review, a national survey has been executed. The responses to this survey provided data with which a quantitative analysis will be performed. Finally, the results of the quantitative analysis will be discussed in order to be able to answer the research question and add knowledge to the scientific research field of DDD.

2. Literature review

The structure of this literature review is created by deconstructing the concept of data-driven decision making (DDD) into several parts. DDD is a management innovation within decision making but has to do with technological advancements as well. Since to many organizations the use of DDD as a practice is new and this research is concerned with its adoption, DDD as a practice is framed as a management innovation with a technological core. Support for the use of this taxonomy and logic is found in E.M. Rogers’ seminal work on the diffusion of

(12)

innovations. According to him: “any idea, practice or object that is perceived as new by an individual or unit of adoption could be considered as an innovation available for study” (Rogers, 1963/2003).

In essence, the main point of this thesis is to understand what the best and essential circumstances are for an organization to be able to adopt such an innovation. Since organizations are not fixed constructs and change and improve themselves within a dimension of time these circumstances are looked at from a time perspective that relates to progress and change as well. Because of both, we need to contextualize even more on what an innovation in this research entails. The context in which this research looks at innovations is two-sided. It looks at an innovation as a new practice organizations can adopt. But, it looks at the circumstances that such an innovation needs to be adopted within a certain context as well. DDD is a practice and is perceived as ‘new’ by many organizations because they are not yet applying it within their organization. More support for DDD being an innovation can be found in Schumpeterian theories on creative destruction. In that sense DDD is an innovative practice because it has the capacity of replacing another, more traditional type of decision making (not based on data and predictive analytics) (McAfee & Brynjolfsson, 2012; Schumpeter, 1942). Furthermore, it should be contextualized on what basis a practice like DDD will be adopted within organizations. The reasons for adoption will be viewed at from arguments from the resource- (RBV) and knowledge-based view (KBV) that support adoption as the result of seeing true benefits to an organization’s performance (Barney, 1991; Hitt et al., 2001). Also of major importance is approaching DDD adoption through a lens of new institutionalism and factors derived from neo-institutional arguments (Powell, 2007).

To resume, we have to cope with an innovative practice that should mainly be explained through the concept of management innovations, but has a strong relation with technology. Firstly, the implications and context with regard to management innovations will

(13)

be deliberated with specific attention to factors that influence the adoption of such innovations. Complementarily, technological innovations will be discussed. Secondly, specificities on research regarding DDD will be introduced. Finally, the theories and factors introduced in the management, technological and DDD part lead to three frameworks through which the adoption of DDD according to literature should approached. These constructs function as a base to the formulation of hypotheses that relate to DDD adoption that will serve as a blueprint for building a contribution to knowledge on the current state and determinants of DDD adoption in businesses throughout the Netherlands.

2.1 Management Innovation

Research that focuses on innovative management practices is important for several reasons. Firstly, because of the effect management innovation has on the performance of organizations. Some authors even describe management innovation as “the most important and sustainable source of competitive advantage for firms because of its context specific nature among others” (Birkinshaw & Mol, 2009). However, in reality it is not that clear what is meant with ‘management innovation’. In order to be able to perform a clear research without too many ambiguities the following definition about these two terms will be used. “A management innovation is both an idea about what might work and the implementation or introduction of that idea within an organization” (Birkinshaw & Mol, 2009).1 This definition is used by respectable scholars and also formally used by the Department of Trade and Industry of the United Kingdom. From here we extract the definition of a management practice as the result of a management innovation after it has been implemented and used for a specific purpose. The practice is used by a manager in order to make decisions and get things done effectively through people (Luke, 2011). A management practice can take several forms and can influence many processes within organization. Among them: human resources,

1_{Which Birkinshaw & Mol again derived from the classic book Diffusion of Innovations by E.M. Rogers}

(14)

strategy, R&D and so on. In this research an even broader perspective is applied. An analysis will be performed on DDD which has the capacity to influence all these processes due to its nature and potentially broad applicability. Some scholars have proven that the use of this type of decision making is significantly beneficial to a firm’s performance (Brynjolfsson & McElheran, 2015; 2016). Since the relevance and definition of management innovations now have been established, in the following section a background about the research field and its theoretical foundations will be explained.

2.1.1 Reasons for adopting an innovative management practice

The scientific field on innovative management practices consists of three areas of research, of which the last one will be central in this thesis. The first area focuses on specific management practices such as the M-form or Total Quality Management (TQM) and investigates diffusion patterns across organizations, industries and countries. The second area focuses on the market for management practices. Especially, understanding the why and how of the popularity of particular practices (Birkinshaw & Mol, 2009). The third area of which this thesis will be part, is the area that examines a broad range of internal and external firm characteristics that influence the likelihood of the adoption of a(n) (innovative) management practice. It needs to be emphasized that the idea of taking the firm as unit of analysis is much needed in the current scientific discourse. Without an understanding of why and under what conditions an organization is able to and receptive to adopt an innovative management practice, the real relationship between management innovation, its adoption and firm performance cannot be understood (Birkinshaw & Mol, 2009; Daniel, Myers & Dixon, 2012). This real understanding relies heavily on the comprehension of the specific management innovation that is of interest in this research.

The adoption of an innovative management practice by organizations can be seen as the result of several forces and circumstances. The relationship between resource dependencies and firm objectives will be discussed. In this part both resource-based view and

(15)

knowledge-based view arguments are important contributors (Hitt et al., 2001). Furthermore, the relationship between an organization and its institutional framework will be discussed in order to be aware of potential institutional factors that influence the adoption of innovative management practices. Also, more attention will be given to establish the relationship between management practices and performance effects.

With regard to a resource- and knowledge-based view, arguments that will be central in this research are related to firms that possess large amounts of resources and knowledge because of their context or specific characteristics are likely to successfully introduce (more) new practices (Barney, 1991; Hitt et al., 2001; Birkinshaw & Mol, 2009).

The theories that support the major influence an organization’s institutional framework has on an organization are usually captured under the name of new institutionalism.2 These theories suggest that organizational practices and structures generally are reflections or responses caused by institutions embedded in an organization’s environment (Powell, 2007). The mechanism that leads to responses created by the institutional framework has been thoroughly researched. In their seminal work DiMaggio & Powell (1983) reassess the iron cage proposition of classical theorist Max Weber. Max Weber’s iron cage theorizes the irremediable effect of rationalization and bureaucratization on society. DiMaggio & Powell (2001) contest his theory that drivers of bureaucratization have to do with capitalist ideals, by arguing that bureaucratization is driven by institutional isomorphism. They argue this because bureaucratization has been achieved in principally all organizational forms, but organization are still changing and becoming more homogenous while not necessarily becoming more efficient (DiMaggio & Powell, 1983; Scott, 2001). DiMaggio & Powell (1983) and Scott (2001) identify several isomorphic pressures influenced by an organization’s institutional environment. These are coercive, mimetic and

(16)

normative pressures. Coercive pressures have to do with the organizational search and need for legitimacy. One factor has to do with obedience to regulation. Another factor is when an organization chooses to use the same processes and practices to be able to show that they are doing it in the right way, which will result in external and internal legitimacy. Mimetic pressures emerge as a response to uncertainty. For example, when an organization faces a problem mimetic pressures cause an organization to copy the approach of another organization that does not seem to experience the same problem. According to DiMaggio & Powell (1983) this is often done without necessarily having a positive effect on an organization. Normative pressures are the result of other deeply rooted institutional factors and often are seen as pressures for professionalization. In a particular institutional framework, it is accepted that a certain practice or process is the norm and becomes institutionalized. This causes an even greater impact because it could result in becoming embedded in an education system (or through academic authorities). Another aspect of this type of pressure is the result of relations between professional networks that cause an acceleration of practice diffusion (DiMaggio & Powell, 1983; Powell, 2007). These pressures capture the essence of the possible ways an institutional framework (through a new institutionalist lens) affects the adoption of innovative management practices.

Many studies have found that management in general, not even in a specific form, has a substantial positive effect on firm performance. It exerts a great effect on productivity, it increases decentralization of decision-making and increases the use of efficient sources of communication and analysis (Bloom, Eifert, Mahajan, McKenzie, & Roberts, 2011; Bloom, et al., 2013; Atalay, Hortacsu, & Syverson, 2013; Yang, Kueng, & Hong, 2015). However, aside of neo-institutional factors, in the field of management innovations, as in almost all other research fields related to business, a lot of research on the relationship between a specific phenomenon or movement and firm performance has been done. It is also not unique

(17)

that views expressed in articles on these relationships differ greatly. The main mission of a commercial organization revolves around the creation of added value and thereby generating profit. By keeping up with management innovations companies contribute to enhancing their profits. Innovation of management has been proven to have a positive effect on productivity and efficiency (Birkinshaw & Mol, 2009; Walker & Chen & Aravind, 2015). Walker et al. (2015) have created an all-encompassing integrative article in which both positive and negative results are taken under consideration. Their final conclusion confirms the significant positive effect of management innovations on firm performance.

At this moment it has been made clear under which theories and assumptions management innovations should be approached. The same type of foundation should be established for the DDD specific technological core to management innovations. This will be discussed in the following section.

2.1.2 Technological Innovations

Many scholars have found empirical evidence for classic Schumpeterian theories that suggest a positive influence of technological innovations on both macro and micro levels, in other words, economic growth in general but firm performance as well (Schumpeter, 1942; Fagerberg, 2005). Technological innovations can be a broad variety of things. It can be ‘a tool, a technique, physical equipment, or system by which employees, the units, or the organization extend their capabilities’ (Damanpour, 1987).

This paragraph on DDD being a technological innovation, is essential to the understanding of DDD as a management practice. The technological innovation that has allowed for the development of DDD as a management practice and instrument for decision making within organizations is the emergence of ‘big data’. The innovation of ‘big data’ and all the possibilities that are associated with this phenomenon are becoming more and more urgent for businesses in order to be able to keep up with their competition. What this phenomenon exactly entails, how we define what big data is, remains a problem. It is

(18)

remarkable to see that many authors that have published about big data, mysteriously have tried to avoid having to formulate a workable definition (Brynjolfsson & McElheran, 2015; 2016; Brynjolfsson et al., 2011; Kaisler et al., 2013; George et al., 2014). At least, no consensus has yet been found and the definitions that were used in papers usually have been customized to a specific research not to the big data phenomenon. In the following paragraph some ideas about a possible definition of big data that were found in literature will be discussed in order of an attempt to do find a workable definition.

To avoid the lack of a clearly defined definition about which consensus exists being a restraint, the definition as formulated by Villars, Olofson & Eastwood (2011) will be used in this thesis because of its broad formulation and clear applicability to DDD: “big data technologies describe a new generation of technologies and architectures designed to economically extract value from very large volumes of a wide variety of data, by enabling high-velocity capture, discovery, and/or analysis”. attention will be given to describing as precise as possible the known characteristics of big data. An important characteristic of big data is the large volume of data that is used (or at least attempted) to capture value for all kinds of organizations. While big data at first was solely an enormous and continuously growing amount of data with which it was tried to analyze patterns, now it is also used as a means for executing predictive analytics (George et al., 2014). Attributes of (big) data like variety, velocity, value and complexity should be taken into account as well (Kaisler et al., 2013; Villars et al., 2011). These applications are valuable for a broad variety of organizations and processes that take place within these organizations. The outcomes that spring from these applications can contribute to efficient production, quality control, efficient procurement, strategy execution and strategy formulation. These are all kinds of organizational processes that influence or are influenced by managerial decision making. Therefore, it is interesting to focus on the influence big data can have on decision making

(19)

within organizations. That is, when we stand back from these ‘micro’ processes like production, procurement or even strategy execution, to get to know more about the bigger picture. The bigger picture then is about in which way all potential data within, as well as from the outside the firm can contribute to better management through enhanced decision making. To, as a firm, be able to realize these potentials, internal and external characteristics of an organization are important. Just as they are for management innovations in general. Following the context and technological background towards DDD that has now been established, the consequences of these technological innovations will be explained. This will be done in the next section and moreover, the few and for this research very important articles that have placed DDD as the central topic within their research will be introduced. 2.2 Data-Driven Decision Making

Traditionally, decisions within organizations were made according to an unwritten rule called HiPPO (Highest Paid Person’s Opinion) (McAfee & Brynjolfsson, 2012). This person earned the privilege of making decisions or delegating decision making authority to persons lower in the hierarchy. Even though this person had earned this privilege and made decisions based on experience and intuition, there was no full guarantee or completely solid rational foundation on which decisions were made. More and more technologies are being developed and introduced that aid the gathering, structuring and interpreting of data more thoroughly than was possible before. These developments create a sense of urgency for businesses to engage in using data in their decision making processes (Economist Intelligence Unit, 2012). The same research, commissioned by Capgemini, showed the large potential within organizations to try to capture extra value by improving their decision making and make us of the data at hand (Economist Intelligence Unit, 2012).

Brynjolfsson could be named one of the founders of the academic research field of active applications in the decision making process of ‘big data’ within US business. The relevance of doing this research originates from the emergence of a real data revolution in the

(20)

last couple of years. More and more companies have started to use data analytics to make strategic decisions (Brynjolfsson et al., 2011). Brynjolfsson finds that the use of DDD, in general, results in a higher firm performance. This conclusion is a confirmation of the inferences that could be made from the theory on management and technological innovation in the paragraphs above. In their sample of 179 large publicly traded US companies that adopted DDD have an output and productivity that is 5-6% higher than expected given other influencing factors such as IT usage and other investments. Since we now know that the adoption of DDD in general causes a significant increase in output and productivity, the questions that remain are: Why do not all companies adopt such practices? Which companies do adopt DDD? Which companies benefit from adopting DDD? Brynjolfsson & McElheran (2015) performed a very technical research on the US manufacturing market to create a better understanding of what makes companies adopt DDD practices. Not only for scholars this information is valuable; for CEOs and other decision makers within firms this information could be very useful in pursuing increased efficiency and higher performance and thereby taking their business to the next level. To be able to use the construct of DDD it is important to concretize its definition. In the introduction of this thesis a ‘definition’ of DDD was introduced as when an organization ‘emphasizes decision making based on data and business analytics’. However, what does that entail? To give substance and draw a line between organizations that did and did not adopt DDD, Brynjolfsson & McElheran have defined the following criteria (2016). Survey questions were asked about the use of key-performance indicators in an organization, the use and availability of data to support decision making and about the use of short-term, long-term or short- and long-term targets within their organization.3 If the answer to the question about use and availability was in one of the top

(21)

two categories4, and if the answer to the question about targets was answered with both short- and long-term targets and the organization makes use of 10 or more KPIs, then the organization is seen as one that has adopted DDD.5

The main concepts that are of relevance to this research have now been introduced and substantiated. In the following sections several frameworks will be introduced that directly follow from two types of arguments that were introduced throughout the previous sections. The first type of argument is related to the theories that were introduced through which management innovations are viewed. The second type of argument is related to contextually relevant factors that have to do with technological innovation or DDD specifically.

2.3 Frameworks

As Birkinshaw & Mol (2009) have made very clear, the reasons for adopting innovative management practices have received relatively little scientific attention in the past decade. The studies that have are dusty or have a conservative approach to them. Different aspects are of importance when approaching the adoption of a management practice. As mentioned before, internal and external characteristics of an organization should be taken into account (Birkinshaw & Mol, 2009).

Internal characteristics could be viewed at from different perspectives. For example, through a lens of the resource-based view within organizations, which translates into an organizational framework of approaching DDD adoption. The contents of this framework can either have a contributing or negative effect towards the adoption of an innovative management practice. Furthermore, external factors should be taken into account. A broad

4_{Answers that correspond to the top 2 categories respectively are: (1) decision making relies heavily on data,}

(2) decision making relies entirely on data. And (1) a great deal of data to support decision making is available, (2) all the data we need to support decision making is available.

(22)

variety of reasons can be identified under which an organization decides to implement changes. When taking a new institutionalist perspective into account, it could be argued that reasons for adoption are not solely based on internal company characteristics and traits but as well dependent on the institutional framework an organization operates in (DiMaggio & Powell, 1983; Powell, 2007). Thus, when discussing the adoption of a management practice whilst concentrating on firm characteristics, firm characteristics that are related to a firm’s institutional framework should be taken into consideration. For example, Birkinshaw & Mol (2009) argue that management innovation and implementation is a consequence of an organization’s internal context and an external search for new knowledge. Factors like these can be seen through a neo-institutionalist lens and translated into a network framework. Important factors that are attributable to this framework are often difficult to isolate because of their comprehensive and far-reaching roles. Therefore, they often affect or give substance to other frameworks as well. This same argument accounts for certain aspects of other frameworks. For example, the RBV also explains the outcomes of the external search for knowledge (Birkinshaw & Mol, 2009; Barney, 1991).

Literature suggests that specific internal characteristics are of major importance when deciding on the adoption of an innovative management practice as well. These internal factors have to do with organizational roles, top management in particular. These types of internal characteristics will be associated with both the resource- and knowledge-based view and translates into a managerial framework of approaching DDD adoption. Furthermore, some articles suggest differences between the adoption of innovative management practices also exist between industries and countries (Agarwal et al., 2014). Such findings can be translated into country framework.

The reason for approaching the adoption of DDD through the above-discussed frameworks is because the studies at hand offer no innovative or consolidated methods.

(23)

Either, a research copies an already existing framework and applies it to a different management practice, or the other way around. Combining concepts and methods to be able to create a more powerful framework and to execute a stronger analysis has not yet been done. Even in the very topical publication of Brynjolfsson & McElheran (2016) that studies the same management as this thesis practice (DDD in the US), does not consolidate the available knowledge on methodology to create a more powerful model. This thesis will try to change that by integrating several research methods that have been used on similar topics. The project could have been the building of a correct, new and comprehensive framework to approach research gaps and problems associated with management practice adoption. However, because of the scientific urgency for more research on big data related problems the choice has been made to focus on a specific management practice. A new framework will be used to analyze the determinants of adoption within an organization. Of course, it would be a good outcome if the frameworks are applied to other cases in the future.

In order to be able to formulate hypothesis that support the objectives of this research in the following paragraphs literature will be discussed to give substance to, and add to the explanatory value of the organizational, network, management and country framework.

(24)

2.3.3 Conceptual model

NL Framework6

US Framework (Brynjolfsson & McElheran, 2015; 2016)

2.4 Hypotheses

To be able to find answers to the question what determinants for DDD adoption in Dutch business are, the explanatory factors are divided into three ‘frameworks’. These frameworks present mutually exclusive factors that influence the adoption of DDD. The frameworks are: organizational, network and managerial.

6_{Within conceptual model in parentheses the corresponding hypotheses numbers and signs corresponding to the}

hypothesized effects.

Note: H11 sign depends on the industry. More information with respect to this variable under section 2.4.4.

Organizational framework

Firm size (H1+), firm age (H2+), firm

structure complexity (H3+),

structured management (H4+)

investment in IT (H5+)

DDD Adoption

Convenience characteristics _{DDD Adoption}

Comparison Comparison

Hypotheses related to US-NL comparison arrow: - Country characteristics (No hypothesis;

see section 2.4.4) Institutional/network framework

Organizational learning (H6+)

Management framework

Top manager age (H7 –), education

(H8+), tenure (H9 –), experience

(H10+) Control variable

Industry sectors (H11+/–)

(25)

2.4.1 Organizational framework

2.4.1.1 Organizational size, age and structure

Organizations have different capabilities in terms of being able to implement and make use of technological advances. In his seminal work Damanpour (1987) has found several organizational factors that influence the adoption of technological innovations. The five organizational factors that Damanpour (1987) describes are specialization (variety of specialists within an organization), functional differentiation (degree to which an organization is divided into different units), professionalism (education and experience of employees), size and slack (power to innovate, formally defined as the difference between the resources a firm has and what it minimally requires to maintain operations).7 The two former are supported by the KBV, while the latter two are supported by RBV argumentations. (Professionalism will be discussed under section 2.4.3).

Many scholars have researched the impact of a firm’s size on its innovation capacities. The effect of firm size on a company’s innovation capacity can be argued in both ways. On one hand, it is likely that smaller organizations are more innovative because of several reasons: faster decision making process, less bureaucracy, more flexibility in structure, easier to adapt and improve and smaller companies often experience less difficulty in accepting and implementing change (Damanpour, 2010). However, larger companies are likely to be more innovative as well because they usually have more resources (both financially and technically) available which allow for hiring of skilled and professional workers, economies of scope that allow for spreading risks and absorb innovation costs, and possibilities to establish and maintain scientific facilities and so on (Damanpour, 2010). Damanpour (2010) distinguishes between process and product innovations and argues that firm size is more positively associated with process than product innovations. Small firms are likely to invest more in product innovations because these are seen as better instruments for

(26)

creating growth. Larger firms benefit more from spending resources on process innovations because they are likely to affect an organizations output in its entirety instead of only in a niche (economies of scale).

For example, the adoption of Total Quality Management (TQM) as a management practice has been a hot topic within social sciences. This new type of approaching management was very innovative at the time. Temtime (2003) has argued that most research on TQM has been performed on large organizations, thereby neglecting SMEs and the potential size effect on the adoption of this practice. The implementation process of TQM is likely to be similar and probably even contributes to the possibilities of DDD within an organization. As mentioned above the size of a firm is often associated with higher available resources and capabilities. DDD, just like TQM, is a management practice that does not entail a ‘one-size-fits-all’ approach, and therefore is a demanding process for all organizations that adopt it. A positive relationship between the adoption of DDD firm size should be assumed (Temtime, 2003; Damanpour, 1987; Agarwal et al., 2014). Therefore, the following hypothesis is formulated:

H1: The size of a firm is positively related to the adoption of DDD.

The relation between DDD adoption and firm age is ambiguous. The age of a firm is often seen as an indicator of external legitimacy, staying power but too the pervasiveness of internal routines (BarNir, Gallaugher, & Auger, 2003). It could be argued that a younger firm does not yet have the needed resources for adopting data intensive management practices. On the other hand, it could be argued that older firms that already experienced a professionalization process are surlier in implementing new practices. Age is a difficult concept within business because in most common economic theories the age of a firm has

(27)

implications. Just as with size it is assumed that a smaller company is younger and often looking for growth. This however, is not convincing reasoning since not all companies are looking to become market leader or to keep growing. It is important to take these implications that underlie the age factor into account. However, because it can be concluded that age will have an effect, the following hypothesis is formulated:

H2: The age of a firm is positively related to the adoption of DDD.

Evidence is found for a significantly higher likelihood of DDD adoption within American manufacturing firms when they are part of a multi-unit, international firm (Brynjolfsson & McElheran, 2016; Mol & Birkinshaw, 2010; Agarwal et al., 2014). This could entail being a production plant with a headquarter at another location, being an establishment of a larger corporation or being a subsidiary of a large multinational corporation (MNC). The mechanism that causes this effect has to do with the higher complexity of an organization that has more than one and often international units. If an establishment is part of such an organization this is associated with more complex intra-company communication and management. Higher complexity is logically associated with a higher need for practices that help in overcoming the restraints of such complexities. These practices could be DDD or other management and technological innovations.

H3: A firm being part of (multi-unit and international) complex organization (as a subsidiary, plant or establishment) is positively related to the adoption of DDD.

(28)

Some ‘dusty’, however, still important articles in the field are related to a major management change that business all over the world experienced in the 1960s. TQM was a real innovation (or fad) back then, and in the decades that followed, and is still used within many businesses. The essence of TQM is about continuous improvement of an organization’s products and processes with emphasis on the customer and the firm’s strategy (Young et al., 2001). That does not sound too exciting, because it is pure logic that a management practice should be in place to serve an organization’s strategy. However, at the time this type of management was very innovative and required a complete new perspective on doing business and generating profits from organizations. It does not come as a surprise that many researchers were keen to contribute to knowledge on this new process. TQM has been implemented in a broad variety of organizations (private, public, profit, non-profit and so on) (Temtime, 2003). This specific management innovation is an example of another advanced type of management. If an organization is applying lean management methods, which are methods that are associated with for example Total Quality Management (TQM) and continuous improvement approaches, then those managers are likely to have particularly high standards in relation to the use of data or not (Brynjolfsson & McElheran, 2015). If a firm has experience and is professional with its management they are more inclined to adopt similar or more advanced practices within their organization. If an organization has internalized this form or a similar form of management this results acting on a basis of structure. A structured approach like this allows for a structured changing process within organization. The adoption of a management practice like TQM also has been proven to have followed strong adoption patterns related to neo-institutional arguments. Therefore, the following hypothesis is formulated:

H4: The implementation of structured management is positively related to the adoption of DDD.

(29)

2.4.1.2 Capital investment

The adoption of DDD is inherent to complex and time-consuming implementation process. To start firms usually go through a lengthy process of learning what they know. Much knowledge inside firms is tacit. Employees often know more than they can tell (Polanyi, 1969). Many of the data that is available within the organization is often unstructured which causes unease and is for many employees too difficult to interpret (The Economist Intelligence Unit, 2012). Another part is the often necessary update of an organization’s IT infrastructure. Therefore, Brynjolfsson & McElheran (2016) argue that firms that have recently made high IT investments are more likely to have adopted DDD.

H5: Investments in IT are positively related to the adoption of DDD.

2.4.2 Network framework

A significant effect on the adoption of an innovative management practice has also been found in network and institutional effects. These effects are associated with theories on early and late adopters of new practices and techniques (Rogers, 1963/2003). The more organizations in a firm’s network have adopted the practice, the higher the likelihood of adoption for the firm (Young et al., 2001). The mechanism that causes this effect are ‘normative pressures for conformity’. Effects like these are seen as central within institutional theory literature and are called isomorphic pressures (DiMaggio & Powell, 1983; 2012). The number of learning sources an organization uses provides a good view on the degree of embeddedness in its institutional environment (Brynjolfsson & McElheran, 2015).

Furthermore, as an organization it is difficult to be able to discover and implement the many organizational changes that are needed to use DDD effectively. For this reason, organizations need instruments through which they learn about DDD. These instruments can

(30)

be of broad variety of institutional entities. For example, consultants, competitors, suppliers, customers, trade associations, conferences, new employees, headquarters or subsidiaries (U.S. Census Bureau, 2015). This relates to what the opportunities within the organization are, and how to adopt and implement in an effective way (Suddaby & Greenwood, 2001). To find a confirmation of the thought that DDD adoption is a result of organizational learning and a consequence of an organization’s institutional framework the following hypothesis is formulated:

H6: The number of learning sources an organization consults, is positively related to the adoption of DDD.

It should be taken into account that indeed it is not possible to draw a simple line between what is and what is not an effect of, or should be seen as, a neo-institutional factor. For example, in the next paragraphs factors are introduced about top manager characteristics. It could be argued that the behavior of top managers is a consequence of their embeddedness (Powell, 2007). However, for the sake of clarity these factors will be discussed within the managerial framework.

2.4.3 Managerial framework

From literature a dichotomous view has emerged. Some articles give attention to the characteristics of employees and managers in general (Brynjolfsson & McElheran, 2016; Damanpour, 1987). This research does not disagree with this statement, however, support was found for a different and better approach to explaining the adoption of DDD. Since the adoption of a new management practice often follows a top-down process, it is even more valuable to include characteristics of a firm’s top management (Damanpour & Aravind, 2011; Daft, 1978). In the earlier mentioned study about TQM, Young et al. (2001) have emphasized the importance of the ‘human’ side in whether an organization adopts an innovative

(31)

management practice or not. In particular, they have empirically confirmed that the characteristics of the top manager in an organization has significant influence on whether a practice is adopted or not (Wamba & Carter, 2013; Young et al., 2001). They define top manager as the person within an organization (or an establishment of the organization) that exerts the highest amount of power in relation to decision-making, and often is the director. The reasoning behind the decision to integrate top management characteristics levels instead of more general characteristics on management and employees has to do with concerns about the representativeness and relevance of the latter. This research is about DDD and its adoption. If lower managers or other employees get authority to make particular decision within an organization these will be given to them accompanied by instructions on how these decisions should be made. DDD is part of a management philosophy and approach to decision making. Again, such an approach is implemented top-down. Moreover, research has shown that top management characteristics better explain why organizations decide to move forward with investment in any innovation (Wamba & Carter, 2013). Therefore, the most relevant explanation of DDD adoption will be found by taking a closer look at characteristics from top management (Young et al., 2001; Damanpour & Aravind, 2011; Wamba & Carter, 2013).

Aside from the effect a top manager can have within the organization the manager is active in, it should also be taken into consideration how the top manager is affected by his/her external environment. In particular with regard to their experience with DDD and their education levels. It could be argued that the behavior of top managers is a consequence of their embeddedness in an institutional context, which often is a result of education or experience within a certain context (Powell, 2007). Nevertheless, as is mentioned before it is not that easy amalgamate these characteristic. Top managers should also be seen as resources

(32)

and simultaneously as holders of (tacit) knowledge within a firm and therefore affect the likelihood of DDD adoption.

2.4.3.1 Top manager age, education, tenure and experience

Scholars have argued that the age of managers has a negative effect on the adoption of technological changes but as well for process changes in general. The older the manager the more socialized into the ‘prevailing routines’ and the greater ‘psychological commitment to them’ results in a low willingness to commit to change and innovation. This fact in combination with a decrease in flexibility from a cognitive standpoint results in an assumed negative relationship between age and adoption of DDD (Wamba & Carter, 2013; Damanpour, 1987; Young et al., 2001).

The level of education is commonly recognized as an important predictor of the adoption of any innovation. The level of education appears to determine the responsiveness of top management to ‘new ideas, the feeling for new innovations and to understand the necessity to create a favorable atmosphere for its adoption (Damanpour & Aravind, 2011; Wamba & Carter, 2013). Therefore, we formulate the following hypotheses:

H7: The age of a firm’s top manager is negatively related to the adoption of DDD.

H8: The educational level of a firm’s top manager is positively related to the adoption of DDD.

With regard to tenure somewhat of the same arguments as for top manager age count. However, there exist more conflicts with regard to the mechanism that causes the proposed negative relationship between tenure and DDD adoption. From one point of view it could be argued that an increase in tenure leads to top managers that are stale and resistant to change. However, it could also be said that with increasing tenure top managers are more experienced

(33)

and therefore better capable of coping with challenges (all, organizational, political and cultural) associated with the adoption of DDD (Young et al., 2001). However, a relatively recent article has performed a meta-analysis on the relationship between tenure and engaging in innovation-related activities and found a negative however, not significant relationship (Ng & Feldman, 2013). The evidence seems not convincing and no consensus has been reached among academics. Therefore, the following hypothesis is formulated:

H9: The tenure of a firm’s top manager is negatively related to the adoption of DDD.

Previous experience with emphasizing data as a source of decision making can be an indicator of a top manager’s propensity to adopt that practice in the future. This factor of previous experience can be seen both as a top management indicator but as well as an indicator of institutional and network effects at work (Rogers, 1963/2003). When a top manager leaves a previous employment he is not only ‘moving from one network tie to another, but also establishing a link between these’ (Young et al., 2001). A top manager with experience and a bond with a network that is using a similar practice will be inclined to adopt that innovation within the organization of his/her current employment. Therefore, the following hypothesis is formulated:

H10: Previous experience of a firm’s top manager is positively related to DDD adoption.

2.4.4 Other contextual factors

The industry an organization operates in could influence the decision to adopt an innovation. It can be logically inferred that the technological context of the organization and innovation

(34)

is more important for a manufacturing organization than for a tourism agency (Wamba & Carter, 2013). The different industry types between which a distinction is made are the financial sector, manufacturing, energy & resources, consumer and health & pharmacy. To address these expectations, the following hypotheses are formulated:

H11: The industry an organization engages in influences the adoption of DDD.

2.4.4.1 Country framework

Agarwal et al. (2014) emphasize the importance of a specific country context in the case of Australia. Noteworthy, is the fact that the Australian research is focused solely on Australia, therefore the results feed the curiosity of executing a real comparative analysis. In my opinion country characteristics are of diminishable value when solely taking into account organizations from that country. It must be said that the Australian research has extracted most of its data from the World Management Survey (WMS). As the title of this survey suggests they did not only have data available for Australia. However, the number of participating countries in this survey has been limited to eighteen. Since no Western European countries have been taken into account and the ‘population’ of this research seems to be very limited, curiosity for a comparative analysis between our context (the Netherlands) and the United States (and perhaps even the Australian context) is increasing. It must be celebrated however that scholars have gathered such large amounts of data and are undertaking such ambitious projects to gather data on management practices from all over the world. Unfortunately, for many countries the data is full of gaps or entirely missing. In this observation more support is read for the use of a self-executed survey to perform this research. Again it must be noted that the research methods performed in this thesis are significantly different from that in the Australian context because it focuses on DDD as a specific management practice.

(35)

Since the probable differences in DDD adoption that follow from country specific contexts are not possible to statistically test or measure within the scope of this thesis, no hypothesis will be formulated. However, in the discussion section some attention will be given to this phenomenon.

3. Methodology

With regard to DDD some research has already been performed. This research has focused on the US. As region to analyze for this thesis the Netherlands is chosen. A similar research will be performed, however, this project will be more extensive. The research of Brynjolfsson & McElheran (2016) shows a solid framework that is not yet optimal. I argue that, the results from their study suffer from omitted variable bias, because they do not take the managerial framework and some factors from the organizational framework into account. Therefore, I propose a different and more comprehensive approach. This approach will be presented in a structured manner. To this end, three dimensions have been developed which will be empirically tested. Another dimensions that allow for comparing the situation in the US and the Netherlands are not empirically tested. However, observable differences in outcomes will be explained through use of sources from the above literature review. The empirical analysis will test what factors can be named determinants of the adoption of DDD within Dutch business.

3.1 Design

3.1.1 Method: Survey

The main objective of this research will be explanatory. The research question will be answered through the execution of a longitudinal survey design (See Appendix). All participants will receive a sample of questions from a similar survey (Management & Organizational Practices Survey) that has been executed in the US, by the U.S. Census Bureau (U.S. Census Bureau, 2016). Brynjolfsson & McElheran (2016) have used the data

(36)

from this US survey to research the adoption of DDD in the US. To be able to assure a similar approach that possibly allows for a comparative analysis, questions in this research will be asked identically to questions in MOPS. The survey questions are built in such a way that a quasi-panel structure will emerge (e.g. answer the question for the situation as it is now and as it was 5 years ago) which will allow for a longitudinal analysis (fixed effects).

The survey was administered online by using Qualtrics software. Since the survey questions have been used in a U.S. Census bureau survey it was not necessary to perform a pre-test. However, to assure the right understanding of the questions asked in the Dutch context the survey was pre-tested with desirable outcome by twenty peers. These peers gave a response from the perspective of a fictional organization. The survey was administered in two languages, Dutch and English. The translation of the survey was done through back-translation (see Appendix).

3.1.2 Sample

The researched population are all organizations that do business within the Netherlands. The sample is built by extracting contact data from organizations that are registered at the Dutch chamber of commerce (KvK). A hard condition for being able to use the data of an organization was the availability of an e-mail address where the survey could be sent to. The contacts extracted from KvK were complemented by the contact database of student organization Integrand Nederland. Thus, the potential participants included in the sample were contacted on the basis of a non-probability sampling technique. The number of firms that were invited to participate in the survey was around 8,114. After having sent two reminders (the 1st_{after 2 weeks, the 2}nd_{after 4 weeks), the total number of responses was 470} (respondents that made a start with the survey). In order to be able to fulfill conditions for doing a (panel) regression analysis some transformations were done and some responses needed to be dropped. Counter-indicative variables from the survey were recoded. Many survey responses appeared to be invalid because of a major number of missing values. Three

(37)

strategies were adopted to assure the dataset being analyzable and to assure that the estimates were not going to be biased. The combination of strategies that were used are: list wise and pairwise exclusion of survey responses that had unlikely answers or did not fill in the questions essential for the model. For the IT investment and revenue question many missing values were registered. In order to still be able to integrate this variable a tactic of mean substitution was used to fill a relatively small number of missing values. For these questions in advance it was expected that respondents were going to be hesitant to disclose such information. Only the sheer number of 107responses were found to be valid (which results in 178 observations). Furthermore, some data were skewed or kurtosis. To normalize the data transformations (logarithms) were performed (Federici, 2016).

3.1.3 Measure description

Dependent variable – DDD Index

The dependent variable in this research is the degree to which an organization has adopted DDD. This degree is measured by building a normalized DDD Index on a scale [1,10]. The index is built by combining four questions from the survey. These questions correspond with the earlier introduced definition of DDD (Brynjolfsson & McElheran, 2016). In this research the survey questions that are included in building the DDD Index correspond to question 18, 19, 20 and 21 of the survey. Respectively, these questions ask the organization about their use of key-performance indicators (KPIs), the use of targets and the availability and use of data to support decision making. The scale is established by taking the sum of the normalized responses (some questions had answering possibilities 1-4 and some 1-5. They were all reduced to 1-4 summed, divided by the total maximum score and multiplied by 10 to create the [1,10] scale. For all four questions participants were asked what the situation in their organization was by either using a four- (question 18 & 19), or five-point (question 20 & 21) scale.

(38)

The DDD Index scale has high reliability, with Cronbach’s Alpha = .714. The corrected item-total correlations indicate that all the items have a good correlation with the total score of the scale (all above .30). Also, none of the items would substantially affect reliability if they were deleted.

A principal axis factoring analysis (PAF) was conducted on the scales. The Kaiser-Meyer-Olkin measure verified the sample adequacy for the analysis, KMO= .665. Bartlett’s test of sphericity c2_{(6) = 192.103, p < .001, indicated that correlations between items were} sufficiently large for PAF. An initial analysis was run to obtain eigenvalues for each component in the data. One component had an eigenvalue over Kaiser’s criterion of 1 (value = 2.171) and in combination explained 54.3% of the variance. In agreement with Kaiser’s criterion, examination of the scree plot revealed a leveling off after the first sector. Thus, one factor was retained and rotated with a Oblimin with Kaiser normalization rotation. Table 1 shows the factor loadings after rotation. The items that cluster on the same factors suggest that factor 1 represents the DDD adoption scale.

Table 1 – Principal factor analysis of DDD Index (~208 obs.)

Factors Eigen value Proportion of variance

Factor 1 2.171 0.543 Factor 2 0.809 0.202 Factor 3 0.702 0.176 Factor 4 0.317 0.793 Scales 1. Top 2 categories for 'availability of data' 2. Top 2 categories for 'use of data' 3. Track 10 or more KPIs 4. Use of short- and long-term targets Factor 1 loadings

1. Top 2 categories for 'availability of data' - 0.743

2. Top 2 categories for 'use of data' 0.675 - 0.867

3. Track 10 or more KPIs 0.277 0.363 - 0.434

4. Use of short-term and long-term targets 0.320 0.355 0.298 - 0.452 Note: Calculated using the Dimension Reduction Factor Analysis in SPSS 22.

The adoption of data-driven decision making in the Netherlands : organizational, network and managerial determinants of DDD adoption