The success factors of a digital transformation towards data driven maturity

(1)

The success factors of a digital transformation

towards data driven maturity

MSc in Business Administration - Digital Business

Name Elles te Riet

Student number 11147369

Supervisor Dr. Prof. H. Borgman

University University of Amsterdam

Date 23-06-2017

(2)

Statement of originality

This document is written by Elles te Riet 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.

Acknowledgement

I would first like to express my gratitude to my thesis supervisor Dr. Prof. H. Borgman of the University of Amsterdam. I would like to thank him for leaving his door always open for discussion and for showing great enthusiasm and engagement on the topic. His critical though positive remarks resulted in a pleasant and valuable learning process.

Furthermore, I would like to thank all participating companies and employees for your

involvement in this research. Without your openness and insights this research could not have been conducted successfully.

(3)

Abstract

The speed at which organizations are required to constantly adapt to the ever-changing landscape has exponentially increased in the current digital era. The rapid growing impact of new technologies is forcing organizations to transform towards a data-driven oriented business. Considering the Davenport Maturity Model, different levels of data driven maturity at organizations can be identified. What are the drivers for successfully transforming to the next stage? This case study research intends to define factors of impact, derived from interviews across c-level managers, middle managers and data-analysts from three low and three high performing data driven organizations. It reveals insights into what factors are promoting data-driven transition for each DELTA element at different stags of maturity; it shows how differences in perceived organizational maturity are demanding different change management approaches; and last how different structures of organizing analytics impact the success of the transformation. It is a theoretical contribution to existing change management literature as it provides greater understanding of organizational change during the digital transformation towards a data-driven organization. It contributes to the usability of the Davenport Maturity Model, by giving a deep insight on the necessary factors to successfully move to higher stages of maturity for each DELTA element. Moreover, this study has a societal contribution for practitioners actually facing the challenges of digital transformation, to achieve faster success by giving priority to the most successful determinants of becoming a data driven organization.

(5)

Introduction

Data-driven decision-making is about building abilities, tools and most important a culture that is acting on data (Anderson, 2015). Taking advantage of the available data has become of

critical manner for organizational success, making it important to derive insights from data

and to use them for smarter decision-making (Olszak, 2016). The importance for companies to use data-driven decision-making will only grow in the future, because predictive analytics are defined as the new big differentiator for competitiveness (Davenport, 2007). Organizations that use analytics show improved financial performance, faster decision- making, increased productivity and reduced risks and costs. Companies defined as top-performers in data-driven decision-making are on average 6% more profitable and 5% more productive than their competitors (McAfee, 2012).

However, businesses are struggling to fully realize the efficiencies of digital capabilities when undergoing digital transformation (Schwab, 2016). When investing in new digital initiatives like data analytics, it remains the most significant challenge to successfully manage the required change in the organization, often being the consequence of low success rates (Global Agenda, 2014). Hence, to become a data-driven organization, good execution of the organizational change is a prerequisite to realize benefits.

Throughout the post decennia organizational change processes have deeply been analyzed, resulting in an extensive amount of literature on different change management approaches (Kotter, 1995; Lewin, 1948). Various authors have stressed that being able to manage organizational change is a prerequisite to keep up with competition (Ford, 2008; Kotter, 2008;

Warren, 2014). However, transformation in the age of the digital revolution seems to be of

total different scope and speed than previous organizational changes. One might argue that there are new and different drivers for success to manage this data-driven organizational change towards data-driven maturity.

(6)

The definition of successful analytical businesses is discussed through the widely accepted model of Davenport (2010), building an analytics maturity model encompassing the three most vital areas: people, technology and processes. There are five stages, measured along five elements with the acronym DELTA: Data, Enterprise, Leadership, Targets and Analysts. Organizations are located at different stages of maturity with regards to different elements. Despite the fact that maturity models define different stages, they not provide the necessary steps organizations should take to transition to more mature stages. Furthermore, no previous research in the change management literature investigated factors determining successful organizational change towards a data-driven decision-making culture during digital transformation. To start filling this gap, the purpose of this paper is to extend the theoretical knowledge on what drives organizations forward to become more data-driven. The following question will be central to this research: What are the success factors for organizations to

reach data-driven maturity? To answer this question, the study will focus on the three

following sub-questions: What are the drivers for success for each DELTA element in order to move to the next level of data driven maturity? Are there perceived differences between different layers in the organization? And last, how is the organizational structure of analytics impacting this transformation? This is researched through multiple cases studies via interviews across c-level managers, middle managers and data analysts from 3 low and 3 high performing data driven organizations to give a first view on the insights. Furthermore, direct observation is used to investigate cultural aspects and the structure of analytics within the organization.

This research has a theoretical contribution by comparing organizational change during the age of digital revolution in contrast to earlier organizational changes. Furthermore, it contributes to the usability of the Davenport Maturity Model, by giving a deep insight on the necessary elements to successfully move to higher stages of maturity for each DELTA

(7)

element. Moreover, this study has a societal contribution for practitioners actually facing the challenges of digital transformation, by intending to define relevant factors that answer the crucial question on how to become more data-driven as an organization.

The remainder of this paper is organized as following. First, existing change management literature will be assessed. Furthermore, the definition of a data-driven business will be clarified. Thereafter, it elaborates on the research approach and gathered field data. Findings of the various case studies are presented as results, and further evaluated in the discussion section to examine the three working propositions. Then, the findings are concluded and followed by managerial implications. Last, the limitations and suggestions for future research are provided.

Theoretical Background

In section 2.1 a deeper understanding through historic literature on the evolution of organizational change is established. This helps to better understand differences between the processes that companies are facing today in the era of digital transformation, compared to earlier organizational changes. Thereupon, the findings of this study can be placed to a better degree in the context of general change management practices. Two theories in specific from Kotter (2012) and Warren (2014) will be highlighted on grounds of their relevance to change management practices during the digital transformation. Elements of the digital transformation will be further elaborated on in section 2.2. In section 2.3 and 2.4 the practice of data-driven decision-making and analytics are explained. Last, in section 2.5 the Davenport Maturity model will be introduced.

2.1 The evolution of organizational change

Charles Darwin: “It is not the strongest of species that survive, nor the most intelligent, but the one most responsive to change.” For decades organizational change has been a

(8)

prerequisite to stay competitive, forcing companies to be in constant change (Conner, 1992). One might say that there are four eras with each its own framework to understand the process of organizational change (Warren, 2014). First, during the era of scientific management early

in the 20th century, the concepts of quality control and standardization were central, with

focus on productivity and output. Second, when the human relations took over in approximately 1930, it shifted towards theories of intrinsic motivation. Feelings of satisfaction, motivation and engagement became the terms people were concerned about. Around the eighties during the third so-called social-technical era, the complex interaction between social systems and technological advancements made it important to balance the two. Key was focus on the interaction between processes, systems, people and the corresponding development of skills, to achieve high employee participation on all fronts. However, learning was viewed as a goal with an end, rather than a continuous process.

Nowadays, we live in the fourth and current era, in which a holistic approach is taken for complex adaptive systems. In order to achieve sustainable growth over the long run, different new initiatives and technologies need to be initiated and the gathered insights are constantly shaping the process of organizational change. It is perceived as pervasive, touching continuously all aspects of the business (Mullins, 2005). It can be stated that organizational change is germane to all organizational aspects, so an enterprise-wide approach is required to manage the constant changing conditions. The stakes organizations nowadays are dealing with during their organizational change, has never been has high since the Industrial Revolution (Beer & Nitin, 2000). So, the difference is that today organizations are faced with a continuous process of change that is constantly forcing the organization to adapt to new situations in high speed. It does not follow a simple predictable path, but rather is a non-linear process.

(9)

2.1.1 Structuring the organizational change

Kotter (2012) analyzed over hundred companies in their attempt to become more

competitive. He found that traditional companies, with hierarchical processes and systems, are good at executing day-to-day operations, but lack the flexibility to cope with the complexity of the rapid changes occurring nowadays. Similar struggles are faced by organizations trying to become more data-driven. They face the difficult dilemma on how to structure their analytics as such, that it remains in line with the operating business while ensuring innovative relevance (Davenport, Morrison and Harris, 2010).

Kotter (2012) suggests working with a second operating system that complements your already existing structure, leaving the operational unit free to work as optimized as possible. This second system is devoted to create and implement strategy. It increases the reaction speed, agility and creativity by which the company is assessing the business. Together it leads to handling the operational needs in an efficient way, while being continuously able to adapt to the environment and implement new initiatives. In this dual operating system, volunteers are assigned as change agents to carry out the second operating system. A sense of urgency needs to be created to let more people stand up and faster the process of change. Mobilization will only take place when the naturally born coalition of volunteers feels the desire and shared purpose to make it a success. Besides, it should be a combination of heart and head. People should believe and feel that they give meaning to a better future by doing their work. Clarification of how the initiatives are helping to achieve future goals is highly needed. Furthermore, soft elements like leadership, vision, inspiration and celebration should replace hard measurements like reviews, accountabilities and reports.

When this is accomplished, Kotter (2012) argues that massive numbers of employees should gather together in order to make the change happening on large-scale. Inefficient processes and hierarchies should be removed, to enable the necessary freedom to create real impact

(10)

across silos. Short-term wins should be generated, recognized and collected to track progress and thereafter communicated to enthuse the volunteers to stay on board as promoters. Acceleration is sustained by increased credibility through relentless initiation of new changes until the final goal is reached. Then the two different systems should act like one organization. Information and activities need to be shared constantly among all stakeholders in the organization. Finally, change should be institutionalized. Here, the new behavior needs to be linked to organizational success, ensuring employees do not fall back into old habits.

2.1.2 Managing the organizational change

Various authors have stressed that being able to manage organizational change on constant base is a prerequisite to keep up with competition (Ford, 2008; Warren, 2014). The process of organizing and managing change is a never-ending process, touching all types of employees at various levels of the organizations. It goes beyond individual projects, requiring enterprise-wide capabilities for managing initiatives (Warren, 2014). Some basic elements of managing organizational change have been explored by numerous studies throughout the years and are stable through all different types of transformations (Kotter, 1995; Beer et al., 1990; Miles, Creed and Coleman, 1997; Warren, 2014). These include that effective change programs start with the central board initiating the activity, establishing urgency and clearly communicating the activities undertaken. Then the managers and their teams who are most willing to change start with the execution. These managers are crucial to get on board in order to effectively manage the change from top-down (Warren, 2014). Rejection for accepting the change occurs when people can not visualize the new reality. Therefore, it is best to work with concrete business questions with clear objectives. Meanwhile, a visionary leader needs to communicate an inspiring vision, in which the followers believe and see a meaning and purpose (Mumford, Scott, Gaddis and Strange, 2002). The behavior of a visionary leader is a central component in the transformation process (Bass, 1985). As humans are social creatures, the social

(11)

reinforcement of others is affecting the belief and opinion an individual has (Sherry, 1997). Trust can reduce vulnerability and help the employee to identify why, what, how and when interchange is taking place (Benbasat, Geven and Pavlou, 2010). People who have trust in the leader that is telling him/her to change their behavior, will be more likely to adapt to the leaders’ desired actions. Then likeminded employees from other units of the organization will be reached and every corner will be touched. Thereafter, the roles and relationships of employees need to be reorganized, the transitions clearly monitored and changes need to be anchored in the culture.

2.2 Digital transformation

Companies are in constant change, where digital transformation is a new dimension in change management, making organizations strive for understanding on how to deliver new digital strategies. It is an organizational change of a complete different scope and speed. Many definitions exist, though in this research digital transformation is defined as: ‘the use of new digital technologies (mobile, social media, embedded devices and analytics) to enable major business improvements’ (Matthews & Brueggemann, 2015). It is leading to lower costs, increasing profits, the creation of new products and services, and to prediction of current and future business models (Kotter, 2012). Still, organizations struggle to fully realize the efficiencies of digital capabilities when undergoing digital transformation (Schwab, 2016). First, this is due to the exponential increasing speed of change by which organizations are required to constantly adapt to new processes and developments. This is reflected by ‘Moore’s law’, stating that the processing power of computers doubles every two years (Moore, 1975). Second, where organizational change in earlier decades was about adapting your company to a certain development in the market, companies now need to be constantly able to adapt to any changes in the market: a total different organizational change process. Like Andrew McAfee and Erik Brynjolfsson argue in the Second Machine Age (2014),

(12)

technological advancements are leading to unprecedented disruptive change. The impact of exponentially growing technologies (like autonomous drones, 3D printing, intelligent robots, machine learning techniques) will without doubt impact how people run their organizations in the future (Gartner, 2013). Third, due to dependency on the complex data infrastructure during a digital transformation towards a data-driven organization, the barriers faced by organizations are different.

2.3 Data-driven decision-making

Nowadays everything can be measured, and every important decision should and can be supported by using available data. Even though it might seem challenging, top performing analytical organizations compared to low performers, make double the amount of decisions for day-to-day operations on rigorous analysis and see the use of data analytics as a key differentiator (LaValle et al., 2011).

Todays’ organizational changes towards a data-driven organization demand complete different skillsets of people. These include in the broad sense technical and quantitative skills, interpersonal and communication skills, a sense of creativity and lastly business knowledge (Davenport et al., 2010). A popular estimate states that more than half of the children going to primary school today will later perform jobs that do not even exist today due to the impact of digital. When technology keeps making progress like it is doing today, it will leave behind a few or perhaps a lot of people in their current roles. Data driven working is not just something you need to learn alongside your usual execution of tasks, it also requires a change in working in the core of your everyday job. It demands life-long learning in order to entail the constantly changing essential skills. Critical issues are the power of human ingenuity and creativity, to fully cope with this increased pace of change. As such, when a digital tool is introduced to people, various factors influence the people’s intention toward use, attitudes and actual usage (Ngai, Chan and Poon, 2007; Childers, Peck, Carr and Carson, 2001). When innovative

(13)

technological systems and new working methods cross the paths of employees, often a threatening feeling occurs. People become resistant when their work environment changes towards new structures. The use of analytics for decision-making is a required change in behavior, and since people are creatures of habit, when feeling comfortable it is difficult to change this behavior. However, learning to successfully manage the change in behavior is highly important, as The Global Agenda (2014) reported that businesses that implement new digital initiatives like data analytics, encounter the biggest challenge in managing the required organizational change as such that employees at all levels are committed to succeed. Consequently, it is found that despite the availability of quantitative models, optimization methods and advanced business analytics, ‘best-guesses or gut-feeling’ remains the method that most employees still rely on (Anderson, 2015). Their decisions are based on some information, but it remains mysterious on what exact information outcomes are based. Various researchers have shown that next to rational reasoning, emotion and intuition are unconsciously playing a large role when people make decisions. Nobel laureate Simon (1967, 1983), introduced bounded rationality, refining the rational choice by including situational and cognitive constraints. This explains why the human brains show difficulties in making decisions fully based on the overloading amount of data subjected in daily business. Our rational brain is better suited for easy problems, where complex problems are better suit to the emotional brain. In complex situations there are too many variables, trade-offs and constraints to consider, after which humans tent to go for their gut- feeling or intuition rather than seeking a rational for their decision.

2.4 Analytics

Among scholars, data analytics is becoming a hot topic (Wamba et al., 2017). Also, in the annual presentation of Forbes on top digital trends, the use of big data and analytics is one of the constant trends appearing on the list. IBM reported that in the last two years, 90% of all

(14)

data available today was produced. This indicates the exponential speed at which data is predicted to grow. Therefore, taking advantage of available data has become critical to organizational success (Olszak, 2016).

The definition of Davenport and Harris (2007) for analytics is the following: “extensive use of data, statistical and quantitative analysis, explanatory and predictive models and fact-based management to drive decisions and actions”. Being an analytical company means using data, analysis and systematic reasoning that is leading to better decision-making (Davenport et al., 2010). Multiple researchers (McKinsey, Deloitte, SAS) found that it pays off to use analytics: organizations who use analytics show improved financial performance, faster decision making, increased productivity and reducing of risks and costs.

Analytics can answer fundamental questions about the business. It goes beyond business intelligence practices like visualization and reporting data. Instead, it is about linking the explanatory variables beyond the surface of data to create business outcomes (Baesens et al, 2016). Davenport (2010) organized six type of key questions addressed by analytics across two dimensions: time frame: past, present and future & innovation: information and insight

(figure 1).

Figure 1: Six types of questions data and analytics can address

Past Present Future

Information What happened? A) Reporting

What is happening now?

B) Alerts

What will happen?

C) Extrapolation

Insight How and why did it _happen? D) Modeling

What’s the next best action?

E) Recommendation

What’s the best/worst that can happen?

(15)

All activities drive knowledge, though reporting (A) and alerting (B) are not defined as being data-driven. These activities are simply stating what happened in the past or present, without any recommendation on solving future situations. Extrapolation (C) can be seen as the ‘danger zone’: one might argue that with this type of questions naïve predictions for the future are made solely based on information, without any insights on why or how it has happened. Wrong or misleading conclusions are the results, leading to wrong decisions. Therefore, causal factors should be understood through digging down on the reasons why the data shows certain numbers, corresponding to the dimension of ‘Modeling (D)’. Truly data-driven organizations fall under Recommendations (E) and Predictions (F), where recommendations and plans can be formulated and predicted based on real understanding. Most organizations go from A to F, possibly skipping some steps. Truly data driven organizations focus on the bottom row which include causal explanations, where forward-looking activities are highlighted. Using analytics should therefore not be a tool to solve a particular problem, but an embedded organizational capability that is constantly supporting decision-making (Becker et al., 2012). It needs to be part of normal processes in the business so that it eliminates the gap between insights, actions and decisions.

2.5 Davenport maturity model

Maturity models are popular in a broad range of domains and industries (Weber et al., 2008; Scott, 2007) and show growing interest on academic level (Becker et al., 2010). It is used to assess situations as-it-is, to provide guidance for improvement and finally to control the progress (Iversen et al., 1999). Maturity models include a sequence of increasing stages from initial to maturity level, where each stage classifies a different state of the organization. First, the present maturity level is defined, representing the current capabilities regarding the specific application domain (Rosemann & de Bruin, 2005). Thereafter, the gap towards the desired maturity stage is analyzed in order to determine the deviation. Maturity models are

(16)

based on the assumption that the organizational evolution towards higher stages is a predictable pattern. The organizational capabilities should evolve in a logical stage-by-stage approach. Though, it might be argued that this oversimplifies reality (de Bruin et al., 2005). Hence, the use of a maturity model in this research is not for the purpose of predicting a step-by-step recipe. The different stages of the model are used to simplify the rather complex process of defining different levels of analytical success. It helps to determine the drivers for success at various organizations in a clear and consistent way.

For data-driven maturity in specific, several maturity models are developed. The

maturity model of Saxena and Srinivasan (2013), exists of three balanced dimensions: culture, capability and technology. They found that organizations often lack behind on technology but excel with regards to capability. Another example is the research of Cosic, Shanks and Maynard (2012), who developed a business analytics capability maturity model. Four capability areas are defined: culture, people, governance and technology. A third example is from Comuzzi and Patel (2016), who created a model around big data maturity consisting of six stages and five domains (data, organization, strategic alignment, information technology and governance). Last, researched at more than hundreds of companies over more than a decade, the so-called DELTA model from Davenport addresses the different stages of success when using analytics. It is selected as focus in this research, because it is one of the most well-known analytical maturity models among scholars and has a broad range of applications in research (Lismont et al, 2017). The following part explains the model.

2.5.1 Five DELTA elements

The Davenport DELTA model has defined the different elements organizations need to focus on in order to succeed with their data-driven transformation This acronym is segmented on five topics: Data, Enterprise, Leadership, Targets and Analysts. In figure 2, the attributes per

(17)

stage for each element can be found. In the following part, the relevance for each element will be shortly described.

Element 1: Accessible, high quality Data Clean and good data is defined as the prerequisite to be analytical. Data needs to accurate, integrated from various sources and easily accessible by employees. Important is that the data ads relevant new information to the existing knowledge.

Element 2: Enterprise orientation An enterprise perspective is important because analytical applications invariably impact multiple business units. When organizations only initiate small local activities without an overall perspective, no significant impact will arise. Furthermore, to establish cooperation on analytical initiatives on enterprise level, management of data, analysts and technology on local level would not be efficient.

Element 3: Analytical Leadership Analytical leadership distinguishes itself from regular leadership by the true passion for managing by facts rather than on gut feeling. The long-term goal of these leaders is to become analytical in their decision-making across the whole enterprise, not only in easy areas for quick wins.

Element 4: Strategic Targets No leader will make any investments without seeing the potential return. Analytical targets are therefore set in order to determine priorities and to set investments at the distinctive capabilities, in order to serve customers in a competitive way. Two basic targeting activities are setting the ambition and finding opportunities. The use of targets is of all DELTA dimensions most connected to other elements.

Element 5: Analysts Analysts are important because they build and maintain models in order to hit analytical targets of the organization. Furthermore, they enable the people in the business to use and apply analytics at large level.

(18)

Figure 2: Davenport Maturity Model - DELTA Stage 1: Analytically Impaired Stage 2: Localized Analytics Stage 3: Analytical Aspirations Stage 4: Analytical Companies Stage 5: Analytical Competitors Data Inconsistent, poor quality, poorly organized Data useable, but in functional or process silos Organizations begin to create centralized data repository Integrated, accurate, common data in central warehouse Relentless search for new data and metrics

Enterprise n/a Islands of data, technology and expertise Early stages of an enterprise-wide approach Key data, technology and analysts are centralized or networked All key analytical resources centrally managed Leadership No awareness or interest Only at the function or process level Leaders beginning to recognize importance of analytics Leadership support for analytical competence Strong leadership passion for analytical competition Targets n/a Multiple disconnected targets that may not be strategically important Analytical efforts coalescing behind a small set of targets Analytical activity centered on a few key domains Analytics support the firm’s distinctive capability and strategy Analysts Few skills, and these attached to specific functions Isolated pockets of analysts with no communication Influx of analysts in key target areas Highly capable analysts in central or networked organizations World-class professional analysts + attention to analytical amateurs

2.5.2 Five stages of maturity

Nowadays organizations recognize that to achieve high performance, the use of information throughout their business is a critical factor for success. Not every company is able or desires to put analytics at the core of the strategy, but there are always ways to better use information to become more powerful. Organizations face different starting points, various rates of progress and precede different capabilities in order to succeed in their progress. The five stages as defined by Davenport (2010) are described in the following part.

(19)

Stage 1: Analytically Impaired Several prerequisites are lacking to work with analytics on a serious level. Analytical skills, acceptable quality of data and senior management commitment are not present at this stage. Organizations on this level show inability to convert data into actions.

Stage 2: Localized Analytics: In this stage analytical activity does take place, but only

at islands without coordinated strategic targets. Organizations in this stage often struggle with inefficient tools and technologies on local level, being unable to use predictions created by the one unit in order to improve other units.

Stage 3: Analytical Aspirations The organization has established basic analytical

capabilities, is aiming for an analytical future and is already working on clear initiatives to become more data-driven. The value is recognized and these organizations have started with integrating several data systems. Though, the progress is slow and the organization still faces difficulty to improve at one or two DELTA elements.

Stage 4: Analytical Companies The organization managed to apply analytics on regular

basis, realizing the benefits throughout the organization.Technological and human resources

are in place (or easily acquired), but analytics are not yet grounded enough in the strategic focus to reach full competitive advantage. A clear vision and creativity is needed to become truly analytically competitive.

Stage 5: Analytical Competitors Both externally and internally the organization is

portrayed as an analytical competitor. Analytics has become a distinctive capability by using it on routine basis. Large-scale results are achieved, because leadership shows full commitment to make it an enterprise-wide activity. The workforce continuously is trained on new tools to keep up with the rapid pace of emerging technologies.

(20)

2.5.3 Organizing Analytics

Among scholars there has been discussion about optimal ways to organize analytical talent

(Saxena & Srinivasan, 2013; LaValle, 2011; Davenport et al., 2010). Although there is no best

way to organize analytics, according to Davenport (2010), the best fitting model ideally ensures the following aspects: valuable analysts should bear up with the enterprise perspective, are working on the most important analytical tasks and projects, face un-going opportunities for job development, and are challenged in their work so they stay highly satisfied. The way analysts are structured in the organization should therefore be in line with the desired coordination, aligned and group composition. When looking to factors that foster data driven transformation, it is important to take into account the impact that each structures has on the effectiveness. For multidivisional corporations five basic organizational structures are defined and further explained in the next part (figure 3, p. 21).

i) Decentralized

The groups of analysts are directly associated to their own business unit. Within this group, various analytical projects are executed without any corporate structure. This structure does not facilitate that analysts from different business unit communicate or share of best-case practices. Rotation and borrowing of analysts is difficult, and effective goal setting of enterprise priorities does not take place.

ii) Centralized

Even though all analysts are assigned to different functions or business units, they report and work from one central analytics group falling directly under the corporate board. The unit is funded by the central organization and employs analysts on the projects with strategic priority. Though, through this centralized structure the business might feel distance with the data analysts because initiated processes are perceived as dictated and obligatory to follow.

(21)

Figure 3: Organizational structures for organizing analytical talent

i) Decentralized ii) Centralized

iii) Center of Excellence (COE) iv) Consulting

v) Functional Corporate Unit Analytics Group Analytics Project Unit Analytics Group Analytics Project Corporate

Analytics Group Unit

Analytics project Unit Analytics project Corporate COE Unit Analytics Group Analytics Project Unit Analytics Group Analytics Project Corporate

Analytics Group Unit

Analytics project Unit Analytics project Corporate Unit Analytics Project Unit Analytics Group Analytics Project Direct association Consulting Community member

(22)

iii) Center of Excellence

In every business unit with appetite for analytics there is a group of analysts and all of them are members of the centrally organized center of excellence. In this community analysts learn from each other by sharing experiences both online and offline.

iv) Consulting

In this structure analysts are hired as consultants to execute analytical projects for business units. It is the most market-driven model, where consolidated analysts have an enterprise-wide view on all activities going on in the company. While being resident in the units, reporting still happens to the central consulting organization.

v) Functional

In the functional model, analysts are providing services to all units of the corporation, but are mainly executing analytical services for their primary unit. In this approach, migration to another unit can take place when the project at their own primary unit is finished. In this situation the analysts remain reporting to their own managers. In this structure the analysts often work at the marketing or sales departments (e.g. the “customer knowledge group” or “business intelligence group”).

Research Design

This chapter describes the research method. The structure is as following: section 3.1 describes the research question and working propositions, followed by section 3.2 explaining the sample selection. In section 3.3 the procedure is discussed and lastly the approach for the analysis is further enhanced in section 3.4.

3.1 Research question

This is multi-case study concerning the success factors for organizations to become more data-driven during their digital transformation. The purpose is to extent the knowledge on the

(23)

drivers for success for organizations to move towards higher levels of data-driven maturity. The following research question is central: What are the success factors for organizations to

reach data-driven maturity?

The interviews investigate possible drivers to move as a company to the next stage on the data Davenport Maturity Model (figure 2, p. 18). By drawing on previous theory, a large number of potential drivers that impact the transformation towards a data-driven organization are identified. To exclude the possibility for missing elements not yet mentioned in the relatively sparse amount of previous research, 20 short questionnaires (part A) and 4 additional interviews with experts who have cross-company experience in the field (part B) are conducted before the start of the actual interviews for the case-study. The list of participants and the short questionnaire can be found in appendix 1 and 2. This pre-investigation is not part of the actual case study, but executed in order to strengthen, validate and reshape the composition of the working propositions developed by theory (Wilson et al., 2002). This has let to the following three working propositions (the order of this list does not give any information on the relative importance):

1. Different elements of the Davenport DELTA model have different factors moving

organizations forward from the one stage to another. Each element shows different relevance regarding the organizational level of maturity.

2. Different layers of the organization have different perceptions of their organizational

level of data driven maturity. Therefore, the layers demand different types of change management approaches to successfully transform towards a data-driven organization.

3. The structure of analytics in the organization impacts the implementation and the

degree of success of each DELTA element during the organizational change towards a data-driven transformation.

(24)

These working propositions are tested by case studies, providing a possible indication of which elements are most strongly influencing data-driven maturity.

3.2 Sample

This qualitative multiple case study format will allow exploring in real-life from multiple perspectives the unique and complex process of digital transformation (Simons, 2009). The multiple cases create more robust outcomes on the propositions because of varied empirical evidence. The reason for conducting descriptive case studies is the focus of answering “what” and “how” questions, rather than the why, to explore the so far no clear single set of outcomes (Yin, 2009). It allows exploring differences or similar results between cases, and aims to find replications across cases to predict future results by analytical generalization. The qualitative method allows gathering additional information in more detail, and to make small iterations based on revealing patterns between propositions and data.

3.2.1 Cases

The combination of literature and extensive experience of the researcher in the field ensures that underlying propositions are rich enough to build rival explanations with six case replications. The cases are chosen through information-oriented selection. To obtain information about various circumstances in different stages of data-driven maturity, the sample includes three low analytical mature and three high analytical mature cases to facilitate comparison (Berg-Schlosser, 2009). The selected cases are cross-industry, with no further intention to show industry differences. There are no names provided due to privacy concerns expressed by the participating organizations. In figure 4, an overview is provided. The selection of multiple cases strengthens the generalizability and replication logic for general findings. Though it has to be noted that the selected organizations are active in different industries, implying that direct comparisons should be deliberately taken care of. By

(25)

Figure 4: Overview of cases sample

Case Sector Case Sector

A Insurance D Insurance

B FMCG E Logistics

C FMCG F Automotive

Figure 5: Overview of unit of analysis

Function Level Business Unit Years in organization

i. C-level Board CEO, CIO 4 year – 15 years

ii. Middle-manager

Head of Business Unit

Sales & Marketing, Business Intelligence, IT

2 year – 18 years

iii. Data analyst

Team-member Sales & Marketing, Analytics Business Intelligence, IT

6 months – 4 years

3.2.2 Unit of analysis

In total 18 semi-structured interviews with three individual employees at six companies are conducted. The unit of analysis consists of three individual employees working at different layers of the organization: one c-level manager (i) one middle manager (ii) and one data analyst (iii). The C-level managers are the highest-level of executives, having titles starting with ‘chief’. In this research it can be the CEO (chief executive officer) or the CIO (chief information officer). The selected middle managers are in charge of the decision making team in one of the departments within the organization, however not involved in decisions affecting

(26)

the whole organization. The data analyst in this research is one of the employees who uses data, statistical and quantitative analysis and models, to drive his or her daily actions on fact-based decision making. The different levels of analysis allow finding patterns of evidence across different layers within and across cases. Figure 5 shows an overview of the unit of analysis.

3.3 Procedure

To frame the levels of successful transformation in the best possible way, the maturity level is measured on a 1-5 scale along five dimensions of the Data Driven Maturity curve of Davenport (Davenport et al., 2010).

The case study consists of executing three separate actions. The first action took place one week before the start of each interview. Namely, participants were asked to place his or her organization on the right position on the maturity curve for all DELTA dimensions. They received the DELTA model (figure 2, p. 18) seven days before the interview date by email, and were asked to return the filled-in model two days before the start of the interview. This helped the researcher to analyze the indicated stages of maturity beforehand, to enable a more thorough analysis during the actual interview.

The interviews were the second action. During the interviews further analysis of factors took place by in-depth conversations to get a deeper understanding on what moves organizations to their current level of data-driven maturity. The employees are asked for their opinions on what determined their success so far, the challenges they face moving to next stages and to name activities and examples related to these drivers and challenges. Two main questions are asked for each DELTA element with the davenport model as talking piece:

1. Success factors: Which drivers moved your organization to your current level?

2. Challenges: What are the challenges your organization is currently facing in moving forward to the next stage?

(27)

To ensure that from every interviewee the same information from the general topics is collected, a standard interview protocol is used in order to ensure consistency and reliability between the interviews (appendix 3). The interview questions are concentrated on historical events. Both on single and general-case level: asked to individual employees and to compare findings across multiple employees. Questions cover all working propositions, with the ability to go more in detail on specific relevant topics. To maintain an exploratory aspect of the research, emerging side-topics during the interview were allowed and further elaborated on. The interview is split into the five dimensions, all discussed proportionally in the 40 minutes. Furthermore, for each DELTA element four to five statements are used as guidance to provide a certain focus throughout the conversations (appendix 4). They were defined before the interview started, after the investigation of literature in combination with the expert interviews. The sole purpose was to ensure a certain degree of consistency and relevance on the topics discussed. These statements are not measured on scale during the interview, neither shown to the interviewees. This maintained the adaptability and degree of freedom to get relevant information from the interviewees.

The third action was direct observation at the organization before and after the interviews. Observations take place at several departments by making notes on elements of culture, type of data analysis and the way analytics are structured within the organization. These multiple independent sources of evidence allow viewing the problem from different directions and to go beyond the initial impressions. These observational notes corroborated evidence from the interviews, by which inconsistencies are solved by deeper probing of the real meaning behind the findings of the interviews. The multiple research method ensured validity through triangulation.

(28)

3.4 Analysis

All interviews are recorded, to supplement the notes for adequate records. Relevant parts of interviews are transcribed and systematically coded along the 5 Davenport DELTA dimensions. Thereafter, the pertinent comments are grouped together per level of analysis and per DELTA Element (see results tables 1,2 and 3).

For analyzing the results, explanation building is used by a set of causal links on why or how something has happened (Yin, 2009; Miles, 1994). Initial predictions are compared against the case evidence, and this process is being repeated until a satisfactory match was obtained. To prevent leaping conclusions based on limited data of only a few cases (Kahneman & Tversky, 1973), the transcription and analysis of the data took place in two rounds. Transcribing the recordings and the first round of within-case analysis was executed immediately after all interviews. The simultaneous process of data collection (interviews and field notes) and within-case analysis helped to gain familiarity with the data and to accelerate the speed of cross-case analysis in later phases. Furthermore, it revealed insights for possible adjustments to the data collection (Eisenhardt, 1989). In order to prevent confirmation bias and to ensure the maintenance of objectivity, building conclusions by cross-cases analysis took place in the second phase of analysis after conducting all case studies. This limited the biases towards confirming the researcher’s preconceived notions (Patton, 2001), through objectively addressing rival explanations and potential counter arguments. The cross-case analysis took place for all three working propositions to investigate the within- and intergroup differences and similarities. The data analysis is executed along the five elements of the Davenport DELTA, with regards to the three different levels of analysis and in relation to the four types of organizational structures for analytics.

(29)

Results

Figure 6 provides an overview of the cases with their indicated level of maturity and analytics structure. All interviewees indicated their organizations on different elements at stage two, three, four or five. The total performance level is calculated by adding the indicated scores for the five elements from all three employees of the organization. Next, it is divided by 15 to get the average performance (three employees times five elements to score). If the total average of all 5 elements was at stage 2 or 3, it is referred to as “low” and at stage 4 or 5 is referred to as “high”. The analytics structure for each case study is determined through direct observation and through conversations with employees.

This research has multiple levels of analysis at each single case study. Therefore, a sample of the relevant comments with the accompanied case is categorized into: c-level manager (table 1); middle manager (table 2) and data analyst (table 3).

Figure 6: Overview of the cases and their industry, maturity level and analytics structure

Case Sector Maturity Level Analytics Structure

A Insurance Low Functional

B FMCG Low Decentralized

C FMCG Low Decentralized

D Insurance High Centralized

E Logistics High Center of Excellence

(30)

Table 1- Sample comments of c-level manager Sample comments

Data “When external parties require you to open up your data, it becomes more important that there is a strong data government” (Case D)

“In this early stage, there is limited time and resources available to invest in a central data governance system” (Case C)

Enterprise “The best approach is to start with the department where the ROI on using data analytics to solve the business case has the biggest change of success, to let this success story go through the organization like oil” (Case A)

Leadership “Middle-management often makes decisions based on gut feeling due to their experience, decreasing the amount of fact-based decision making” (Case E) “Urgency is often felt outside the company during trainings, summits or business trips to places like Harvard, Silicon Valley or data academies” (Case D)

“The change in behavior of ‘older managers’ to become more data driven is best reached by showcasing successes of other internal analytical activities so that their competitive instinct is making them willing to change their

behavior” (Case F)

Targets “Using data analytics itself is not a goal, but a means through which we can realize a winning game” (Case C)

“Analytics will support the firm’s distinctive capabilities more if the solved business questions proof profit and usability” (Case B)

“The biggest hurdle to increase speed is time and money made available by division budgets” (Case F)

“We are not directly rewarding our employees on their analytical capabilities, though we realize this is a necessary metric to push ourselves forward” (Case E)

Analysts “Asking open questions instead of closed questions is what managers should learn to fully take the benefits of analysts. Only then the analysts will have the position in the company they deserve” (Case E)

“First you need the mindset, then you can focus on analytical knowledge and skills” (Case A)

(31)

Table 2- Sample comments of middle manager Sample comments

Data “External pressure in the industry by startups and disrupters significantly increased the speed by which data analytics are used” (Case F)

“I’m not so sure what the added value would be of a central data team, we are currently doing quite well without” (Case B)

Enterprise “Being open for new trends and innovations happening in the world, keeps you up to speed on adopting new technologies and tools to become more data-driven” (Case E)

“To encourage more cross-division sharing of data practices, you need two things. That are hiring the right people who have a cooperating attitude & facilitating the process on enterprise level by providing the right tools” (Case F)

Leadership “Hierarchical organizations are slowing down the organizational change due to the risk-averse attitude of middle level managers who feel safe in their current type of structure” (Case D)

“Data-driven organizational change starts top-down, where strong passion, skills and support of leaders is needed” (Case A)

“It is a challenge for us to keep the balance between the entrepreneurial mindset of the company while thinking and acting fact-based” (Case C) “C-level leaders who are passionate about analytics also need the goodwill from other leaders at lower levels, before they can make any changes in the culture” (Case D)

Targets “We are used to knowing what is going to happen, based on our experiences. But to be open for all insecure consequences of in-depth data analysis, admitting you also don’t know the answer already, requires a vulnerable attitude; something we are not used to. And honestly, maybe not want to” (Case C)

“Even though managers see the benefits of analytics when e.g. a new tool or analysis is presented, when it comes to paying for it, everybody takes a step back and puts it ‘on hold for later” (Case F)

Analysts “Data analysts and business people are speaking two completely different languages and lack of mutual understanding limits achieving full potentials” (Case D)

(32)

should be entrepreneurial, be flexible, creative and have analytical skills. Then together they will be able to bridge the gap” (Case F)

“Data analysts need to celebrate their successes more to earn credits and create good-will from the business people” (Case E)

Table 3- Sample comments of data analyst Sample comments

Data “We rather have too much data of which we don’t know what exactly to do with it, than not having access to data at all” (Case B)

“Using external data which we scrape is sometimes easier because of greater freedom to use, store and analyze it. Internal data needs more governance and structure which takes a lot of time” (Case D)

Enterprise “People at other divisions first have to become aware of our existence, thereafter realize our impact on their profit and business return. Only then they will ask us for help throughout the company” (Case F)

“Before we can work on enterprise level, we first need to train other employees to read and use our analysis” (Case A)

“To enlarge the strategic importance of analytics, the central board should open budgets for tools like cloud storage” (case B)

Leadership “You need a visionary leader who is able to transform the business into a digital organizations and who get’s all employees on board” (Case A) “Most digital innovation comes from top-down, though leaders these days should always be open for bottom-up approaches from people close to the clients and support these initiatives” (Case C)

“Every board should have a digital DNA at a board seat so at least someone really understands what is going on in the world” (Case B)

“A risk-rewarding culture would increase the speed of becoming data-driven” (Case D)

Targets “Within the data lab we are encouraged to spend our time 50/50 on the data analytics and half freely at the ‘lab’ part, which helps us to constantly innovate” (Case F)

Analysts “Some divisions are too different in their basics to make any benefit of aligning all data, though for different brands there is huge opportunity for

(33)

complementary values” (Case E)

“It would help if we would have no need to sell our working hours like consultants internally to other divisions” (Case F)

“Constant sharing of best practices improves our work” (case D)

“The big amount of training and data academies facilitated centrally, improve our work on de-central level as well” (Case E)

Discussion

This section elaborates on the cross-case analysis of the results. It discusses the findings for each working proposition as defined at the research design. First, in section 5.1 different factors to increase the maturity level of organizations on the different stages of maturity are discussed for each element separately. Here, the data from c-level managers, middle managers and data analysts is merged to maintain a greater emphasis on the different findings per DELTA element. Second, in section 5.2 the perceptions of the different layers in the organization are elaborated on. Lastly, the impact of the different organization structures for analytics and its corresponding impact on the data driven transformation is discussed in section 5.3.

5.1 Different factors for different stages and elements

Warren (2014) found in his research, change takes place on continuous base with no end or beginning. It is not following a simple predictable path, but it is a non-linear process. There is no need to move equally on all elements to more mature stages, but it is recommended to give each DELTA element in roughly an equal amount of attention. Due to the mutual interdependence of various elements, frustrations will occur when some elements are lagging behind.

(34)

i. Data

The challenge for low mature organizations is that they often have huge amounts of data, but do not know how to analyze and translate it into useful applications (Case B). Meanwhile they often feel triggered by low storage costs and therefore collect as much data as possible. Though, solely collecting data for the sense of collecting data makes no sense. Organizations need to carefully consider how to store the data and how to create value out of it for the whole enterprise. Organizations that already started to clean their data at their most important domains during early stages of maturity, profit from higher data quality in later stages (Case E, F). Emphasizing small progresses in this stage is important to enthuse the rest of the organization (Case A). Moving forward means building an enterprise warehouse or cloud storage solution that is solving issues of data quality and reciprocity. Highly mature organizations realize that investment in a single data governance system is a necessary step in moving forward. Leading to a situation where employees can access all data through self-service and start delivering value themselves. These organizations see using data as a prerequisite to become digitally transformed, rather than a goal itself. To move forward, high mature organizations constantly put effort in combining the available data in innovative ways so they remain competitive. These organizations do this by implementing external sources (e.g. open data) and combining it with unique internal data gathered through distinctive measurements (Case F). Interestingly, highly mature organizations argue that they make less use of and advanced techniques, compared to statements of low mature organizations. This is explained by the fact that more mature organizations simply have broader knowledge of all the technologies available, making them more critical when judging their own use. Namely, they know there is still a broad range of potential available technologies they are not using yet. So, even though organizations perceive not to be on top of the edge, their employees actually are constantly exploring more advanced technologies and tools to implement. In

(35)

contrast, organizations stating they use all advanced technologies, are at lower stages of maturity and using less advanced tools, because they even do not have the knowledge on what is accessible and therefore perceive to use everything that is available. By lacking the more advanced tools, they mainly use data for extrapolation, the so-called danger zone where naïve predictions are made based on information rather than insights (Davenport, 2010).

The biggest success factor to become highly mature is changing the mindset of employees, so that they are willing to train themselves (Case E). It starts with senior commitment to recognize the potential of data-science; from there this sense of urgency should gradually spread to the rest of the employees beginning to show commitment. This is found to be present at all levels of data-driven maturity (Case A, C, D & F). Data should be on top of mind by all employees and should be used as the basis of their decision-making (Case B). Only then all employees will take great care of the data governance.

ii. Enterprise

For the investigated companies a certain level of enterprise-wide data-driven maturity is desired. However, successful organizations start by carrying out small projects with the highest probability of success, lowest amount and scale of risks involved and with activities that are replicable to other units. Low mature organizations established success by showcasing financial relevance of the business questions solved through using data (Case A, B, C). Moving up means starting with the next department where the ROI on data analytics can be achieved fast, after which the best practices can be implemented at other departments (Case D, E). Highly mature organizations always focus on the real business impact from data analytics in every project. In that way a snowball effect is created by validation of first actions by proved successes (Case F). Consequently, sharing the successes and accomplishments is a major success factor in moving forward. Because it gradually spreads through the company, managers of other departments see the success and feel the urge to be part of it (Case E).

(36)

A clear road map is used in all researched cases to determine the priorities for investment of resources on an enterprise level. In this way, employees can better visualize the new reality on all relevant domains and their feelings of rejection are diminished (Shamir, 1998). Organizations that have the ambition to quickly move forward on their data-driven maturity level think ahead in their strategy. For low mature organizations (Case A), this is shown through their investment in new tools and technologies on enterprise level already at early stages. In this way the basis for a common data language and practice are created throughout the entire enterprise. This increases the speed by which organizations can scale towards enterprise-wide analytics in later stages.

Highly mature organizations go even further. In addition to facilitating necessary tools and resources for either central warehouses or cloud storage solutions, the organization is centrally facilitating the sharing process, by organizing cross-departmental meetings and communication channels (Case D, E, F). Furthermore, the Data Analytics departments of highly mature organizations actively cooperate with the Human Resource department. (Case F). This helps recruiting employees for all departments with a cooperative attitude, analytical skills and broad sense of creativity, helping to fully capture the benefits of data science on enterprise level (Case F).

Central responsibility is a key differentiator in moving forward to a more enterprise-oriented view (Case D, E, F and to some extent Case A). Day-to-day operations always diminish the available time to implement new habits on enterprise level (Case B, C). Therefore, this research has shown that when a person or team is centrally responsible, a data-driven culture throughout the entire enterprise is faster established.

iii. Leadership

Data-driven awareness and passion showcased by the leaders is a prerequisite to successfully transform at all stages of the maturity curve. Bass (1985) found that leadership is the most

(37)

deciding factor for transformational success. Although in different degrees of intensity, leaders of all cases showed visionary leadership and passion for analytics. This is valuable, since the analytical leader needs to be visionary, involving the creation of future-oriented images on the work-floor to shape the behavior of the employees (Shamir, 1998; Stam, Knippenberg and Wisse, 2010). Organizations need a leader who understands the importance of digital transformation. Specifically, the extent to which the leader acknowledges and actively emphasizes the importance of analytics, is impacting the general level of analytical knowledge among employees at other levels. Namely, this analytical leadership leads to increased awareness at the employees which are thereafter more likely to develop analytical skills. Differences in this aspect are both found between cases showing low maturity and between cases showing high maturity (Case A vs. Case C; Case D vs. Case F). Important is that top-management critically judges themselves whether they really possess the skills, because in some cases (independent from maturity level) they perceive themselves to be more skilled than colleagues judge them to be. To move upward on maturity, leaders should constantly invest time and money in learning new necessary skills, keeping themselves up to date on all what is available. Leaders of high mature organizations look beyond internal sources for educational purposes and realize that constant learning is key. Examples are data science academies (Case E), business trips to places like Silicon Valley (Case A) or through online education (Case D).

Next to developing understanding of data analytics, it is the role of the leader to create a sense of urgency through identifying and sharing his or her vision of becoming a data-driven organization. He or she should have the skills to make the employees follow his/her direction. The drive and passion shown by the leaders has an enormous impact on the positive feelings of the employees involved in the change, leading to faster transformation (Case A, B, D and

The success factors of a digital transformation towards data driven maturity