Designing big data-enabled organizations
A study on how organizational design affects value creation in the Telecom industry.A
Author: João Laranjeira Student number: 10394443 Supervisor: Dr. Ranjita M. Singh Second Supervisor: Dr. Arno Kourula University of Amsterdam
2
Preface
This thesis explores the impact of organizational design in value creation within the Telecom industry.
The topic of this thesis stems from my personal interest in two topics: big data and organizational design. Both topics relate to my study field, namely strategic management. My choice of selecting the Telecom industry as a case study was facilitated due to my close connection with some industry experts.
Therefore I would like to thank all the participants for their inputs in this thesis in particular to my friend Rui Ferreira and to my thesis supervisor Dr. Ranjita Singh. To Rui, who has inspired me in identifying an adequate research topic and shared his insights and experience with me on a daily basis. To Dr. Ranjita, who has helped me throughout the thesis in achieving rigor and relevance.
3
Abstract
Big data is predicted to become a driver of competitive advantage for organizations, underpinning new waves of innovation, productivity and growth. Recent research elaborates on the benefits of big data. However, little research has been done in exploring how the organizational design influences the organization's ability to create value from big data. This thesis explores this research gap and takes an exploratory approach on how value is created from big data in the Telecom industry. It uses organizational design and value creation theory to uncover the impact the organizational structure and processes have in the value created from big data. By combining both theories, this thesis provides considerations on how different organizational design impact value creation and how they affect the organization's capabilities to explore big data.
Big data brings new opportunities to organizations to out-compete existing players by creating new sources of competitive advantage. Yet, to grasp these opportunities the design of the organization has to support value creation of big data.
4
PREFACE ___________________________________________________________________ 2ABSTRACT __________________________________________________________________ 3
1. INTRODUCTION ___________________________________________________________ 6 2. RESEARCH BACKGROUND ________________________________________________ 7 2.1THE RISE OF BIG DATA ______________________________________________________ 7
3. RESEARCH MOTIVATION AND RESEARCH QUESTION ______________________ 9
4. BIG DATA AND BIG DATA ANALYTICS ____________________________________ 11
4.1WHAT IS BIG DATA? _______________________________________________________ 11 4.2WHAT IS BIG DATA ANALYTICS? _____________________________________________ 12
5. THEORY REVIEW ________________________________________________________ 13
5.1REVIEWS ON ORGANIZATIONAL DESIGN THEORY ________________________________ 13
5.2REVIEWS ON VALUE CREATION THEORY _______________________________________ 14
6. RESEARCH FRAMEWORK ________________________________________________ 15
7. HYPOTHESES ____________________________________________________________ 16
7.1STRUCTURE ______________________________________________________________ 16
7.2PROCESSES ______________________________________________________________ 18
8. RESEARCH DESIGN AND METHODOLOGY ________________________________ 20
8.1CASE STUDY APPROACH ___________________________________________________ 20
8.2RESEARCH SAMPLE _______________________________________________________ 20
9. DATA COLLECTION AND ANALYSIS ______________________________________ 21
9.1METHOD OF ANALYSIS _____________________________________________________ 21
10. ANALYSIS AND RESULTS _______________________________________________ 22 10.1VALUE CREATION ________________________________________________________ 22
5
10.3DEVELOPMENT OF CAPABILITIES ____________________________________________ 27 10.4STRUCTURE _____________________________________________________________ 28 10.4.1 JOB SPECIALIZATION ...28 10.4.2 CENTRALIZATION ...29 10.4.3 DEPARTMENTALIZATION ...31 10.5PROCESSES _____________________________________________________________ 32 10.5.1 DATA GOVERNANCE ...32 10.5.2 INFORMATION SHARING MECHANISMS ...32 10.5.3 KNOWLEDGE RETAINING MECHANISMS ...3311. DISCUSSION ____________________________________________________________ 33
11.1IMPACT OF ORGANIZATIONAL STRUCTURE ____________________________________ 33
11.2IMPACT OF ORGANIZATIONAL PROCESSES ____________________________________ 35
11.3FORMS OF COOPERATION _________________________________________________ 37
11.4THEORETICAL AND MANAGERIAL IMPLICATIONS _______________________________ 37
11.5LIMITATIONS ____________________________________________________________ 38
12. CONCLUSION ___________________________________________________________ 39
13. REFERENCES __________________________________________________________ 41
14. APPENDIX ______________________________________________________________ 43
14.1OTHER FIGURES AND TABLES ______________________________________________ 43
14.2QUOTES ________________________________________________________________ 46
6
1. Introduction
Telefónica, one of the biggest mobile network operators (MNOs) operating in Europe, has recently formed a new global business unit aimed to monetize big data under the name of Telefónica Dynamic Insights (TDI).
TDI's first product, Smart Steps, provides retailers and other third parties location based insights and other visual charts of mobile users movements throughout the day. Telefónica's Smart Steps is helping retailers in understanding their consumer movements, their demographics and their preferences. By relying on those insights, retailers are able to target consumers more precisely and consequently creating more cost efficient marketing campaigns.
There are other components involved in the creation of new revenues. Deutsche Telekom (DT) for example, has created an innovative service by leveraging their customer location based data. The mobile operator sends an SMS to clients arriving at designated ski resorts, informing them about simple insurance packages suitable for their vacation. This innovative insurance service is based on improved client segmentations and risk profiling also derived from customer data. By better segmenting its customer base, DT is able to calculate different service premiums.
The value for the customer is clear, by following small and easy steps, the customer can use his mobile phone to buy an insurance service on spot with more comfort and at lower cost.
7
2. Research background
In this section the research background is separated into two parts. The first describes the rise of big data as a management and academic topic. The second focuses on describing the motivation for future research on big data, this section also includes the research question of this thesis.
2.1 The rise of big data
Big data became one of the trendiest management topics in vogue 1. The proof that big data is helping organizations to improve their current operations and creating new business opportunities has led to the exploration of big data by different industries. At the same time, many well known strategic consulting companies have become highly involved in providing big data solutions and in helping their clients in implementing the necessary organizational changes to explore big data.
As Biesort et al. (2013) mentioned "The payoff from joining the big-data and advanced-analytics management revolution is no longer in doubt."2. More recent research on organizations exploring big data also highlights the benefits it brings to organizations. Barton and Court (2012) for example, found evidence that organizations which explore big data and inject big data analytics in their business show productivity rates and profitability that are 5% to 6% higher than those of their competitors. Other recent articles suggest that organizations using data analytics to support decision making are more productive and experience higher returns on equity than competitors that do not use these analytics (Brynjolfsson et al., 2011).
Figure 2: Visualization of data and data value Source: “Big data: The next frontier for innovation, competition, and productivity”, McKinsey Global Institute, May 2011.
1 http://www.nytimes.com/2012/08/12/business/how-big-data-became-so-big-unboxed.html?pagewanted=all&_r=0 2 http://www.mckinsey.com/insights/business_technology/big_data_whats_your_plan
8
Despite the perceived benefits, developing the right big data analytic capabilities is far from easy. The necessary investments to put these capabilities in place can be expensive and the disruptive organizational changes difficult to implement. As Carlos Pinto, Principal at Analysys Mason commented in response to the news about Telefónica Dynamic Insights3: “The organizational challenge is one of the most important that Telefónica will face. Big Data is a completely different type of product to what they are used to and it needs to be packaged in a different way”.Recent literature on the topic also suggests that big data challenges lie within the organization. Brown et al (2011) remarked that radical customizations and new business models will become the new hallmarks of competition as organizations explore big data. According to the authors, big data will become a new strategic asset which will be explored across business units and functions, representing a key basis of competition between companies. Defending the same idea, Manyika et al. (2011) highlight important challenges that organizations need to address in order to capture all the benefits of big data. According to them, organizations will need to have the right people and process in place to capture value from the use of big data.
The lack of scientific research on the topic can be explained by its novelty. Whereas previous buzz research topics such as data mining have been exhaustively explored, big data remains a relatively new outlook in the field of business management, as depicted in the table below. Hence studying the impact of organizational design is important to the study field.
Figure 3: Different academic fields researching big data. Source: Halevi and Moed (2012),page 4.
The research of big data is increasing within the academic community due to its interest across distinctive academic fields. Evidence of that is the exponential increase of published articles since 2000.
3
http://www.eurocomms.com/features/analysis/8593-telefonica-breaks-cover-as-first-big-data-telco
9
Figure 4: Time line of Big Data as topic of research. The dotted line represents the exponential growth curve best fitting the data represented by the blue bars. This shows the number of Big Data articles increasing faster than the best exponential fit. Source: Halevi and Moed (2012), page 4.3. Research motivation and research question
While most of the literature was initially focused on understanding how big data can be beneficial for organizations, it is still not clear how organizational design affects big data value creation.
Brown et al. (2011) and Manyika et al. (2011) suggest that radical customizations and organizational changes will be required to explore big data. The authors mention the way companies are presently organized as a potential obstacle to the exploitation of big data. For instance, for most companies a huge part of data accumulated traditionally lies within departmental "silos", which can slow the process of exploiting data and difficult data sharing, ultimately affecting an organization's ability to create value with it. Brown et al. (2011) also foresee " complications" faced by all organizations exploring big data. Among them are issues such as shortage of big data specialists and threats concerning privacy and security around big data. All these issues are related to organizational design, as it will be explained in the next chapters.
According to the authors, organizations should consider whether they are organizationally prepared to exploit big data and able to overcome its threats. But how can they do it? And most important, what is important to change in their organizations in order to do it? From the above, it seems important to study the impact of organizational design in value creation. Hence, this paper aims to explore this research gap and contribute to the emerging literature of big data in the business management field.
10
Industries such as telecom, utilities and banking have been pioneers in exploring big data. Though every industry is virtually capable of exploring big data, some have special characteristics that make them able to monetize it. For the industries mentioned above the most obvious characteristics are: the high number of customers per company, the different customer data sources such as demographics or behavioral at their disposal, the high number of customer transactions. Telecom companies in particular have been dealing with analytics for a long time, given their extensive client base. Depending on the population of each country, MNOs can have 5, 10, or even 20 million customers each. As it is shown in figure 4, the telecom industry is the industry leading the race to monetize big data.By analyzing only one industry better comparisons can be made, due to the fact that all companies have similar business models and organizational designs.
The choice of the telecom industry is supported by a recent edition of the telecom magazine; Telecom Asia. It conducted a query directed to telecom operators around the region asking them to mention their vision about the industry's priorities for the forecasted year (telecom business outlook 2012). Unsurprisingly, big data was the most cited priority which the industry leaders expect to drive growth.
11
Figure 4: Industries leading monetization of big data. Source: "Big data global trend study", Tata consulting services (2013), page 43.The fact that mobile operators have been exploring data for a long time can enhance the probability to extract more meaningful findings from this research.
All of the above has led to the formulation of the following research question:
How does the organizational structure and processes influence value creation from big data?
4. Big data and big data analytics
This section elaborates on the concepts of big data and big data analytics. It aims to familiarize the reader with both expressions and provide a definition for both expressions based on the existing literature.
4.1 What is big data?
Figure 5: IBM 305 RAMAC 5MB hard drive being fork lifted into a plane, the HDD weighed over a ton.
The often called "digital era" revolutionized the way we handle and store data. The amount of data stored and computation capacity grew exponentially during the last decades. According to Manyika et al (2011), the amount of data stored worldwide grew from 50 Exabyte in 2000 to 300 Exabyte in 2007 (1 Exabyte = 1018 bytes), while the computation capacity more than duplicated (from 3 to more than 6 million instructions per second). At the same time, data storage became cheaper in parallel to the increase in HDD capacity. As an example, today it costs 600$ to buy a disk that can store all the world music. This evolution led organizations to start exploring data like they never had before.
Big data lacks a consensual definition, even though it is an often mentioned term.
Manyika et al (2011) for example, define big data as "datasets whose size is beyond the ability of typical database software tools to capture, store, manage, and analyze". Whereas Zikopolous
12
et al. (2012) claim big data to be composed by four main characteristics (presented below). Their definition is frequently cited and it is used for instance in the popular article "Big data: the management revolution" by McAfee and Brynjolfsson (2012).According to Zikopolous et al. (2012), big data can be disaggregated into four dimensions (also known as 4 V's): Volume, Velocity, Variety and Veracity. Subsequently, big data is composed of large, dynamic and complex datasets that cannot be entirely analyzed by the use of traditional software tools and analytic techniques within tolerable time frames and tools.
Volume can be understood as the large volume of available data in an organization. This is a key characteristic of big data. In telecom for instance, it allows operators to potentially analyze individual customer data, performing sub-segmentation and develop customized approaches. Variety refers to the various different data sources (CDR, data sessions, social networks, internal reports) and formats (alphanumeric, XML, audio, etc.).
Velocity can be understood as the frequency with which data is generated.
And finally, veracity or value of big data which refer to the quality or trustworthiness of the generated data. In order to extract valuable data, organizations need to set up the right methodologies and analytical frameworks, i.e., using tools that help handling big data's veracity: transforming data into trustworthy insights while discarding noise.
Below are presented the characteristics of big data in the telecom industry.
Figure 6: Big data in telecom industry.
4.2 What is big data analytics?
In this paper, big data analytics is defined as the capabilities in place in order to explore value in big data, i.e., the mechanisms used to uncover unknown correlations, hidden patterns and other useful insights. Big data analytics is not only about implementing IT infrastructures, but rather about the development of the right methodologies and analytical frameworks to take advantage of that infrastructure. The lT infrastructures are the tangible assets that serve as the basis of big data analytics. Examples are data warehouse and software tools such as Hadoop. Big data analytics is therefore the capability created to explore value in large, dynamic and complex data sets through the use of the organization's IT structure.
•Telecommunication services have typically very high penetration rates both in developed and underdeveloped markets
•Due to the relatively reduced number of players in each market, a typical telecom operator has a large % of the population in its customer base
•Large telecommunication groups have access to worldwide customer database Quantity of customers
•The average customer from a telecom operator generate data entries on a daily basis •Frequency will tend to increase with the widespread of internet services into a potentially a
continuous generation of data Quantity of data
•Telecom operators data includes different data dimensions including telecommunication patterns, location, devices used, content accessed, online transactions, demographics, etc •Growing services such as mobile payments, M2M and other services related to NFC will increase
further the diversity of data available Diversity of data
13
5. Theory review
This section aims to serve as the theoretical base of the research and to contextualize the reader with two common strands of business studies research: organizational design and value creation theory.
The earliest academic research on big data was mostly focused on uncovering the benefits derived from exploring data. Henceforward future research should explore how organizations should evolve in order to better explore big data.
This thesis aims to fill this gap. The first part of this section is focused on uncovering existing literature on organizational change and big data, at the same time on using organizational design theory as the base of the research framework that is presented below. In the second part, a prominent value creation theory, the resource based view, has been reviewed in order to understand how mobile operators can move to positions of competitive advantage.
5.1 Reviews on organizational design theory
Most authors explain organizational design as a concept where the structure is connected with other dimensions in the organization in order to create synergies among them. These systems are interconnected by the organization's vision and a common strategy.
Nadler & Tushman (1997) describe the organizational design as a system composed by the formal and informal organization, business processes and human resources. Similarly, Galbraith (2002) uses similar concepts to explain the organizational design. According to the author, organizational design is composed by five dimensions: strategy, reward systems, organizational structure, processes and lateral links, and human resources. Galbraith's definition is commonly cited in organizational design studies. His concept his more clear and of better use during this research. Therefore this concept will be used throughout the research due to the easy fit between it and the case study.
Galbraith describes organization design as an ongoing process of reshaping structures, processes, reward systems and people conduct. In order to create an effective organization focused on achieving its strategy.
In order to build the research framework, Galbraith's Star model has been adopted (figure 7).
The Star model provides a simple visualization of different dimensions that influence organizational behavior. By manipulating these dimensions, the organization can potentially change its desired behavior, ultimately affecting its performance as well as its culture.
14
Galbraith mentions that "a company's organization is the accumulation of all previous strategies and structures the company has adopted."In one of his latest papers, Jay Galbraith (2012) builds on Alfred Chandler (1962) work "Strategy and Structure" in order to explain the evolution of organizations along the last century. Galbraith builds on the idea that "structure follows strategy" and consequently develops his concatenation concept. Concatenation can be understood as the accumulation of simple strategies into increasingly more complex structures. This concept is proven along the last one and a half century of organizational structures evolution. Six major strategies were then identified by the two authors. During the nineteenth century, Chandler (1962) identified three growth strategies: volume expansion, geographic dispersion, and vertical integration, resulting in three common structures: U-form, M-form, and H-form. As Galbraith explains, subsequently to the international expansion era (mid-1900s), a fashion for a new structure design started to emerge: the matrix structure. Organizations changed their structure by embedding the geographical activities into the functions and business.
The last concatenation of organization design started during the 90's, as a result of a new customer-focused strategy. The premise for change came from the increasing customer buying power and the necessity for more product customization instead of stand-alone products. The result was the creation of platform structures that can be easily and quickly modified to meet the unique need of the customers.
Following the six major strategies which drove the evolution of organizations during the last one and a half century, the author questions which next major growth strategy will be in place during the present century. According to him, a rapidly emerging candidate is big data.: "Big data could very well be the next strategic emphasis of the future enterprise organization" (Galbraith, 2012). Galbraith mentions that companies are aggregating their independent databases, which are traditionally held by different departments or business units. In the big data era, the trend is to centralize these databases in the company. In the author's perspective, centralization allows organizations to run engines and algorithms in order to generate valuable insights from big data. Therefore, companies will increasingly change their organization design in order to explore this precious asset. Centralization has been mentioned has an important change. Yet, many other dimensions of organizational design might also impact value creation. Therefore, as it is shown below, centralization and other dimensions of organizational design will be explored throughout this research.
5.2 Reviews on value creation theory
In order to address the research question this sub-chapter will explore one particular theory of value creation: the resource-based view (RBV). The resource based view fits well with organizational design since it analyzes the resources available in the organization in order to explain value creation and positions of competitive advantage. During this research, value creation is defined as the process of creating new revenues and/or creating efficiencies by reducing the costs.
The RBV of the firm views the firm as a bundle of resources and capabilities. The value creation results from the unique combination of a bundle of complementary and specialized resources and capabilities. These strategic assets are heterogeneous within the industry, scarce, difficult to trade and imitate. The key point of the theory is to identify key potential resources, evaluate if they fulfill the VRIN criteria and ultimately protect them (valuable, rare, inimitable and non substitutable). According to the RBV, these criteria will create a source of competitive advantage which will lead to supra-normal performances of the firm.
A resource is valuable only if it leads to a firm's cost reduction or increase its revenues compared to an alternative situation where the firms does not possess those resources (Barney, 1997). In order to be valuable, the resource has to be rare too (Dierickx and Cool, 1989; Barney, 1986).
Inimitability is another necessary condition, i.e., if competitors are not able to duplicate these resources perfectly, then the firm might have a source of sustainable competitive advantage (Peteraf, 1993; Barney, 1986). The underlying premise of the inimitability is causal ambiguity, i.e., the resource is unknown to others, thus non imitable. If the competitive advantage stems
15
from knowledge-based resources than the effect is escalated, since this type of resources are more likely to be idiosyncratic to the firm who deploys them (Peteraf, 1993).The last criterion is substitutability. Even if the resources are valuable, rare and non-imitable (Dierickx and Cool, 1989), each criterion alone is not an enough condition to be a source of sustainable competitive advantage. Instead, resources have to simultaneously fulfill all the VRIN criteria.
Big data and big data analytics can be understood as part of a new set of strategic assets organizations have to deploy ("Over time, we believe big data may well become a new type of corporate asset that will cut across business units and function much as a powerful brand does, representing a key basis for competition" - Brown, Chui, Manyika, 2011), i.e., they are respectively a resource and a capability that potentially can lead to a firm competitive advantage.
Using organizational design theory together with RBV, the Galbraith's five organizational dimensions can be measured against their capacity to create value from big data and to lead to positions of competitive advantage. For instance, this research will explore how structure leads to value creation from big data and ultimately how it can lead to positions of competitive advantage.
Big data has been a new source of value creation for some MNOs, with it mobile operators are increasing efficiency in their operations and are capturing new business by providing big data insights. For instance, big data is being used to tackle a common plague common to all mobile operators: customer turnover, also known as, churn. In the telecom industry, the loyalty level is particularly low when compared to other industries, where customers move from one provider to another relatively faster than other industries. Therefore, reducing churn rates became a strategic priority for most mobile operators. Big data is a resource some are using in order to detect churn patterns. With the right capabilities, churn can be predicted in advance and measures taken to retain the customer. From the above it is easy to understand that if combating churn represents a major strategic priority for mobile operators, then to big data must be given equal importance since it can be an important strategic asset to fight churn. The questions that arises is whether big data can lead to positions of competitive advantage. It would be wrong to assume that big data alone can lead to positions of competitive advantage. It doesn't matter how much quantity or variety of data a mobile operator has if he is not able to extract value from it. Big data as just data has no value and cannot lead to positions of competitive advantage. This means that an operator with five million customers, does not have a relative advantage to others with inferior numbers of customers, if he is not able to extract value from it.
Therefore this research will use the resource based view to explore the resources and capabilities within the organizational design that lead to value creation and competitive advantage. In order to do it, it will access whether organizational design can enable value creation from big data.
6. Research framework
Building on Galbraith's conceptualization of organizational design, the Star model, the following framework was developed in order to address the research question.
16
Figure 8: Theoretical framework.This thesis explores the relation between some of those dimensions and the organization's ability to create value with big data, both internally and externally. Internally, it means all the value creation outcomes that improve the organizational efficiency, such as OPex and CAPex reduction, customer service improvement, churn reduction, etc. This side of value creation concerns only the MNO's traditional business model (e.g.selling voice calls). Externally, it refers to all the new revenue sources created from new business models (e.g. Selling insights based on big data), this side of value creation is also known as monetization of big data.
Due to time limitations this paper will solely focus on the structure and processes dimensions, since it would not be possible to gain a reasonable level of depth if all dimensions were to be studied within the time limit.
The choice of these two dimensions, was based on the relatively higher level of discussion around topics related to the structure and processes dimensions. These topics are discussed below.
7. Hypotheses
In the theoretical framework presented above, several dimensions have been drawn, which are the exact organizational dimensions Galbraith (2012) mentions in his Star Model. Galbraith (2012) divides the structural dimension into specialization, distribution of power (centralization) and departmentalization and the processes dimension into vertical and horizontal processes. The proposed conceptual framework presented below, aims to explore the impact of each one of these sub-dimensions in creating value with big data. Some sub-dimensions have been adapted in order to fit into the big data topic.
7.1 Structure
Specialization refers to the type and number of job specialties used to perform the work (Galbraith, 2012). In this paper, the concept was adapted. Hence, job specialization refers to the type of specialized personnel necessary to explore big data. Galbraith mentions that new job specialization affects the organization structure by creating new interdependencies between
17
employees. According to him, big data brings changes in the way these interdependencies are managed.According to Davenport and Patil (2012), organizations need specialized workers (data scientists) to develop their big data analytics. The authors argue that a new generation of data scientists is necessary for organizations to implement big data analytics and to explore big data. The authors claim that many of the skills and knowledge necessary to crack big data are rarely taught in traditional statistic university education and therefore most of the statisticians are not yet prepared for it. They go further by saying that these people are extremely hard to find and will be in short supply as the demand continues to increase. Thus it makes these data scientists rare.
However, if these scientists are necessary to exploit value from big data, so too are the necessary analytic tools and technology in place to analyze it (McAfee and Brynjolfsson, 2012). According to the authors technology is a necessary component to implement big data analytics and ultimately improve firm performance. These tools became relatively cheap and widely available. As an example, Hadoop, one of the most used and well know big data analytical program is an open-source software. From the RBV perspective, these tools are not the most strategically important assets since they do not fit the VRIN criteria (they are widely available, easy to copy and imitable). It means that owning them will not provide a source of competitive advantage. Instead, data scientists seem to be an important strategic asset for companies exploring big data and important enablers of value creation. This hypothesis will explore the role of data scientists in developing big data capabilities and their importance for value creation from big data.
Nevertheless, data scientists might not be the only new job specialists an organization needs in order to create value with data. As it was mentioned above, Carlos Pinto mentioned that big data also needs to be packed and sold to new clients. This suggests that other job specialists might be necessary in order to create value from big data. Hence this hypothesis will explore in which circumstances other job specialists are necessary in order to create value with big data.
H1: Data scientists and other data related job specializations are necessary to enable value creation using big data
Distribution of power or centralization is also incorporated in Galbraith's (2012) framework. Centralization can be understood has the vertical distribution of power within an organization. This hypothesis aims to explore how centralization of big data analytics enables value creation from big data.
Within the big data world, centralization became a frequent topic of discussion. There are divergent opinions on whether organizations should centralize big data analytics or not. Big data analytics can bounce between two level ends of centralization. In the most centralized manner, analytics are located and owned by a separate unit, at the other end big analytics are separated into silos across different organization departments and/or business units. In a recent study released by Tata consulting services4, it was found that centralizing big data analytics in one separated unit is important to create big data capabilities. According to the study, the major challenge big data brings to the companies is getting different business units to share information across organizational silos. By centralizing analytics, the study showed companies could "preserve the data scientists independence", i.e., provide unbiased insights and advice and at the same time to provide a more attractive career path, ultimately important in order to retain these scarce data scientists.
Contrasting opinions5 mention centralization to carry several disadvantages and difficult to implement. For instance, centralizing data processing and databases can be very costly.
4 http://sites.tcs.com/big-data-study/big-data-skills/ 5 http://www.waterstechnology.com/inside-reference-data/news/2226828/ubs-exec-dont-centralize-processing-of-big-data
18
Despite the opposing opinion, literature evidences the benefits it can have for new projects such as big data. Koberg el at. (1996) found that a high level of centralization is beneficial for innovation in new ventures, since the higher levels of the organization have greater freedom to be assertive and commit resources (Miller, 1987). This perspective is coherent with Tushman and O'Reilly's (1997) position of ambidexterity. They defend that organizations should complement its existing business models through the creation of autonomous innovating units. Answering this hypothesis will help to understand the advantages and disadvantages of centralizing big data analytics and for instance explain the pros and cons of creating autonomous business units as Telefónica did with TDI. Accordingly, the second hypothesis is:H2: Centralization of big data analytics enables value creation using big data.
Departmentalization is the criteria used to form departments at each level of the structure (Galbraith, 2012). Typically departments are formed by functions, products, processes, geographically or by customer. However some organizations use matrix structures, where two or more dimensions report to the same entity within the same level. Departmentalization is an important driver that mostly depends on strategic decisions. For instance, with functional departmentalization the most recalled advantages are economies of scale and efficiency, whereas in product and geographical departmentalization, these are respectively higher flexibility and customer proximity. Studying the impact of departmentalization is important for the organizational design field.
As Galbraith (2012) suggests, big data can be the very next candidate in the concatenation process explained above. This also means that the departmentalization styles of companies using big data will change in order to better explore their databases and analytical capabilities. This hypothesis aims to explore the impact of different departmentalization styles in enabling value creation using big data.
H3: Departmentalization enables value creation using big data.
7.2 Processes
Galbraith (2012) mentions two main forms of processes within the company, horizontal and vertical. Vertical processes are those aimed to allocate resources and deploy talent. While the horizontal processes are related to the organization's workflow, such as product development, insights sharing, cross-learning, etc.
According to Galbraith (2012), big data will contribute to create tools that will manage the increasingly interdependencies in today's complex organizations. This interdependences respect to the number of vertical and horizontal process organizations have. Regarding big data, one of those important tools will be data governance, since it is used to manage those interdependencies by creating management conducts of data within the organization.
Firstly, in order to fully exploit the data, organizations must first make sense of it. Data as to be treated in the right fashion, thus organizations should seek the right governance modes in order to track and manage data. According to Zikopolous et al. (2012), "organizations that don’t plan for the governance of their big data systems from the start end up falling behind and must face significant additional costs to retrofit governance into their ecosystems.".
From a resource based view perspective, data governance seems to be an important resource to be deployed. Data governance can create mechanisms that facilitate the management of data within the company, therefore creating transparency on the use of data, improving data security. These mechanisms can enable value creation by enabling information sharing within the organization. Moreover, data governance can avoid dangerous data leaks that can severely damage the company business or reputation, which ultimately can have an impact on value by dropping revenues (Verizon and NSA example6).
6
http://www.channelpartnersonline.com/news/2013/06/nsa-verizon-scandal-to-have-a-major-business-impa.aspx
19
Hence, the fourth hypothesis is:H4: Data governance enables value creation using big data.
As it is mentioned in the first hypotheses, organizations must hire and retain data scientists in order to develop their analytical capabilities. It is a fact that without these capabilities, no organization is able to create value from big data. However, other capabilities are equally important in order to create broader big data competencies.
From a resource based perspective, organizational learning is a capability or resource that is important to deploy in order to build or maintain positions of competitive advantage (Smith et al. 1996). Organizations must set up the right learning mechanisms in order to create and develop their capabilities (Zollo and Winter, 2002), the experience must be accumulated and the resulting knowledge articulated and codified within the organization. This means that organizations must have mechanisms that allow cross-learning, but also processes where information is codified and shared across the organization. This hypothesis aims to explore how these mechanisms affect value creation with big data.
Therefore the last hypothesis is:
H5: Learning mechanisms and information sharing processes are enablers of value creation using big data.
20
8. Research design and methodology
This chapter describes the methods used in the research, the aim is to inform the reader about the methodology used, sample selection criteria, data collection and method of analysis.
In order to answer the research question, the research was made using an exploratory research design, thus qualitative methods have been used in data collection and data analysis. Due to the newness of the topic in academia, one could say that the most valuable sources of big data in telecom reside in specialists who have been working in the industry and got involved in big data projects. Therefore, the data collection method will include interviews with three different kinds of specialists in the telecom industry. The following sub-chapters will discuss the most important elements of the research design.
8.1 Case study approach Advantages
The research question aims to explore how the organizational structure and process affect mobile operators. According to Saunders et al. (2007), research methods that aim to answer the "how" and "why" question types can be better addressed by using a case study method.
Due to the newness of the topic, this research aims to find patters across different firms.
Using a case study strategy has the advantage to better understand the context involved and to allow the researcher to obtain deeper insights about the processes performed in each organization (Morris & Wood, 1991). Another advantage of using case studies stems from the non manipulation of the researcher in the studied events (Saunders et al., 2007; Yin, 2009). In order to better explore the research question, different firms have been studied (figure 27 in the appendix).
Disadvantages
Despite the advantages mentioned above, the case study approach has some commonly mentioned disadvantages which tend to affect the validity and reliability of the research. Two disadvantages have been found.
Qualitative research methods usually have a low level of codification and standardization of data compared to quantitative research methods (Eisenhardt & Graebner, 2007; Gibbert & Ruigrok, 2010). During this research this disadvantage was attenuated by drawing a clear research question and framework. Moreover, the independent and dependent variables were well defined and the connections between them made clear. This allowed me to create a clear coding scheme and consequently improve the level of standardization of the data collected. The second disadvantage is that environmental factors are not controlled, which according to Saunders et al. (2007) can affect the causal relationship between variables. Two steps were taken in order to attenuate this effect. First, this research presents in appendix a chain of actions applied during the research (Yin,2009). Second,during the data collection process unambiguous question were used during the interview. Following these recommendations, the research could overcome those initial challenges and increase the validity and reliability of the research.
8.2 Research Sample
Due to the newness of the topic, the method used in order to obtain a sample was self-selection sampling and snowball sampling (Saunders and Lewis, 2012), i.e., some members were hand-picked and most of them recommended by earlier interviewees due to their involvement with big data projects. The criteria used for sample selection was based on the working experience each candidate had in telecom industry and the level of involvement with projects related to big data. Moreover, the respondents were found to be working in three different sub-parts of the telecom industry: MNOs, telecom consulting firms, telecom associations. The advantages stemmed from these differences are measured in terms of a trade-off. MNOs employees yield more depth to the case study due to their knowledge about their company structure and process. Telecom
21
consultants and telecom association's employees on the other side, provide broader insights about the industry best practices.All the interviewees were based in Europe, Africa, Middle East and Asia.
Figure 27 in the appendix provides information about each respondent's current employer, job title and employer website.
9. Data collection and analysis
This section will present the data collection methods and methodology used in the analysis section. The appendix presents a chain of evidence (Figure 28) with the purpose of increasing reliability of the methodology used (Yin, 2009).
The data collection method chosen was semi-structured interviews and all the interviews have been conducted by using Skype® due to the geographical dispersion of the interviewees. Each one of the interviews took an average of 42 minutes and all have been recorded in audio files. Altogether 12 interviews have been administrated, however from the 10th interview onwards, data saturation was achieved, i.e., no more other insights have been found. From then onwards repetition was common and new data has not strongly contradicted the previous findings (Bryman, 2004).
Following the research framework presented above, the interview was composed of three main dimensions:
a) Section about value creation.
b) Section about the impact of structure on value creation. c) Section about the impact of processes on value creation. Note: the final Interview guideline can be found in the Appendix.
Two criteria to select each respondent was used. Firstly, each candidate was asked about his background experience in telecom industry, furthermore each of them had to have more than 3 years experience in the telecom industry. Secondly, each candidate was asked about his knowledge on the topics discussed, consequently each candidate had to have been directly or indirectly involved with at least one big data project.
When the candidates fit these criteria, the interview was conducted. Recommendations from Saunders et al. (2009) were followed in order to ensure the rigor of the data collection. The structure of the interview was read out loud. To each candidate was given the option to keep confidentiality of their name and company and asked if the interview could be recorded. Subsequently it was asked to each interviewee about his familiarization on the three parts of the interview. The interview proceeded by the order each interviewee mentioned to feel more comfortable with. Naturally, some interviewees provided more depth than others in each of the three parts of the interview.
The first two interviews served as a good pilot test for the remaining ones. Due to time limitations and small response rate on certain questions, some questions have been excluded and others reformulated in the proceeding interviews. As a result the average time per interview
was reduced and better insights have been gathered.
9.1 Method of analysis
All of the interviews have been transcribed manually into the Doc format and each one has been summarized by first and last name, company name, job title and gender. After the transcription process was complete, each one of the interviews was codified by using the Nvivo 10® software (Figure 29 - coding scheme in the appendix).
Due to the length of in each interview, two codification stages were applied.
Firstly, a coding scheme was created based on the proposed research framework. The three main parts of the interview have respectively been coded into three major codes. Furthermore, each one of the three major code nodes has been split into sub codes nodes in order to match the interview sub topics (e.g. The structure part, into centralization, job specialization and departmentalization). The coding process was done by dragging meaningful chunks of text to the respective codes and sub codes.
22
Using the Nvivo 10® query tool, more specifically, "word frequency", "text search" and "coding", resulted in simpler and better visualization of data. Patterns have then been identified within each coding node. So for example, within the value creation main node. Value creation barriers were frequently mentioned across the interviews. (Figure 29 - coding scheme)Secondly, to each one of the patterns has been attributed new code, i.e., coding was done by a data-driven manner (Fereday & Muir-Cochrane, 2006).
The methodology used to analyze the codified data was focused on the pattern matching analytic technique, this technique served to compare empirical patterns found within the codified data against the predicted propositions. Positive matching helped to increase the internal validity of a case study (Yin, 2009). The coding scheme is provided in the appendix.
10. Analysis and results
Following the application of the pattern matching analytic technique, several patterns have been identified within each sub-code. This chapter is therefore divided into three parts that correspond to the design of the conceptual framework presented above.
10.1 Value creation
Two particular ways of value creation have been identified. The most common way to create value from big data is internally ("by running your business better"). This stream of value creation is composed by all the internal benefits that stem from the exploitation of big data within the company thus resulting in an increase of revenues and/or in a reduction of costs. The second form is externally ("creating new business opportunities"). This stream of value creation is a result of new revenue sources derived from the monetization of big data. It currently represents the biggest challenge for mobile operators since only a few of them are creating value in this segment.
The main difference between the two is the distinct business model underlying them. Regarding the internal form, mobile operators aim to improve their traditional business of selling voice, SMS, etc. In the external form, mobile operators adopt a new business model by providing customer data to third parties.
Figure 30, Figure 31 and Figure 32 in the appendix provides a simple visualization of the different forms of value creation from big data.
Churn reduction
The most commonly mentioned way to create value internally was through churn reduction. Big data analytic capabilities help operators identify churn patterns in their sets of customer data allowing the possibility of identifying customers that are at risk of churning. With the right mechanisms in place, mobile operators can avoid churning and retain the customer.
Currently some operators are fighting churn by creating micro-pricing plans, by enhancing customer experience or even by launching micro BTL campaigns to potential churning customers.
Improved cross sell and up sell
Big data improves customer profiling. With better knowledge about their customers, telecom operators are able to better cross and up sell their products. The result is an increase of ARPU (average revenue per user), one of the most important KPIs used by mobile operators.
Customer experience
Mobile operators that use big data are more able to monitor their customer complaints and understand their needs. By accessing more complex information, customer service can be improved thus leading to an increase of customer satisfaction and reduction in churn. As interviewee 4 (KPN) mentioned, most of the operators have weaknesses concerning their customer care service. Those weaknesses are reflected on their traditionally low NPS (network
23
promoter score). This problem can be tackled using big data through the improvement of customer life management.Figure 10: Exemplification of customer life management using big data, provided by interviewee 10 (AM).
Network / CAPEX optimization
Using location based data, telecom operators are able to trace the movements and habits of their customers. Using this information, mobile operators can better design their network systems in order to optimize their capacity. Network optimization has several benefits. First, it allows the reduction of capital expenditures (CAPEX) by allocating towers accordingly to location usage. Therefore it minimizes the overall costs of the network deployment. Second, it improves the quality of the services provided. Since each antenna has a maximum capacity of usage, a trade-off between quality and capacity is always made.
Improved marketing campaign performance
New sources of data, such as behavioral data, location based data, commuting data, allow better customer segmentation. Mobile operators have been fitting their marketing campaigns to their customer using micro segmentations to reach their targets, allowing them to be more successful in capturing and retaining their customers.
Improved pricing
Pricing is an important tool used mostly used by the marketing department in telecom operators. Understanding customer's price elasticity allows the operator to optimize its price plans and to increase its margins.
Nevertheless, as mentioned by interviewee 9 (DP), to solely use big data analytics in order to calculate customer price elasticity is not correct.The control group technique has also a role to play in these finding price elasticities.
Improved decision making
The least mentioned benefit was improvement in decision making. Big data helps operators in creating more insightful reports within operations, which are used to improve their decision making.
The recent increase in new technologies such as SAS platforms, is allowing better visualization of data prompting better and faster decision making.
Interviewees were also asked whether big data is improving automatic decision making on the frontline level. Despite some evidenced benefits, automation is still not in place for most of the mobile operators due to organization hindrance.
24
Figure 11: Big data for internal decision making, provided by interviewee 10 (AM).Monetization of big data
Monetization of big data represents the newest revenue source amongst mobile operators. Some operators are packaging data and selling customer based insights to other third parties. Another way of monetizing big data has been by providing customized services such as insurance services, traffic insights or commuting reports.
Several sources of data are being combined by some mobile operators in order to sell big data related services to third parties. From instance EE (UK mobile operator) is selling services based on customer location to city halls in order to help them predict commuting patters about their residents. T-Mobile uses a combination of customer location and demographics based information to sell insurance services to their customers. Therefore, monetization of big data represents a new source of revenue for some operators and it represents a whole new business model and new customer base for mobile operators.
25
Mobile advertisementBased on better customer segmentation, some mobile operators have been using customer data to provide advertising services to third parties. Two ways have been evidenced: mobile operators can either directly use their client segmentation to provide mobile ads to third parties or share customer information with advertising agencies. Though mobile advertisement had been already a revenue source for most operators. It was still mentioned by respondents as a service that has been improved by using big data. As interviewee 4 mentions (quote in the appendix), with the right use of customer data, mobile advertisement can be more efficient by targeting better their own and third party customers.
10.2 Barriers to value creation
Throughout the process of data analysis, several obstacles of value creation have been identified. These obstacles were then codified under "barriers to value creation" by identifying patterns across the interviews.
These findings became an important addition to this research, since they have neither been initially deducted from the theory nor included in the research framework.
Figure 12: Value creation barriers.
Lack of capabilities has been identified has the most common barrier mobile operators face
when exploring big data. This obstacle can be divided into two categories based on their different effects in value creation.
Internally, as interviewee 6 (DP) and 11 (AM) mentioned, mobile operators have been using analytics for quite a long time (Quote 1 and 2 in the appendix).
The relative strength of analytics in telecom compared to other industries can be attributed to their large customer base and historical use of analytics (quote 3 in the appendix).
26
As technology changed, mobile operators became more able to analyze different sources of data and use them for internal value creation purposes (quote 4 in the appendix). However as Interviewee 9, 11 and 4 mentioned, technology is not a sufficient condition to create value, the right capabilities have to be in place (Quote 5, 6, 7, 8 and 9 in the appendix).As the level of data complexity intensifies the lack of more advanced analytical capabilities becomes more evident. According to some respondents, big data analytics capabilities reside mostly in people and therefore processes must be put in place in order to fully leverage their skills.
Externalization of data also brings new challenges to mobile operators. It evidences even more their lack of capabilities in monetizing big data. The justification partly lies in their difficulty to deal with different types of data than the traditional sources they were used to (Quote 10 in the appendix). Moreover it requires more customized solutions regarding the services provided which also represents a big challenge for most mobile operators (Quote 11 and 12 in the appendix).
Evidence was also found that analytical capabilities can evolve in a path dependent manner and are related to the market maturity level that mobile operators are competing on. As some respondents mentioned, mobile operators operating in more mature markets such as Europe have better analytical capabilities than others operating in emerging markets such as Africa, Middle East and Asia.
Creating value externally means moving away from traditional business models.
As interviewee 10 and 9 mentioned, mobile operators are now dealing with completely different products than the ones they have traditionally dealt with ( sell voice, SMS, and internet related services to a large customer base). Today some are selling data related services to a different customer base, mostly other organizations. Whereas before the business was B2B, with big data it moved to B2C. Moving from a mass market company to an enterprise business company. Considering this, most operators are not yet prepared to sell data-based services when compared to other bigger and older players present in the market (Quote 13 and 14 in the appendix).
The last identified obstacle was the misalignment of strategic priorities. Some mobile operators have not considered big data a top priority to be explored. According to some respondents, externalization of data doesn't create significant value in the short term, thus some operators prefer to maintain their main focus on increasing their traditional revenue stream (Quote 16 and 17 in the appendix). Uncertainties associated with the risk of regulation infringement were also mentioned to impact the mobile operators' priorities (Quote 18 in the appendix). Furthermore market maturity was found to also influence this obstacle (Quote 15 in the appendix).
Market maturity
From the above findings, some respondents regarded the value creation obstacles as being driven by the maturity level of the market. In immature markets the relative level of analytics performed is low when compared to more mature markets.
Figure 13: Market life cycle effect on revenues.
As interviewee 9 (DP) mentions, the evolution of the markets towards a more mature level made MNOs look into big data to create efficiencies in their operations. On the other side, MNOs
27
operating in emerging markets are focusing their priorities towards network deployment in order to increase their market penetration.10.3 Development of capabilities
This section analyzes the different modes used by operators to explore big data. Two main forms have been identified on how operators develop their capabilities: organic development and strategic partnerships. Outsourcing analytics was mentioned has a common way some mobile operators use in order to fill their capabilities gap.
No particular way of developing capabilities has been found. Nonetheless, as interviewee 10 (AM) mentioned, this choice depends on the data complexity that each operator faces (Quote 19 in the appendix).
Outsourcing
As some respondents mentioned, it is very common to see operators outsourcing analytically based tasks to third parties (Quote 20, 21 and 22 in the appendix). According to interviewee 11 (AM), the biggest advantage to outsourcing is the mobile operator's ability to quickly and easily incorporate into their operations the wide supply of available technologies from third parties. Outsourcing is done by buying third party base platforms and the support services these outsourcing vendors provide (Quote 23 and 24 in the appendix).
Strategic partnerships
Strategic partnerships assume the form of a bilateral relationship between an operator and a partner towards a common objective. According to some respondents, partnerships are mostly used in situations where operators are externalizing data.
Partnerships were mentioned to help mobile operators in surpassing their capabilities gap (Quote 25, 26, 27, 28 and 29 in the appendix).
Advantages Respondents Disadvantages Respondents
Access to Know-how 9/12 75% Data leakage 5/12 42% Access to market 2/12 17% Sharing value 3/12 25% IT scale and experience 2/12 17% Slow development of
analytics
1/12 8% First mover advantages 2/12 17%
Figure 14: Advantages and disadvantages of partnerships.
Several advantages of strategic partnerships were identified. The most referred advantage was the access to the partner's know-how (Quote 30 and 31 in the appendix). By partnering with other big data specialists, telecom operators benefit from the know-how of the other party has. This advantage was mentioned as particularly important in the development of big data capabilities since it allows a faster development of the telecom own capabilities by using existing knowledge and experience.
From the external point of view, partnerships allow mobile operators faster access to the
market, by exploring the partner's existing client portfolio and by having privileged access to his
distribution channel (Quote 32, 33 and 34 in the appendix). Therefore, partnerships bridge mobile operators customer gap. Moreover by using the partner's existing customers, mobile operators avoid delays in sales and also join the credibility status their partners have. As interviewee 4 mentioned, EE is partnering with other companies in order to explore existing enterprise customers this partners have. It allows faster sales and reduction of costs in start up sales.
Mobile operators can harvest first mover advantages by exploring untapped customers and by establishing an image of credibility in the market (Quote 35 and 36 in the appendix). As it was pointed by interviewee 2. By partnering with GFK, TDI's benefited from GFK's market research
28
service reputation. Complementary, interviewee 4 mentioned these advantages to stem for the opportunity to build customer relationships ahead of their competitors.Mobile operators can leverage their data through the use of partnerships since in certain cases their partners have a bigger IT scale. By leveraging on their partners' data, mobile operators are able to complement their sources of data. For instance, mobile operators have an extensive customer base information (demographics, location) about their customers, yet some of them uses third parties in order to gather market intelligence (market dimension, segmentation information). Furthermore the partners have experience selling knowledge to other third parties which can help MNOS surpassing their difficulties in selling insights in B2B environments (Quote 37 and 38 in the appendix).
On the other side, partnering can also bring disadvantages to mobile operators. Without the right governance modes, data can leak outside the company to competitors or to other third parties (Quote 39 and 40 in the appendix). Allowing partners to access their databases without any control mechanisms can be dangerous for mobile operators. As some respondents mentioned, data can eventually fall in the wrong hands such as the ones of competitors or even come out in public. It represents a risk for these organizations not only because data leakages can have legal consequences for organizations, but to also have a negative impact on their reputation.
Despite the privilege access to their partners know-how, in some cases partnering might block the development of internal capabilities (Quote 41 in the appendix). As mentioned by interviewee 1, if this knowledge and experience in not transferable an codified inside, then the operator will not develop their big data capabilities, and will become more dependent on their partners.
Finally when partnering with other parties, mobile operators are obliged to share the value created from their new business (Quote 42 and 43 in the appendix). In this case, the profits resulted from providing big data related services are not totally retained by the mobile operators. Organic development
Organic development is a form of developing big data capabilities inside the company without the support of other third parties. Organic development was mentioned to be preferred when the level of analytical complexity is low. One of the disadvantages (4/12, 25%) is the huge amount of resources and investment required to develop the capabilities in house (Quote 44, 45 and 46 in the appendix). Moreover, the relatively long time required to develop internally these capabilities is regarded as a disadvantage (4/12, 33%). On the effectiveness of the organic development, as interviewee 1 mentions, UK's mobile operators have not been successful in monetizing big data on their own. Most mobile operators did not have the right commercial and marketing structure to package the product and sell it. In order to monetize big data, these mobile operators opted for partnerships.
10.4 Structure
This section aims to answer the first three hypotheses on the impact of organizational structure on value creation from big data.
10.4.1 Job specialization
From the research framework presented above, literature suggests that data scientists are the most important job specialists in order to explore value from big data. This sub-section explores the role of these people, their skills and their availability. Further other important job specialization types were deemed necessary in order to create value from big data.
Roles & Skills of data scientists
Role identified Respondents
Modeling & Coding 5/12 42%
29
Skills identifiedBroad understanding of data 4/12 33% IT and analytical knowledge 3/12 25% Figure 15: Roles and skills of data scientists.
Evidence from the interviews showed that data scientists are a key asset that mobile operators have to deploy in order to create value from big data. Mobile operators are dealing with increasingly different sources of data, therefore it is important to retain data scientists given that they are the ones able to reconcile and find value out of big data (Quote 47 in the appendix). Along with the data collection process, each interviewee was asked about the availability of data scientists in the market.
Respondents (%)
Low 6/12 50%
High 3/12 25%
Figure 16: Availability of data scientists.
Most respondents said that data scientists are difficult to find and that supply is not equally dispersed across the markets. Some suggested that certain markets like the Asian market have a larger supply of data scientists due to specialized university programs (Quote 48 in appendix). Some respondents mentioned data scientists as very difficult to retain due to industry characteristics regarding the pay level (3/12, 25%). An example provided mentions that the telecom industry pays lower salaries in comparison to the finance industry. Considerations about different markets have also been made, for instance, IT personnel is mentioned to be paid better in the USA compared to Europe (Quote 49 and 50 in the appendix).
As it will be analyzed further, data scientists are an important resource to deploy when developing big data capabilities. The fact that data scientists are hard to hire and retain, became an important obstacle against value creation from big data.
Other job specialization types have also been identified as important when mobile operators
want to monetize big data. Despite the importance of data scientists in unlocking value from big data, other job specialists are required in order to sell it. As some respondents mentioned (4/12, 33%), new marketing and sales team have to be in place in order to package and sell data to other third parties. (Quote 51 in the appendix)
10.4.2 Centralization
The second hypothesis suggested that centralization of analytics has a positive effect on big data value creation. While most operators are still far away from centralizing analytics, most of the respondents don't doubt that centralization of analytics can create advantages to mobile operators. Below is provided the status of centralization in the cases studied, as well as the advantages and disadvantages that come along with the centralization of analytics.
According to the respondents, analytics are mostly located in the marketing department, yet other departments also perform their own analytical tasks.
Department Respondents (%) Quotations
Marketing 7/12 58% "So usually where big data has been sitting has been in the marketing department, in the business intelligence unit."
Silos 6/12 50% "Analytics is everywhere in your company, it is always distributed"
ICT 3/12 25% "It falls under what we call enterprise business units, normally under there what they do is they have all the
