Master’s Thesis Data-driven Control

(1)

Master’s Thesis

Data-driven Control

The influence of Big Data Analytics on Management Control

Systems in the energy sector.

By

Max van Wamel

S3540332

Supervisor: Dr. A. Bellisario

Co-assessor: Prof. Dr. I.J.J. Burgers

June 24, 2019

MSc Business Administration: Organizational & Management Control

Faculty of Economics and Business

University of Groningen

(2)

Abstract

Nowadays, there is a growing interest in big data and big data analytics (BDA) in the energy sector. BDA becomes more and more important in the energy sector due to the energy transition and the digitalization of smart energy systems. BDA has proven to support decision-making on multiple grounds. Additionally, BDA influences the management control system (MCS) in various positive ways. However, research is still lacking evidence and only a few scholars attempt to study BDA in relation to MCS. MCS in relation to BDA has not been investigated in the energy sector before. This paper conducts a multiple-case study in-depth qualitative research among four companies in the energy sector to investigate the relation between BDA and the design and use of MCS. Doing so, this paper concludes that BDA mostly has enabling effects such as enabling organizational learning and enabling the MCS to be more dynamic and pro-active. Contrary, this paper also concludes that BDA has a disabling effect, namely lack of understanding BDA that disables the quality of MCS.

Key words: Management control systems, Smart-energy systems, Big data, Big data Analytics, Energy

(3)

List of tables

Table 1: Overview interviews ... 8

(4)

1

1. Introduction

Currently, the field of big data is rapidly growing due to the development of technologies, such as sensor technology, cloud computing and smart mobile devices that generate huge amounts of big data. Big data is defined as huge data sets with great variety of data sources and formats (Samuel Fosso Wamba, Akter, Edwards, Chopin, & Gnanzou, 2015; Idrees, Alam, & Agarwal, 2018), high required data processing velocity (Choi, Wallace, & Wang, 2018) and further, it is unpredictable (Beulke, 2011). In the energy sector, energy systems are being digitalized and there is an increase of emerging information technologies (Zhou, Yang, Shen, Ding, & Sun, 2015). Innovations such as the smart grid produce huge amounts and different types of data that generate big data. Although big data generates big challenges in form of variety, veracity, velocity, value and volume (Samuel Fosso Wamba et al., 2015; Idrees et al., 2018; Sivarajah, Kamal, Irani, & Weerakkody, 2017), big data also provide opportunities which can lead to a competitive advantage when they overcome these challenges (Bumblauskas, Nold, Bumblauskas, & Igou, 2017; Samuel Fosso Wamba et al., 2015). To do so, big data analytics (BDA) must be used in a iterative and sensitive manner (Jukić, Sharma, Nestorov, & Jukić, 2015). BDA can be defined as techniques that acquire and analyse intelligence from big data and therefore, extract insights (Gandomi & Haider, 2015).

(5)

2

all the stakeholders of the energy sector (Zhou et al., 2016). Raffoni et al (2018) suggest that further research should investigate the phenomenon BDA in different industries and companies to gain knowledge and understanding of companies which use external, unstructured and huge amount of data. Other scholars suggest to conduct in-depth qualitative based case studies in the phenomenon BDA and management control research (Samuel Fosso Wamba et al., 2015; Sivarajah et al., 2017). For those reasons, this paper seeks to contribute to the management control and BDA research, by investigating the phenomenon BDA on the design and use of MCS in the energy sector conducting in-depth qualitative case studies within 4 companies to answer the research question: “How do big data analytics influence management control systems design and use in the energy sector?”.

The findings show that BDA influenced and changed MCS in different ways. First, BDA supports both the diagnostic and the interactive use of MCS. Second, BDA provides important information and discloses hidden insights that enable organizational learning. Third, BDA ensures change in MCS that is constant-driven and evidence-based which enables the MCS to be dynamic and sensitive to changes. Contrary, due to data pollution and opportunistic behaviour, the findings indicate a disabling effect on the quality of MCS. The paper is structured as follows, first, the theoretical framework provides an understanding of BDA and MCS in the energy sector. Then, the research design and data collection are discussed. Consequently, the description and analysis of findings are given. Finally, the study will provide a discussion and conclusion of the findings.

2. Literature review

(6)

3

2.1 Management control systems

Managers implement controls to keep their organizations on the right path (Merchant & Otley, 2006). These controls come in many forms which vary from simple operating procedures to elaborate processes, such as evaluation reviews (Merchant & Otley, 2006), and are part of the bigger system known as; the management control system (MCS). As Cunningham (1992) describes, MCS embodies the mechanisms and techniques to pursue company objectives, achieve company goals and pursue strategies (Cunningham, 1992). This is in line with (Ferreira & Otley, 2009), who note that MCS assist the strategic process and continuous management in the way of rewarding, control, analysis, planning, measurement and widely managing performance. Thus, MCS is an important system for supporting many organizational practises. Moreover, Merchant and Otley (2006) argue, indifferent what the goals of the organization are, that an organization which is “in control” is expected to have good performance against its goals. However, designing such a MCS is a highly complex assignment (Rotch, 1993). According to Merchant (1998), the design of MCS requires an understanding of the organizations’ desired actions and results and the people, assignments and circumstances being controlled. Moreover, contextual factors such as external environment, technology, strategy and national culture are in prior research often associated with the design of MCS (Chenhall, 2003; Langfield-Smith, 1997). Thus, there are many internal elements in the organization as well as external from the organization to take into account for designing a MCS and therefore, it is a complex process.

Therefore, many frameworks were introduced, one of them is the levers of control (LOC) which is proposed by Simons (1995). The LOC framework was used by many researchers (Ferreira, 2002; Tuomela, 2005; Widener, 2007). Despite its popularity among researchers, this framework still has weaknesses, such as too specific focus on top-level controls. Therefore, the framework does not stress a scope of informal controls well and does not emphasize properly the operational controls in lower organizational levels (Ferreira, 2002). As a result, Ferreira and Otley (2009) propose their Performance Management Systems (PMSs) framework that captures research beyond particular features of management control systems, and the boundaries of existing frameworks, such as Otley (1999) and Simons (1995). Therefore, the framework of Otley and Ferreira (2009) is used in this study to appropriately enter the field and understand and describe the MCS. The PMS framework of (Ferreira & Otley, 2009) emphasizes the design of MCS in eight core components namely, (1) vision and mission, (2) key success factors, (3) organizational structure, (4) strategies and plans, (5) key performance measures, (6) target setting, (7) performance evaluation and (8) reward systems.

(7)

4

and adjusting deviations from predefined standards of performance (Henri, 2006). Standards of performance are referred to as goals and targets, such as key performance indicators KPIs. Moreover, by examining with critical performance variables the diagnostic use monitors and direct implementation of predetermined strategies (Henri, 2006). Hence, as Henri (2006) described, the diagnostic use is a process where feedback is evident and goals and outputs are compared and simultaneously adjusted. This is in line with Otley and Ferreira (2009), who argue that within diagnostic use, feedback information flows are crucial.

The interactive use of MCS is organic and mostly learning-orientated and used for opportunity seeking in the organisation (Henri, 2006). The focus of the interactive use is to create dialogue and displays signals to top managers of the organization (Henri, 2006). In addition, Otley and Ferreira (2009) suggest that feedforward information (interactive use) enables organizations to learn from experiences and generate ideas and redevelop plans and strategies. Additionally, the interactive use aims on strategic uncertainties, in a way of disqualifying and threatening underlying expectations of current strategies and guides this bottom-up (Henri, 2006). Otley and Ferreira (2009) add that, strategic validity controls are needed for re-examining the strategy by, for instance, open discussions where feed-forward information flows are crucial. Hence, information flows are important in helping the managers and controllers to design and use the MCS. Another important factor pointed out in previous literature is the analysation of information or data to help managers set up components of the MCS, such as strategic specific benefits, and identify primary success factors (Raffoni et al., 2018). Since, the ability of analysing data becomes more and more important, researchers refer to the “era of big data” (Choi et al., 2018) and the era driven by technologies, in which big data is emerging and of high interest for the world of academia and the world of business (Dezi, Santoro, Gabteni, & Pellicelli, 2018). However, little is known about how these technologies change the MCS and how these technologies help managers design and use MCS (Samuel Fosso Wamba et al., 2015; Sivarajah et al., 2017).

2.2 Management Control Systems and Big Data Analytics

(8)

5

BDA can be defined as techniques that acquire and analyse intelligence from big data and therefore extract insights (Gandomi & Haider, 2015). BDA allows managers by giving more updated and concrete information, which helps to improve the quality of decisions which indicates moving from a more perceptive to a more evidence-based decision-making (Ittner & Larcker, 2005). Hence, to process large volumes of data to get valuable understanding and facilitate evidence-based decision-making, organizations need efficient methods (Gandomi & Haider, 2015). These methods refer to BDA. In detail, BDA analyses big data and can be carried out in the way of, descriptive analytics, diagnostic analytics, predictive analytics and prescriptive analytics (Kibria et al., 2018; Raffoni et al., 2018). Firstly, descriptive analytics are the most basic analytics and do closely examine past and current data (Raffoni et al., 2018). These are considered as the preliminary step to the effective application of following types of analytics (Raffoni et al., 2018) and is evident in the measurement of key performance indicators (KPIs) (Kibria et al., 2018). Second, diagnostic analytics look at past performance to determine what the problem is and why it happens (Raffoni et al., 2018). This method uses techniques like drill-down, deep learning, data discovery and correlations to figure out for instance faulty KPI’s (Kibria et al., 2018). Third, predicative analytics is a method to make predictions (Kibria et al., 2018; Raffoni et al., 2018). These deliver usually predictive forecasts based upon real-time and archived data (Kibria et al., 2018; Raffoni et al., 2018) and managers can do experiments and test new business models followed by an evidence-based decision-making approach (Raffoni et al., 2018). Examples of techniques for predictive analytics are machine learning, data mining examine scenarios in time series, Monte-Carlo simulation, pattern recognition and alerts, and forecasting. Finally, prescriptive analytics are aiming to show which action must be taken and suggesting decision options with the implications (Kibria et al., 2018; Raffoni et al., 2018). Prescriptive analytics are considered the most valuable kind of analysis due to having robust techniques, such as virtualization, edge-computing, network slicing, policy and subscriber management and 5G (Raffoni et al., 2018). Moreover, it can show the best course of action for KPI’s, can give recommendations and set up management performance rules.

(9)

6

(interactive control), is supported by their business performance analytics. Therefore, referring back to (Ferreira & Otley, 2009), information flows, feedback (diagnostic use) and feedforward (interactive use) information flows should also be supported by BDA. As a result, BDA helps managers and controllers with the design and use of MCS. Altogether, prior research shows that big data and BDA has an influence on MCS, however, it is not clear how BDA helps managers and how this influences the design and use of MCS. To study the phenomenon of BDA, an industry in which an appropriate amount of external and unstructured big data sets are available for the application of BDA is favourable (Mayer, V., Schönberger, K., and Cukier, 2014). Thus, studying the phenomenon BDA should have a lot of potential in industries which are digitalized and where are multiple unstructured data streams available.

2.3 Management control systems and Big Data Analytics in the Energy Sector

Energy systems are being digitalized and there is an increase of the amount of emerging information technologies within the energy sector (Zhou et al., 2016). Moreover, the introduction of smart energy systems in the energy sector, such as the smart grid, generates huge amounts of customer and production data, which may not be equal to that of other industries and thus, depicts a big challenge to analyse and process this huge amount of data (Zhou et al., 2016). The smart grid is defined as “the next generation power system able to manage electricity demand in a sustainable, reliable and economic manner, by employing advanced digital information and communication technologies” (Diamantoulakis, Kapinas, & Karagiannidis, 2015, p.94). The large amount of production and consumer data generated by the emerging information technologies (smart grid) and the innovations due to big data are changing the traditional energy industry (Zhou et al., 2016). As a consequence, the energy industry faces challenges on multiple grounds, such as, energy efficiency and environmental issues (Zhou et al., 2015), renewable energy management (Altin et al., 2010), operational efficiency and cost control (Momoh, 2009), consumer engagement and service improvement (Aalami, Moghaddam, & Yousefi, 2010) and system stability and reliability (Amin, 2008). Therefore, according to Zhou et al (2016), BDA can be an efficient decision support for many stakeholders of the smart grid, such as producers, operators, customers and regulators.

(10)

7

3. Methodology

3.1 Research design

Because of the explanatory nature of this paper, this research is the most relevant for qualitative research. In order to answer the research question, I conduct multiple in-depth case studies to get a deeper understanding of the phenomenon BDA on MCS. Multiple case studies consist usually out 3 to 5 cases and mostly investigate explanatory research questions by using semi-structured interviews. With multiple case studies replicate the pattern-matching helps strengthen the results and therefore enhances the robustness of the theory (Tellis, 1997). This is subsequently beneficial for this study.

There will be four case companies examined in the Netherlands. Purposeful sampling will be used to illuminate the question underlying the study. Purposeful sampling is about selecting information-rich cases studies in depth (Welch & Patton, 1992). In this research the variant criterion sampling is used which makes use of predetermined criterion of importance (Welch & Patton, 1992). As Welch and Patton (1992) noted, “Criterion sampling can add an important qualitative component to a management information system or an ongoing program monitoring system” (Welch & Patton, 1992, p.177). For this sample are two important criterions composed. First, the companies selected for the sample need to be companies from the energy sector. The reason is that within the energy sector is a huge availability of customer and production data that provide huge unstructured data sets (Zhou et al., 2016). Therefore, it is more likely that companies in the energy sector use the application of BDA and as Raffoni et al (2018) suggested, the application of BDA is more favourable in this situation. This is important for investigating BDA on management control systems in depth. Second, the managers and controllers of the companies need to be familiar with BDA for understanding how this phenomenon influences the MCS. To make sure this criterium is recognized, this research uses snowball sampling or so called chain sampling for sampling interviewees. This to make sure that get key and information-rich interviewees (Welch & Patton, 1992). At least 2 people were asked before getting to the interviewee to ensure information-rich interviewees in the different cases. Controllers and managers are sampled because these practitioners are both mainly involved in the design and use of MCS (Ferreira & Otley, 2009; Simon, 1995), and are confronted with big data and the help of BDA (Raffoni et al., 2018; Sivarajah et al., 2017).

3.2 Data collection

(11)

8

Table 1: Overview interviews

Case company

Company expertise

Number of

interviews Duration Role

Trapeco Network operator 2 3 hours

Strategic advisor/ Manager D&I

Tilenenergie Network operator 1 35 minutes Controller ICT Enerkie Energy supplier 2

1 hour 40 minutes

Manager digital/ Product owner

Datacom Data trader 1

1 hour 20 minutes

Energy consultant/ Project manager

The interviews are taken semi-structured by asking open and closed questions to collect data about the phenomenon BDA and its influence on the MCS. In order to appropriately reflect the research question the interview is divided in two sections. The first section of questions reflects the influence of BDA on the design of MCS. The second section of questions reflects the influence of BDA on the use of MCS. This is important for not mixing up elements by the interviewer and interviewees. In the framework of Ferreira and Otley (2009), this distinction is also made. The interview questions are inspired by the literature framework and in the appendix the interview protocol is attached.

3.3 Data analysis

(12)

9

4. Findings

This section is organized based upon the informant’s description of MC. This section shows how BDA influenced the design and use of MCS in the case companies. Moreover, the findings show mostly enhancing effects, but on the other hand, they show a disabling effect. The figure below contains the data structure which was derived from the analyses during the study.

Figure 1: Data structure

(13)

10

(AI) modelling, machine learning, deep learning and drill-downs. The findings show that BDA influenced and changed MCS in different ways. Moreover, BDA provides important information and discloses hidden insights that enable organizational learning . Additionally, BDA ensures change in MCS that is constant-driven and evidence-based which makes the MCS more dynamic. However, due to data pollution and opportunistic behaviour, the findings indicate also a disabling effect on the quality of MCS. Lastly, the managers came across challenges and considerations which are not directly involved in answering the research question, but are important to note. For me, as the author of this paper, it is important to communicate those implications of big data and BDA to the reader. Hence, my findings contribute to the research because the challenges the companies and managers face are crucial for the context of the research.

4.1 Enabling organizational learning

The growing use of data has led to a more dynamic and competitive environment in the energy sector. This growth of big data is driven by different sources such as the big data from smart energy systems. Big data is very valuable for the case companies and can be used to find patterns and derive predictions. Moreover, the case companies agree that big data helps them to be predictive and see this as a major impact of big data. As a result, a shift is evident in looking backwards to more proactive ways such as predicting.

´That you look now very much backwards, before the big data era. I think that for sure with big data would you look forward and can be more predictive, I think that´s the biggest impact of big data.´ (G, Tilenenergie)

BDA technologies are able to assist managers of the case companies to be predictive. Additionally, the managers from the case companies made mostly a distinction between operational level and strategic or tactical level. On the strategic level, all sort of BDA are deployed to assist management control and strategic directions. BDA methods, such as agent-modelling, pattern recognition and machine to machine learning assist and control strategic decisions. BDA is able to create insights and assist the strategic level of control by helping managers to direct the company goals and objectives from the strategy.

‘No, that will be for sure, it creates very much insights, but I think that the consequence of course is because the whole discipline is in the beginning of its infancy. It is not very old, we can do a lot more with it. But in the future my expectation is that we will see it more and more.’ (R, Trapeco)

(14)

11

normally need for instance where people calculate 10 or 12 scenarios and make a choice out of that, can a system calculate just 100000 scenarios in the same time or maybe even more.’ (R, Trapeco)

‘Than computers, AI’s are much faster and more efficient than people.’ (W, Trapeco)

‘You are able, I think to work with more scenarios, so than indeed, what you are saying, with the glasses forward indeed.’ (G, Tilenenergie)

On the operational level BDA is also assisted by predictive BDA technologies. Similarly to the strategic level, BDA adds value by creating insights. Additionally, for instance, this is evident in the turn of a customer which is forecasted with a model used by Enerkie. Moreover, at Enerkie, they used a specific forecast model to forecast how much are people willing to pay and with what price offered it is most likely to gain the interest of new clients. This promotes debate and managers need to interpret and discuss the outputs of the forecast model. However, this forecasting model was not based upon big data, but it showed the big potential of BDA and big data in the MCS on operational level.

‘Than we have a second model, because it’s about, how are we going to make money within this company. Well, than there is of course a chance when you ask a lot than you have a smaller chance to score a client, than on the moment when there is a more descent number. And then you get always the question what is actually descent and what is actually good. And that is actually more that forecast model.’ (J, Enerkie)

Network operators have their interest in the electricity cables and their interest is to know everything about these cables. On an operational level, condition-based-maintenance is used to predict where and when the need of electricity is higher or lower. This is control on the resources or assets which can be assigned to quality, staff, skills and material to run the grid. As a consequence, an AI model predicts based upon profiles where electricity is needed. Thus, when these predictions deviate from the strategy, managers will be able to adjust the direction of the strategy because the predictions question if it is the good direction. The AI use support feedforward and feedback information flows which are supporting debate about the results of the model.

(15)

12

However, predicting different scenarios is dependent on which data you have to your availability and how the time lapse of your data is. Big data which is generated on a high frequency and therefore generate on a short time lapse a great volume of data. How shorter the time-frame of data, such as a 15-minute time frame, how more accurate the predictions.

‘So, how finer the data is, so that are 15-minute values, how better you can predict.’ (G, Datacom)

However, a prediction is mostly not 100 percent accurate, so in the end it is important to see what the company actually accomplished. Thus, the actual results can be used as an input and can be compared again to adjust the strategy and direction of the company. Nevertheless, predicting is a really important part in the data-driven world and BDA therefore has become, or is becoming, a key component in the process of setting up plans and strategies and becomes part of the MCS. It facilitates feedforward information flows and enhances interactive use of the MCS.

‘Yes for instance, people eat now asparagus, we have decided that we have 500 products, we have decided that for 20 products we need more and we will buy because our intelligence said that it is eaten more. Was that a good decision? […] And then your big data/analytics is itself part of management control. (W, Trapeco)

Moreover, BDA are really good in finding and recognizing patterns from the past. It helps managers directing and motivating the goals throughout the company. Predictions can identify possible blind spots which are not identified by the managers. Additionally, managers cannot identify the blind spots because they think framed and thus do not have the complete picture. Therefore, on strategic level, it gives managers the information to reconsider strategies. Whereas, on operational level, it helps managers to monitor and motivate particular targets and objectives by correcting them and thus directing the strategy. As a consequence, both diagnostic and interactive control are enhanced within the organization.

‘Because it gives insights, how particular things are more optimal tackled and where you can put your attention and focus on and where precise less. Because there is often a difference between the human acting at which still often have blind spots and think framed. And a model thinks not framed, that shows things which you never had imagined or never had seen and through this you are of course be able to better navigate.’ (R, Trapeco)

(16)

13

creditworthiness. They came up with the idea because the company is mostly lagging 6 months behind and that is most of the times too late to collect the invoices.

‘So, there we have now a machine learning model, a random forward model based upon history that looks to the best indicators of the client, such as, how creditworthiness and the payment amount is done the past years, at us but also at external parties. So, there you get results and it is hard to get those into practice.’ (D, Enerkie)

This is a great example in which the capabilities prediction and identification is used to find important information to improve MCS efficiency in operations. However, the question how this fit into the company’s MCS arises. The answer to this question was that it was a really difficult to take the subject of “bankruptcy” into the MCS. It was hard to determine based upon the model from which point the employees should come in action to solve the problem of the risk client, because of the sensitivity of the subject and the fear of losing clients due to wrong predictions. This shows the challenges in having confidence in the model and the difficulties in implementing the model into the operational MCS.

An important feature of BDA is that it can detect and identifies much earlier the exceptional things out of processes which may trigger to carry out particular controls, such as corrective action. Hence, BDA can detect things such as faulty measures and corrects those. Doing so, with the assistance of the information derived from the BDA, the managers can correct and adjust the measures. Hence, managers direct and motivate based upon accurate goals and objectives.

‘I think that because of big data you are able to maybe detect earlier the exceptions or the exceptional things out of the processes and that is certainly a trigger to carry out particular controls.’ (G, Tilenenergie)

‘Yes I think, look of course you have everywhere some measure mistakes etc, only than you can correct for particular things.’(D, Enerkie)

Another influence of BDA in the control of operations is detecting malfunctions much earlier and prevent these. Moreover, BDA is used to detect local malfunctions since it is able to detect whether those are local malfunctions or are part of a bigger malfunction in the electricity grid. The way how BDA detects these malfunctions is described by R. from Trapeco:

(17)

14

This example shows how BDA helps managers to control their assets and resources. Additionally, BDA detects malfunctions much earlier and therefore managers can correct these deviations (malfunctions) much earlier than they normally would do. However, on strategic level, there are also ways to use BDA on the grid where BDA provides valuable information. For instance, by setting up targets and directions based upon predictions for the load of the grid. As a result, the information derived from BDA can be used to fill in parts of strategies.

‘Through this you are able to act faster. But on strategic level and tactical level you are able to do also very smart things with load on the grid and things like that to the future calculations, emergency calculations and that kind of things. […] Yes, you can actually with that kind of information, you can fill in actually a part of your strategy.’ (R, Trapeco)

In the end, big data is extremely valuable and is full with patterns. By analysing these big data sets with BDA, patterns can be detected and thereby, surprising patterns can be found. As a result, managers can look to those patterns and can ask themselves why the MCS did not identify this pattern. As W. of Trapeco said:

‘Yes and that is the next step a MCS with AI would be amazing.’ (W, Trapeco)

Instead of looking at the risk itself, managers should look on the MCS and investigate why the MCS did not recognize or signal the pattern, fault or opportunity.

‘Yes that is something what going wrong, what everyone does is directly operational dive into the risk control what you do too at the moment. And nobody thinks about I am here the responsible.’ (W. Trapeco)

Finally, it is clear that BDA is embedded in the MCS. Additionally, BDA enhances diagnostic and interactive control by creating and supporting feedforward and feedback information flows. Therefore, managers are able to correct deviations faster and help them to monitor the strategy better. In addition, it helps managers with creating new ideas and new directions which questions the current strategies. Hence, BDA enables organizational learning.

4.2 Constant-event and evidence-based MCS

(18)

15

systems. Another indication and driver is that everything needs to be more and more efficient and cheaper.

‘No no, at very much is still standing, there are companies which are much further, but you must still not forget that energy companies well were somewhat more civil service companies.’ (D, Enerkie)

‘But it is typical one of the sectors that will use more data than ever. […] It has many drivers, it has very much to do with that always everything needs to be more efficient and cheaper.’ (W, Trapeco)

We see that as streaming data. And in the end when you have that all together then it becomes big data. [...] It is all your sensor data together’ (R, Trapeco)

Based upon the interviews, it was clear that the design of MCS was mainly represented by setting up KPI’s to give directions. To give directions, KPI’s were set up from the main goals and objectives of the companies and these were gradually specified down to the right measures to the different divisions. Additionally, this top-down control was to direct and motivate the companies’ goals and objectives through the divisions. From this point, the data generated from the KPI’s were used back as information for generating and adjusting the main goals. Examples of KPI’s were the number of clients and client satisfaction.

‘Yes, that is cascaded actually from top to bottom, from the goals the measures are set up in the surrounding divisions. So from there you cascade it up towards the main goals of the company.’ (R, Trapeco)

The informants refer to the traditional control system which is meant by setting up KPI’s, revised monthly or even yearly and are static from origin. The traditional KPI is static and the energy sector is getting more dynamic and competitive. Therefore, a distinction is made between the traditional KPI and the environment because the traditional KPI is not made to be dynamic and is instead classic, linear and mechanistic. It is not suitable for a big-data driven world, therefore the traditional KPI heydays are over and thus, there is need for a more dynamic kind of KPI.

(19)

16

The process of setting up KPI’s is mostly a slow process where managers set up, measure and adjust the direction of the division. Therefore, management tends to follow the facts instead of being ahead of the facts to appropriately monitor and adjust when needed. Hence, the traditional KPI needs to be more dynamic in a data-driven environment. Therefore, the traditional KPI’s are changed to the “new” KPI’s or core numbers which are more accurate and actual.

‘Then as management you are walking hopelessly behind. […] Yes a core number is not a key performance indicator. And when you want to control in terms of control, we call it nowadays Agile, so show what you made every week and I tell how much is not good.’ (W, Trapeco)

The informants also mentioned, that before the big data era, traditional MCSs were looking too much back to outputs of past events. Afterwards, these MCSs were used for explaining and, describing something what was already happened. This is not necessarily wrong, but it is not useful when the output is only used for describing and decisions in the future are not taken into account. Additionally, the data derived from the outputs should be used for decisions in the future. Thus, a more proactive MCS which supports setting up more accurate KPIs and targets is needed.

‘Decisions for the future, those take you now and within you take into account what happened in the past, that’s how you take decisions. But in the past you don’t have to take decisions of course.’ (W, Trapeco)

Moreover, the changing energy sector forces companies within the sector to change. Additionally, this means that the strategies and KPIs of companies change, too. As a result, there is change in organizational structure because the organizational structure depends on its strategies and targets. In addition, data-driven environments suit flexible strategies that require dynamic structures. Therefore, the case companies introduced Agile and Scrum which is working with short-time frame in multidisciplinary teams.

‘Yes I think that big data support of course the knowledge, the movement, the Agile movement assumes that the knowledge of the organization is mostly in the lower levels of the organization, so, the people who handle the daily particular processes.’ (G, Tilenenergie)

‘Well sure, Agile is naturally used a lot in the big data world, a bit Scrum and Agile.’ (W, Trapeco)

(20)

17

with the Agile and Scrum structure, knowledge should be in the lower levels of the organization. As mentioned above, Agile and Scrum is a dynamic structure. It enables managers to monitor, correct targets and give direction on a continuous basis. Therefore, the feedback loop becomes shorter on operational level which is suitable for dynamic environments.

‘So yes, the product owner I don’t know if you familiar with Agile Scrum? […] But that is nothing else than that I am the switch of clients internal as well as external clients to determine what is the most important what we are going to do. And it is Scrum team, that it is now a man of 6 whom are realizing and developing that.’ (J. Enerkie)

‘With Agile you ensure that you get long-length in teams at which you within that teams on detailed level continuously can adjust etc.’ (R. Trapeco)

Important to note, this is on operational level. So, a distinction needs to be made between strategic and operations MCS. As mentioned above, on operational level Agile and Scrum enable managers to monitor on continuous basis. However, on the strategic level, it is important to stick a longer period on the same track. Consequently, continuous control and monitoring will induce no focus and the company would have no stability in strategic direction, it would be a loose cannon. Agile and Scrum shift responsibility to the operational level to react faster on the environment and create a bottom-up information flow.

That has more to do with working from the Agile work way and where it is more logical because you do quarterly adjustments. […] But it is not true that you continuously adjust on strategic level because otherwise it is just a. […] Well than you just have no focus anymore and you fly to every direction, than you fly today to left and tomorrow to right and. […] In the Agile world you have still the stratification from the daily things at which you adjust, but you have to still think about what you want to do on tactical and strategic level.’ (R, Trapeco)

When you are on the managing operations level, you deem different things to be important than when you are managing on a strategic level.

Yes but if you are looking with a management glasses, if you watch the system dynamical or synthetical than you have always a control system and a control system. And if you look through the eyes of the management system than you try to see other things than when you are the control system. […] When you actual manage you should find other things important than things like the ice cream will be decently scooped.’ (W, Trapeco)

(21)

18

targets and goals. Moreover, Agile and Scrum has not a great interest in KPI’s, spreadsheets and dashboards.

‘Yes indeed, project and product control and that is not a world that directly have much trust in KPI’s and spreadsheets and dashboards and that sort of pranksters.’ (W, Trapeco)

The administrative organization is being digitalized and has further to be digitalized. Moreover, the administrative processes can be done much faster when these are digitalized. However, for instance reports about the KPI’s and targets need to keep up with the changing dynamic and competitive environment. Therefore, you could assume that these reports are increasing because of continuous monitoring.

´At the moment, may a set of KPI’s be agreed on which are valid during the year, and then with big data, you also have to go all the way into the flow, see that you can adjust your management reports and that you, what I think control earlier and acting faster, so that the dynamic also will change.’ (G, Tilenenergie)

Various administrative documents are also a reason why the case companies changed MCS. Moreover, the case company Enerkie started to look more forward 2 to 3 years ago and started shortening reports about KPI’s. They wanted to be more proactive and do proposals instead of preparing backward flavoured reports.

‘But also important is that you have to stop with reporting things and looking backward and bring deepening as much as possible the things. […] It is surely a difference with 2 to 3 years ago, is that we now have said what I just mentioned, those reports, all the necessary of the new KPI’s or else we will of course do it. But we are not reporting anymore all depth layers and wides unless it is really valuable for KPI’s or what else. But rather in this kind of things we want to do proactive proposals.’ (J, Enerkie)

(22)

19

‘You see that the average control system and then of course I do immensely wrong what is thought up today, but the average control system is all straightforward, they are goals and I go there or not. That also lacks the computing power to work on it. What you see is nowadays, you have two things, you have constant-event driven architecture, respond to events, I expect A to happen but B can also and I have plans for that. […] Entire decisions then become such a real-options theory. Computers can calculate much faster if you give value to all those real real-options. […] Well if you put all those real-options together and you build a control system on it that supports real leadership, I should say, really supports decision-making. […] That is, that are computers, AI’s are much faster and more efficient than humans. You will not think about not having a nice birthday party, but it is really an option. It can also rain on your birthday.’ (W, Trapeco)

Additionally, in this way BDA gives managers the room to focus more on the business and to continuous monitoring goals and objectives. Hence, allowing managers to perform the Agile and Scrum structure well. Consequently, managers have more time to direct and motivate the goals and objectives to their employees. Hence, the managers are better able in correcting goals and objectives where needed. As a result, the feedback and feedforward loop are shortened which should have a positive effect on the performance of the company and employees.

‘So yes, because you are closer to the ball, and that what I think, your feedback loop will become shorter and faster and I think that that has a positive influence on performance, you are able to direct and faster control.’ (G, Tilenenergie)

Apart from that, managers are always prejudged. However, a model is non-judgemental and does not take decisions and analyses data without having biases. Therefore, a model does not have self-interest, such as going to football practice or being ill. Moreover, a model does not care about the outcome which makes a model objective.

‘A model is non-judgemental, a human is always prejudged because a human takes always with him luggage from the past. And thinks in patterns too, while a model does not. So, a model shows non-judgemental what is the most optimal outcome or planning or does not matter what it is about.’ (R, Trapeco)

(23)

20

‘And if you now ask and you are sitting with somebody and say what kind of grade would you give me? Then you are yet fast inclined to give somebody a higher grade even when you had a bad conversation. For instance, feel nice feels always a bit unpleasant while if we can analyse your voice it shows that it is a 4, a 2 or 1 than is that much more interesting.’ (D, Enerkie)

During the interviews, terms as autonomous control came into light. Autonomous control is the next step in BDA technologies. BDA is now creating and identifying scenarios to support evidence-based decision-making. However, it is not unthinkable that in the future systems get autonomous which means that these systems are self-employed and make particular decisions, instead of managers making the decisions based upon the different scenarios.

‘Than you are of course not looking anymore to the administration. If you have a Samsung refrigerator, such a beautiful blockchain application have is own service, can order itself. What are you than still to key performance indicating with how many refrigerators are within 5 minutes may or may not have done etc, that will go all autonomous. So, you can look to very different things again.’ (W, Trapeco)

‘Yes the right choices proposed. Sorry, the first intention, the step has to be made that those choices are going to be proposed and show a top 10 and that still a human in the end make the definitive choice. But in the future it is possible that you get autonomous systems, that are being able to do self-employment.’ (R, Trapeco)

However, the great objectivity of information due to BDA, implies that management has to come with facts and evidence to introduce new products. Additionally, with the increased objectivity of data, managers can rely less on gut feeling and should evident their arguments. Hence, this means a more objective MCS based upon facts from data rather than gut feeling.

‘So that soon there will be no way that I can get away with it, for instance if I would say to my division that we have to introduce this product. I would say why would you introduce that product, that it have to be increasingly substantiated.’ (D, Enerkie)

4.3 Disabling quality of MCS

(24)

21

because of a huge workforce needed or mistakes being made. Consequently, dealing with big data is hard and not an easy task.

‘But in practice there are reasonable problems. If you go to real big data, so you get flooded by it. Then there are 2 things, 1 you don’t understand it anymore, you just don’t understand any of it, than you must go through all the datasets and those are not made for you. […] That is very expensive, so many people at the same time working, or there are enormous fundamental mistakes made. It is not such an easy trick, let me put it that way.’ (W, Trapeco)

This is the point where BDA jumps in, to analyse the big data and make the big data understandable. However, using BDA is not an easy task and BDA has its own challenges. Currently, the technical part of BDA is not necessary a problem. However, the output of a BDA is the problem. Additionally, managers and experts do not always know what precisely the output is that BDA generated. Furthermore, it can be even the case that the output is worthless. As a result, resources are spent on something which generates no value.

‘Technical it is quiet easy to get everything together at the moment. Yes that is possible nowadays. And the tools are also very pleasant, you can set an AI to work, but of course there are countless examples where you invested a lot of money and there is no output or if there is some outcome that you not precisely know what is done.’ (W, Trapeco)

Moreover, it is also important to control the BDA for preventing making wrong decisions. A reason for this could be that the input has changed. Another reason could be that the bandwidth has changed. Consequently, it is important to know that an AI model for instance works under particular conditions. When these conditions are not met, it is possible that management faces problems by making wrong decisions based upon the model. As a result, the quality the information flows are damaged and therefore, managers are facilitated with wrong information to create a strategy or to monitor the strategy throughout the company.

‘But you have to of course always have some kind of control on those systems of course, because you don’t want that those systems in one time are going to take wrong decisions because the parama that is put into it has changed or that the bandwidth within falls in 1 time differently. Because models are of course connected to particular, under particular circumstances works well and give results, but if that becomes different than can you face problems.’ (R, Trapeco)

(25)

22

‘No, and then there are of course magnificent stories known about AIs which of course only turned out to have learned to recognize snow by the fact that the sky was so blue. You think that you are discovering snowy landscapes, but the AI only thinks there is no cloud in the sky, that is why I call it snow.’ (W, Trapeco)

Managers are often in the literature referred to as being bounded-rational which means that managers are not able to understand everything. In the interviews, it is referred many times to this phenomenon. Additionally, managers can also be opportunistic which means that they act in their own self-interest. Similarly, managers have all kind of biases and take experiences from the past, those patterns into account when making decisions. As a result, managers make decisions which are not necessarily in the best interest of the company.

‘That information of big data can be much the same, the conclusions you connect to that have the tendency to be different. And where you than justly become, and that always with those management systems, is that people are bounded-rational and opportunistic. Bounded-rational means that you cannot always understand everything., that means that you are rational, you have all kinds of biases, you want to drive fast for instance. […] Yes, You are for instance biased to your own perception.’ (W, Trapeco)

Furthermore, managers come across the phenomenon of data pollution. This means that the data is already filtered or touched by others. As a result, wrong analyses can be done and based upon those faulty decisions can be made. As a consequence, the MCS is not able to fulfil its role as it should be and may guide wrong strategies throughout the company.

‘Than you see just faulty data gives faulty analyses. So, sometimes it is also very dangerous that you are dependent too, because you started something and you have still that human work and then from that moment, that faulty data having in your system and you are going to make decisions bases upon the whole analyse. Totally top, only the data is just wrong than it is sometimes very tricky.’ (D, Enerkie)

Data pollution can also occur due to the fact that managers are opportunistic and behave to their self-interest. As a result, managers can take advantage of the objectivity of data and manipulate it for putting it in their own advantage. When that is the case it can be very detrimental to the company. In that case the MCS is not fulfilling its purpose to control the behaviours of the employees, but instead giving the opportunity to act in their self-interest.

(26)

23

handsome drill-down nonsense. That shouldn’t be happening of course that big data also means big lies.’ (W, Trapeco)

‘Yes that is well, that is a hard one I think. Because yes you know what you get because you have more data and you can get maybe well, a nuanced view, but sometimes data looks very objective but you can also abuse the data, that it fits in your street.’ (G, Tilenenergie)

For companies, especially the energy suppliers have a hard time to get the right data. Energy suppliers, such as Enerkie, have a hard time to get real-time data. Therefore, they should install hardware and sensors such as smart meters to get this real-time data. Additionally, energy suppliers do not have the same protected status as network operators.

‘It is not all the clients too anyway, but it is also not the actual so we don’t know now the usage of our client a minute ago, so for that matter. […] That data is not available for the market so you need to install hardware so sensors and meters at clients.’ (J, Enerkie)

‘That’s why you need al that data, you need smart meters. So, that’s why the output of the smart meters, wherefore you can see that you have better control. […] But they are afraid that, what I could get out of data when you are on vacation, when you are away I get that all from your data.’ (G, Datacom)

The small consumers are protected in the Netherlands while the big consumers are not that intensively protected. That means that companies have to be really careful with the usage of data from smart meters. Additionally, this has to do with General Data Protection Regulation (GDPR) compliances which protects personal data. However, the business consumers are less protected and can be seen as business relations.

‘So, the consumer is the most protected market of the Netherlands, or the most protected market. So, that means that can’t be stored. The business consumer world is being stored, but that is because it is a business market and that is seen differently.’ (G. Datacom)

‘For instance, with smart meter data, first you must be there very careful with that because that is not allowed to do everything with it. Which is very logical because there is GDPR compliance on it.’ (R, Trapeco)

(27)

24

‘Currently, comes with for example recoding for sake of the must of anonymizing personal data. […] Yes, the personal data make is much more complicated. There are many people that in our domain data are working with insight, if you once know the trick applies more is better, that is the whole idea of being data-driven.’ (W, Trapeco)

Many companies know the trick, but the problem with storing the big data requires huge investments in data centres. Therefore, managers are in constant trade-offs about which data is needed and valuable for the company.

‘Yes, only, than are the companies actually going to pay instead of the consumers for the data to get the data. But yes, how are they processing that in the system. What kind of choices are made in there and when is the data, how do you say that? Have much value, when not, let me put it this way, at this one smart?’ (G, Datacom)

The trade-offs are managerial challenges which comes with a data-driven strategy. Big data is valued when you really can do something with it. If big data with BDA can give the needed insights, then big data is very useful and worth the money invested, but when that is not the case, it can be a bad investment.

(28)

25

5. Discussion

The findings show that the change in environment has led to a change in organizational structure, namely Agile and Scrum. BDA enables to perform this structure well and has led to a constant driven evidence-based MCS. Additionally, the findings show that BDA enables organizational learning by enhancing feedback and feedforward throughout the organization. However, the findings also show a disabling relationship between BDA and MCS which is explained by the lack of understanding and opportunistic behaviour of managers.

Historically, the energy sector was originally a civil service sector and has not been neither a competitive nor a dynamic sector. Due to the privatization, digitalization and the energy transition, the energy sector changed and is still becoming more data-driven, competitive and dynamic (Zhou et al., 2016). The findings show that all the case companies in the energy sector change their structure to Agile and Scrum with support of BDA to suit flexible strategies due to this competitive and dynamic environment. This confirms existing research that structures are dependent on key success factors and strategies (Ferreira & Otley, 2009) and, that key success factors with the competence to act fast on market conditions may call decentralization or designs as team-based structure (Johnson, Scholes, & Whittington, 2005). By adopting the Agile and Scrum structure, the case companies ensure more responsibility in the lower levels of the hierarchy to respond quickly and create bottom-up information flows. This study extends existing studies that suggest that BDA supports managers with taking decisions and therefore frame the business processes (Provost & Fawcett, 2013; Raffoni et al., 2018). Because the findings suggest that BDA ensures and supports the organizational structure (Agile and Scrum) by enabling bottom-up information flows providing the teams and managers with insight and information, that enable managers to monitor continuously. Since organizational structure is a crucial control in the organization (Ferreira & Otley, 2009), this paper offers an intriguing finding because it implies that BDA supports and helps managers by carrying out the Agile and Scrum structure in the energy sector.

(29)

26

towards more evidence-based decision-making is done (Ittner & Larcker, 2005). Overall, this study finds that when managers are able to design MCS upon a constant-event driven architecture, then it should support leadership and evidence-based decision-making in the energy sector. In this way BDA ensures the MCS to be dynamic and sensitive to changes in the internal and external environment and therefore, enable evidence-based real-time decisions. This is an interesting finding since Marchant and Raymond (2008) argue that MCS are not dynamic or sensitive to changes in the external and internal environment and that they lack being up-to-date, accurate and do not facilitate confident and fast decisions.

In light of the use of MCS, the findings reveal that BDA supports and enhances both uses of the MCS. In case of interactive control (feedforward) , this study confirms prior research (Raffoni et al., 2018) by showing that BDA provides forecast models and decision-options that assist continuously monitoring and promote dialogue throughout the organization. Moreover, by providing outputs that are negotiable, managers have help to review the underlying expectations of strategies and even to find new patterns (Ferreira & Otley, 2009; Raffoni et al., 2018). Henceforth, BDA is used as a strategic validity control. In case of diagnostic use (feedback), within the case organizations, BDA supports managers with correcting and motivating particular targets and goals by identifying blind spots, such as critical unidentified performance variables (Elgendy & Elragal, 2016; Raffoni et al., 2018; Warren et al., 2015). Consequently, the results indicate that BDA detects exceptional things and processes much earlier, which may trigger particular controls such as corrective action by detecting faulty measures or malfunctions. Thereby, this research confirms the current studies of (Kibria et al., 2018; Raffoni et al., 2018; Warren et al., 2015). It further contributes by adding that in the energy sector, BDA enables the case companies to use condition-based maintenance to detect malfunctions within the grid (Tsang, 1995). Therefore, BDA can provide hidden insights and supports continuous improvement by providing relevant and on-time information to managers and employees (Kaplan, 1990).

(30)

27

organizational learning. In this way, BDA is seen as a decision support system that promotes organizational learning (Bhatt & Zaveri, 2002). Therefore, this study provides an emerging contribution to the organizational learning literature (Argyris & Schön, 1978; Bhatt & Zaveri, 2002; Kloot, 1997; Senge, 1997) by providing evidence to the connection of BDA and organizational learning and thus, a topic for future research.

Finally, in contrast, the findings show a disabling effect of BDA on MCS. Consequently, the results indicate that managers and employees can have a hard time understanding the output of AI models or other BDAs. Thereby, this research adds to the current literature (Bumblauskas et al., 2017; Shah, Horne, & Capellá, 2012) that suggest that non-understanding of BDA is not understood leads to poor decisions. It does so, by providing results that show that BDA models work under particular conditions. When those conditions are not met, it is possible that management would face problems and makes wrong decisions based upon that model. As a result, the information flows are damaged and therefore, managers are facilitated with wrong information. This is problematic because actions that are recognized but rooted in false knowledge will barely yield predicted outcomes and may draw to unintentional negative outcomes (Bumblauskas et al., 2017). Therefore, BDA not only enables the diagnostic and interactive use of MCS, but under particular conditions, consequently disables the use of MCS by disrupting information flows which is caused by inadequate interpreting and data pollution. This is an interesting finding since there is a shortage of BDA specialists (Chen, Chiang, & Storey, 2018; Shah et al., 2012) and human talent is needed to perform tasks connected to BDA (Phillips-Wren & Hoskisson, 2015; Shah et al., 2012).

6. Conclusion

(31)

28

The findings show that the digitalization of the energy sector has its effect on the organizational structure of the companies. The changing environment changed the companies’ strategies and key success factors which were more based on flexibility and data. As a consequence, all the case companies changed their structure to a more dynamic and flexible structure to appropriate carry out the strategies. The findings show that BDA stimulates the performance of the structure by helping managers framing the business processes and enhancing efficiency, motivation and information flows. Thus, the change to a data-driven energy sector has led to a change in structure suitable for a data-driven strategy and BDA stimulated the performance of this structure. Furthermore, the findings show that BDA enables the MCS to be proactive and evidence-based. BDA supports the decision-making of managers by generating and identifying scenarios or decision options. This gives managers decision options and when the particular scenario occurs, managers are able to respond fast by enforcing the plan for the particular scenario. Moreover, BDA provides real-time data and objective information and therefore, support evidence-based decision-making. This enables managers to set-up the best course of action for targets and KPIs by improving for instance the measures of KPIs. Based upon scenarios and options, managers are able to design a MCS which is dynamic and support evidence-based and real-time decision-making. In light of the use of MCS, the findings indicate that BDA supports the diagnostic use (feedback) and interactive use (feedforward). In case of diagnostic control, the findings suggest that BDA supports managers with correcting and motivating particular targets and goals by identifying blind spots. Moreover, BDA provides concrete on-time information to managers and employees that support the diagnostic use of MCS and enables condition-based maintenance. As mentioned above, BDA provides forecast models and scenarios/decision options to assist managers with continuously monitoring and promote dialogue throughout the case organizations. Thereby, BDA enhances the interactive use of the MCS and can be used as a strategic validity control. In case of organizational learning and BDA, the findings show that BDA supports and creates feedback and feedforward information flows within the organization and thus, enables organizational learning. BDA facilitates knowledge creation by detecting exceptional things and provide frequent feedback. BDA is able to provide information that enhances the fit between the organization and the environment and therefore, enables organizational learning. Therefore, BDA can be seen as a decision support system that enhances organizational learning. However, the findings suggest also a disabling effect of BDA on the MCS. This is partly the reason because big data and BDA is in the beginning of its infancy. The findings indicate that managers and employees can have inadequate knowledge about BDA and can take wrong conclusions from the output of BDA models. Moreover, the data can be polluted due to touched and filtered data or opportunistic behaviour of managers. Therefore, BDA under particular conditions, consequently disables the use of MCS by disrupting information flows caused by inadequate interpreting and data pollution.

(32)

29

research. Therefore, the results are not statistical generalizable. Nevertheless, this is acceptable since the study is exploratory in nature. Therefore, future research could shed light in further investigating BDA in relation to MCS in different industries or contexts.

Correspondingly, this research investigated multiple aspects of MCS. However, reward systems were not mentioned within the interviews. Additionally, rewards systems are the result of performance evaluations (Ferreira & Otley, 2009). Altogether, future research could examine the relation between BDA and reward systems.

Due to the explanatory nature of this research the scope is limited. However, the results show an additional finding which is beyond the scope of this research. In addition, this finding may provide a fruitful future research topic. The findings suggest that organizations need to make trade-offs in which data they want to acquire. Due to GDPR compliances the data from smart-energy systems, such as smart meters, are protected and need to be recoded in sake of anonymization. Additionally, storing this anonymous data in data-centres is very expensive and it is even more expensive for companies in which data is not directly the primary goal and income. Accordingly, scholars find that privacy is a major obstacle and challenge of big data and agree that this needs to be investigated (Samuel Fosso Wamba et al., 2015; Idrees et al., 2018; Sivarajah et al., 2017). However, these scholars concentrated on the challenge of being safe and careful with big data and not about the challenge getting big data that is slacked by GDPR compliances. Finally, this is a fruitful future research topic to investigate.

(33)

30

Bibliography

Aalami, H. A., Moghaddam, M. P., & Yousefi, G. R. (2010). Modeling and prioritizing demand response programs in power markets. Electric Power Systems Research.

https://doi.org/10.1016/j.epsr.2009.10.007

Altin, M., Göksu, Ö., Teodorescu, R., Rodriguez, P., Jensen, B. B., & Helle, L. (2010). Overview of recent grid codes for wind power integration. In Proceedings of the International Conference on

Optimisation of Electrical and Electronic Equipment, OPTIM.

https://doi.org/10.1109/OPTIM.2010.5510521

Amin, M. (2008). Challenges in reliability, security, efficiency, and resilience of energy infrastructure: toward smart self-healing electric power grid. In IEEE Power and Energy Society 2008 General

Meeting: Conversion and Delivery of Electrical Energy in the 21st Century, PES.

https://doi.org/10.1109/PES.2008.4596791

Argyris, C., & Schön, D. (1978). A theory of action perspective. Reading (MA): Addison-Wesley. Beulke, D. (2011). Big Data Impacts Data Management: The 5 Vs of Big Data [Blog post]. Dave

Beulke Blog.

Bhatt, G. D., & Zaveri, J. (2002). The enabling role of decision support systems in organizational learning. Decision Support Systems. https://doi.org/10.1016/S0167-9236(01)00120-8

Bumblauskas, D., Nold, H., Bumblauskas, P., & Igou, A. (2017). Big data analytics: transforming data to action. Business Process Management Journal. https://doi.org/10.1108/BPMJ-03-2016-0056 Chen, Chiang, & Storey. (2018). Business Intelligence and Analytics: From Big Data to Big Impact.

MIS Quarterly. https://doi.org/10.2307/41703503

Chenhall, R. H. (2003). Management control systems design within its organizational context: Findings from contingency-based research and directions for the future. Accounting,

Organizations and Society. https://doi.org/10.1016/S0361-3682(01)00027-7

Choi, T. M., Wallace, S. W., & Wang, Y. (2018). Big Data Analytics in Operations Management.

Production and Operations Management. https://doi.org/10.1111/poms.12838

Cunningham, G. M. (1992). Management Control and Accounting Systems under a Competitive Strategy. Accounting, Auditing & Accountability Journal.

https://doi.org/10.1108/09513579210011880

Dezi, L., Santoro, G., Gabteni, H., & Pellicelli, A. C. (2018). The role of big data in shaping ambidextrous business process management. Business Process Management Journal. https://doi.org/10.1108/BPMJ-07-2017-0215

Diamantoulakis, P. D., Kapinas, V. M., & Karagiannidis, G. K. (2015). Big Data Analytics for Dynamic Energy Management in Smart Grids. Big Data Research.

https://doi.org/10.1016/j.bdr.2015.03.003

Edmondson, A. C., & Mcmanus, S. E. (2007). Methodological fit in management field research.

Academy of Management Review. https://doi.org/10.5465/AMR.2007.26586086

Eisenhardt, K. M. (1989). Building Theories from Case Study Research. The Academy of Management

Review. https://doi.org/10.2307/258557

Elgendy, N., & Elragal, A. (2016). Big Data Analytics in Support of the Decision Making Process. In