In control with data

Master Thesis

In control with data

How business analytics is involved in the design and use

of management control systems

By

Sari IJsseldijk

S3273571

University of Groningen, Faculty of Economics and Business MSc Business Administration: Organisational and Management Control

June 24, 2018

Supervisor: Dr. A. Bellisario Co-assessor: Dr. E.G. van de Mortel

1

A

There is growing interest in how technology improves the information flows holding together management control systems (MCS). Despite business analytics (BA) being one of the most influential technological innovations of current times, believed to provide numerous advantages to MCS, little empirical evidence exists since only few have touched upon the subject. Using case studies in ten organisations selected for their variation in industry and size, this paper explores the ways in which managers are using BA in the design and use of MCS. The paper concludes that BA enables visibility and awareness of performance, creating a feeling of responsibility and the opportunity for managers to hold employees accountable for their achievements. BA is however not only used as an alarm that rings when things are not working out. It is a resource used to enlighten management’s thinking in developing plans, that facilitates the proactive monitoring and acting on performance, as well as the execution of effective performance review meetings.

Key words: technological innovations, business analytics, management control systems,

2

T

LIST OF TABLES ... 2

1.

INTRODUCTION ... 3

2.

THEORETICAL BACKGROUND ... 5

2.2 BA: a technological innovation ... 6

2.3 The use of BA in MCS ... 7

3.

RESEARCH DESIGN ... 10

3.1 Research approach and sample selection ... 10

3.2 Data collection ... 11

3.3 Data analysis ... 13

4.

FINDINGS... 15

4.1 Enlighten thinking for backing up plans ... 16

4.2 Proactively monitoring and acting on performance ... 20

4.3 Effective reviewing ... 23

5.

DISCUSSION ... 25

6.

CONCLUSION ... 28

REFERENCES ... 31

APPENDIX. INTERVIEW GUIDE ... 38

L

T

3

1. I

Recent developments in information technology and digitalisation cause an explosion in the amount of data available for analysis. The term used to describe this growing volume, variety, velocity and veracity of data is big data (Visani, 2017). As written by Warren, Moffit and Byrnes (2015), this data can come in many forms, such as video, image, audio and text. Data can bring value to organisations with the use of business analytics (BA), which is currently one of the most influential technologies for organisations (Luftman & Zadeh, 2011) and describes the mathematical, statistical and econometric analyses of business data (Davenport & Harris, 2007). At the same time, organisations nowadays face increasing competition and rising customer expectations, forcing them to lower costs while improving quality and service (Umble, Haft, & Umble, 2003). New BA tools are able to assist by integrating data and thereby providing new insights for decision-making (Shollo & Galliers, 2016), altering how organisations operate and work towards their goals, eventually leading to better performance (McAfee & Brynjolfsson, 2012).

Several authors have pointed out the potential that BA has for management control (MC) purposes. The information in typical old-fashion MC systems1 _{(MCS) is often not up-to-date or accurate because}

it is not sensitive to changes in the firm’s environment, thereby prohibiting the organisation from making fast and confident decisions (Raffoni, Visani, Bartolini, & Silvi, 2018). The purpose of MCS is toguide employee behaviour in a way that increases the chances of achieving the organisation’s objectives (e.g. Flamholtz, Das, & Tsui, 1985; Otley, 1980; Merchant & van der Stede, 2017; Simons, 1995). Over the years, the information flows that hold together MCS have been improved by the development of technology (Dechow & Mouritsen, 2005; Maraghini, 2010). Tools like ERP systems provide a platform for information to flow and thereby enhance MC and decision making (Chapman & Kihn, 2009; Dechow & Mouritsen, 2005; Granlund & Malmi, 2002; Scapens & Jazayeri, 2003). BA is the most recent technological innovation in the area of using information within MCS and transforms the way information is collected and analysed (Warren, Moffit, & Byrnes, 2015). It can provide near real-time insights into several (financial) dynamics (Brands & Holtzblatt, 2015), thereby allowing organisations to move from intuitive to more evidence-based management due to more fine-grained and updated data (Raffoni et al., 2018). By using BA, MCS become more effective and advanced (Silvi, Bartolini, Raffoni, & Visani, 2012) and decision time decreases (Rouhani, Ashrafi, Ravasan, & Afshari, 2016).

Even though the possible advantages of BA seem clear, a lack of research is apparent in terms of how BA is involved in MCS. Some authors have suggested ways of implementing BA into MCS frameworks such as the Balanced Scorecard (Appelbaum, Kogan, Vasarhelyi, & Yan, 2017), while others suggest ways of implementing BA into the organisation overall (Brands & Holtzblatt, 2015).

1_{Note that some (e.g. Malmi & Brown, 2008) refer to combinations of MC practices as ‘packages’ while others (e.g.}

4

Since the way of using MCS can be more significant than their design (Ferreira, 2002), some have made suggestions as to how BA can be used in the design and operation of MCS (Mello, Leite, & Martins, 2014; Visani, 2017; Warren, Moffit, & Byrnes, 2015). Others have proposed potential changes in the control mind set (Nielsen, 2015; Nudurupati, Tebboune, & Hardman, 2016). However, these potential uses and changes are just that, potentials. A more helpful study on the involvement of BA in the design and use of MCS comes from Raffoni et al. (2018), who employed action research to explore the role of BA and thereby provide a five step framework on how BA can be used for performance management. Except for their conceptual framework, which is built on one case study, the potentials stated in the literature are not grounded in empirical evidence.

To bring insight into this novel phenomenon, the goal of this study is to contribute to the current body of literature on BA and MCS by exploring how BA is involved in the design and use of MCS. Several authors state the importance of further investigation and explanation of the design and use of MCS in relation to BA (e.g. Berry et al., 2009; Côrte-real, Ruivo, & Oliveira, 2014; Mancini et al., 2017), particularly in a single study (Visani, 2017). Since BA might be able to overcome traditional issues related to MCS, such as information overload, absence of cause-effect relationships and lack of a holistic view of the organisation (Raffoni et al., 2018), the examination of this phenomenon is of both theoretical and practical relevance. The eventual aim is to use the findings of this study to provide ideas for future research. By studying ten organisations that use BA in the design and use of their MCS, this paper answers the following research question: “How is business analytics involved in the design and use of MCS?”. The findings show how BA is used to enlighten management’s thinking in developing plans and uncover the use of BA in the proactive monitoring and acting on performance, as well as the execution of effective performance review meetings.

5

2. T

2.1 Technological innovations in MCS

MCS are operated by managers to keep their organisations reliably on track. As written by Merchant and Otley (2007), an organisation that is ‘in control’ is likely to achieve good performance against its objectives, regardless of what these objectives are. MCS come in many forms, from simple operating procedures to elaborate processes such as performance evaluation reviews on the efforts of managers over time (Merchant & Otley, 2007). These authors state that both the structure and the operation of MCS are of importance for their effectiveness. According to Ferreira (2002), the way of using MCS can be more significant than their design, so it is not only the instrumental but also the behavioural side of MC that is important (Waal, 2007). In their performance management framework designed for guiding the efficient obtainment of a holistic overview of MCS, Ferreira and Otley (2009) describe the design process of MCS as: (1) vision and mission, (2) key success factors, (3) organisational structure, (4) strategies and plans, (5) key performance measures, (6) target setting, (7) performance evaluation and (8) reward systems. This shows that MCS support a broad range of managerial activities, including strategic processes and ongoing management. In terms of operating the system, Simons’ (1995) research and subsequent contributions (e.g. Chenhall, 2007; Tessier & Otley, 2012) show that in using MCS, information is mobilised in different ways. Any MCS can be used diagnostically and interactively, where the former is a traditional feedback style for motivating, monitoring and rewarding achievement of goals based on critical performance variables and the latter is a feedforward use to stimulate organisational learning and the emergence of new ideas and strategies. Interactive control can be seen as comprising five areas, namely intensive use by senior managers, intensive use by operating managers, pervasiveness of face-to-face challenge and debate, focus on strategic uncertainties and a non-invasive, facilitating and supportive involvement (Bisbe, Batista-Foguet, & Chenhall, 2007).

6

information flows and the subsequent faster and easier access to operational data (Maraghini, 2010). For instance, Malinic and Todorvic (2012) found that the use of SAP reduces the time managers have to spend on traditional management accounting tasks, while increasing the time required for analysing data, measuring performance and strategic reporting. These authors also state that the logic of accounting and control often becomes a subject of evaluation when major IT changes happen. Dechow and Mouritsen (2005) point out that MC in an ERP-environment is no longer only in the hands of the accounting function, but a collective affair where local control issues are used to create notions of global management. Even though several studies have attempted to study how technology impacts managerial reporting and control, these studies have only opened the discussion (Granlund, 2010). It is however clear that technological innovation is a key overarching variable to be considered in the design and use of MCS (Chenhall & Moers, 2015).

2.2 BA: a technological innovation

A recent technological innovation in the area of using information within MCS is BA. This innovation is concerned with evidence-based problem recognition (Holsapple, Lee-Post, & Pakath, 2014) and supports decision making while allowing a better understanding of the business and its markets (Chen, Chiang, & Storey, 2012). BA allows organisations to go beyond traditional business intelligence (BI) reporting (Laursen & Thorlund, 2017), which is characterised by the creation of simple rules and the distribution of known facts to systems and people (Nielsen, 2015). BA focusses on developing new insights based on data, whereas BI provides historical, metric-driven decision making based on for instance the number of units sold. BA includes the BI approach that is focused on the past, but also allows for looking forward. To capture both, the overarching term BA is used for this research.

Organisations are using data in several ways (Evans, 2013; Nielsen, 2015; Raffoni et al., 2018; Visani, 2017); two-thirds of the use is predictive, 25% is prescriptive, and the remaining is descriptive and diagnostic (Amani & Fadlalla, 2017). The most simple type of BA is descriptive analytics, which is the use of past and current data to address what happened or what is happening now by identifying patterns. This can be called BI. Diagnostic analytics provides a deeper understanding of why something has happened and is characterised by techniques such as drilling down into the data. Predictive analytics provides a look into the future by addressing the question of what may happen and delivers forecasts and predictions. Prescriptive analytics actually shows managers what they should do next, which results in performance management rules and recommendations based on the best possible alternative given different scenarios. The last two forms of analytics use techniques such as artificial intelligence, machine learning, mathematical models, data mining and data modelling (Kumar, 2017).

7

2011). These authors state: ‘Using big data leads to better predictions, and better predictions yield better decisions’ (p. 1). Before big data, decision-makers relied too much on experience and intuition. Now that all of these data are available for decision-makers to use, organisations move from intuitive towards more evidence-based management (Davenport & Harris, 2007; Ittner & Larcker, 2005). BA can monitor and assess the competitive landscape of the organisation and give insights into business processes and bottlenecks to improve problem solving and process optimisation. It enables organisations to better understand their business and markets (Chen, Chiang, & Storey, 2012). For performing BA, several tools are available which in essence facilitate data collection from systems, process analytics and create displays and reports (Schlegel, Milbury, Buytendijk, & Sommer, 2014). Examples are advanced analytics systems such as ‘R’ to run analyses, as well as dashboards with drill-down options, that stimulate the use of information by summarising and visualising data (Peyrot, Childs, Doren, & Allen, 2002). The ability of BA tools to dig deep into the source data and the flexibility to change the analysis’ parameters are what make BA a game changer for organisations (Brands & Holtzblatt, 2015).

Since BA provides more accurate and detailed performance information (Fanning & Grant, 2013), it is a useful tool for performance management and control purposes. Several authors have shown that BA can reveal the drivers of performance and can explain deviations from targets (Ittner & Larcker, 2005; Silvestro, 2016). It allows managers to become more responsive to risk and uncertainty and can increase performance awareness (International Federation of Accountants, 2017). The new features BA brings to the table change the way managers can use information in the design of MCS and can alter the execution of MC for good (Granlund, 2010).

2.3 The use of BA in MCS

In this era of modern enterprise systems and the possibilities of using information presented by BA, managers and management accountants ‘can do more than simply monitoring and tracking key indicators of historical financial reports’ (Appelbaum et al., 2017, p. 34). MCS of the past were largely ‘proximity controls’, based on assumed relationships between aspects of performance that are not based on proper data (Silvestro, 2016), whereas now a trend towards evidence-based control is apparent (French, 2014). The challenge for organisations to get value from BA is to process the data into meaningful information to enable decision-making and control (Bititci, Garengo, Dörfler, & Nudurupati, 2012). BA improves the information and its flow dramatically and can therefore potentially alter the execution of MCS for good, by providing the means to better management of performance (Granlund, 2010). For organisations to get value from their data with BA, MCS must be designed to gather, absorb and leverage these new information flows (Elbashir, Collier, & Sutton, 2011). However, the ideas presented in the literature are mostly conceptual and little empirical evidence exists with regards to the phenomenon of using BA in MCS.

8

strategy (Mello, Leite, & Martins, 2014) and can make MC more about the need to ask the right question (Nielsen, 2015). Its use is said to enable the identification of relationships between performance measures and the discovery of new ones (Belfo & Trigo, 2013; Ittner & Larcker, 2001; Mailliard & Mulhall, 2017; Warren, Moffit, & Byrnes, 2015). BA can support both planning and cybernetic controls, as well as administrative and reward controls (Elbashir, Collier, & Sutton, 2011). In terms of the involvement of BA in the use, or operation, of MCS, it is said to aid reporting and performance evaluation by using internal data (diagnostic control) and enable communication between departments, the reassessment of strategy and detection of strategic uncertainties by using real-time unstructured data (interactive control) (Warren, Moffit & Byrnes, 2015).

These potential advantages and uses are just that, potentials, not grounded in empirical evidence.A more helpful study on the involvement of BA in the design and use of MCS comes from Raffoni et al. (2018), who employed action research to explore the role of BA for MC and for operationalising what they call Business Performance Analytics. This term refers to ‘the management and control of the firm’s strategic dynamics and performance through the systematic use of internal and external data and analytical methods’ (Silvi, Möller, & Schläfke, 2010, p. 51). The five step framework they provide shows how BA can be used for performance management, and is based on the framework of Ferreira and Otley (2009) mentioned earlier. The framework shows the steps organisations must take to use BA for identifying causal effects between measures and how this information can be mobilised in the process of using MCS. Based on their case study, these authors agree with others on how BA can be used for MC. In terms of diagnostic control, these authors found that BA can help make critical performance variables visible and provide evidence of hypothesised links among factors underlying the causal performance model. This enables managers to measure results, analyse and eventually reward and correct actions. In terms of interactive control, BA contributes by identifying strategic uncertainties, questioning strategic assumptions, promoting the discussion of current strategies and emergence of new strategies. According to these authors, controls are used more interactively and are used as strategic validity controls by forecasting consequences of events. BA encourages dialogue and can contribute to identifying potential sources of risk.

9

10

3. R

3.1 Research approach and sample selection

Previous literature has stated potential uses and advantages of BA for MCS. However, empirical evidence is scarce, with only two case studies conducted on the subject. In light of the lack of prior empirical research, a qualitative method is best suited for this study and therefore used to explore the phenomenon in depth (Carlsen & Glenton, 2011; Edmondson & McManus, 2007). In order to investigate how organisations are using BA for MCS, a case study design was adopted, to focus the exploration on BA in its real-life context (Yin, 1999, p. 1211). The study is cross-sectional, which means no long-term causal effects were studied (Patton, 2015, p. 255). In line with recommendations of Raffoni et al. (2018) to conduct future research on the BA phenomenon in different contexts, a multiple case design was chosen. The setting of this study was a small sample of ten firms in the Netherlands. As written by Chae, Koh and Park (2018), it is widely noted that the role and use of technology differ with organisations’ sizes and industries due to for instance structural properties and historical intensiveness of IT use (Buonanno, et al., 2005; Chiasson & Davidson, 2005; Mabert, Soni, & Venkataramanan, 2003; Masli, Richardson, Sanchez, & Smith, 2011; Zhu, Li, Wang, & Chen, 2010). As an example for BA in specific, Torres, Sidorova and Jones (2018) state that BA may be enacted and used in different ways for different purposes in different industries and therefore ensured adequate industry representation in their study. Furthermore, within organisations, MCS are not always used in the same way in different hierarchical layers (Ferreira & Otley, 2009). This implies that the use of technology for information flows possibly differs across hierarchical levels. Given these arguments, the cases were selected based on purposeful sampling (Blumberg, Cooper, & Schindler, 2014, p. 174; Patton, 1990, p. 169). The ‘logic and power’ of purposeful sampling lies in the quality of information obtained per sampling unit, not in the number of units included (Sandelowski, 1995). To construct a holistic understanding of the BA-MCS phenomenon in light of the above written evidence (Benoot, Hannes, & Bilsen, 2016; Suri, 2011), a maximum variation sampling procedure was used to select cases that differ in terms of industry and size, and select participants working in different hierarchical layers. This procedure is applied to find shared patterns that cut across cases and derive their significance from having emerged out of heterogeneity (Patton, 1990).

11

extra five organisations from five other industries were added to the sample (Patton, 2015, p. 313). According to Verbeeten, Kolthof and Steenwijk (2017), BA gets little attention in the public sector. However, interestingly enough, a university was found that uses BA in controlling the organisation, which was therefore selected for this study. Besides that, a financial institution and an insurance organisation were specifically searched for and selected, since according to the same authors, organisations in these industries are working with analytics to a great extent. Start-up organisations were not included in the sample, since even though these may use large amounts of (un)structured data, the absence of several controls in these organisations is believed to not be suitable for this study (Sandelin, 2008; Sandino, 2007). In the end, the study design involved a total of ten cases from ten different industries (Table 1).

Informants were identified through snowball sampling, by asking at least two people within a firm about the right person to speak to on this subject matter. Participants were selected because of their direct knowledge of using BA in MCS (Sandelowski, 1995), so they were likely to be able to generate rich information on the research topic (Walsh & Downe, 2006). In light of the maximum variation strategy, informants from different hierarchical levels (e.g. directors, managers, management accountants, data scientists) were selected. Data gathering was terminated when interviews in new cases did not lead to new information (Lincoln, Guba, 1985, 202). To gather in-depth information and be able to demonstrate the robustness of eventual findings (Maxwell, Rotz, & Garcia, 2016), the sample included a total of 15 professionals from ten organisations.

3.2 Data collection

12

Table 1. Description of case data

Organisation Industry Size Hierarchical layer No2_{. Role Duration Observation/data}

DutchUni Education 500 - 1.000 Corporate 1 Board staff 2 hours Performance

dashboard3

Finis Insurance services 5.000 - 10.000 Corporate 1 Management accountant (MA) 2 hours -

FlyUnited Aerospace 1.000 - 5.000 Business unit 1 Revenue manager 2,5 hours Perf. dashboard

Monus Financial services 10.000 + Corporate 1 C-level finance manager 2 hours Perf. dashboard

OneShop Consumer goods 500 – 1.000 Business units 2 Logistics and customer-service manager 3 hours PP of meetings4

Shipon Logistics 10.000 + Staff unit & business unit 2 Data manager and MA 2,5 hours -

SourceD Customer contact 10.000 + Business unit 1 Customer service manager 2,5 hours Excel calculations

TechNow Electronics 10.000 + Staff unit 2 Advanced analytics director and data

scientist

2,5 hours Perf. dashboard PP of meetings PDCA cycle

TelCo Telecom 5.000 – 10.000 Staff unit 2 Finance director and corporate MA 2,5 hours Perf. dashboard

WeTrade Wholesale 1.000 – 5.000 Corporate and business

unit

2 Operations director and corporate MA 3 hours Financial systems

2 _{Number of main interviews. Follow-up interviews are not included in this table, but were conducted with all organisations, except for FlyUnited and Monus.} 3 _{Dashboards showing (near) real-time performance of parts of the organisation.}

13

The main case organisations were the five completed at first, in which I conducted two interviews per case. The informants were very knowledgeable on the subject (minimum of three years of experience on the subject), the interviews took a minimum of two hours per case, and I was allowed to analyse several documents and make observations. This richness in data collection allowed me to only require talking to one or two people at each case organisation to gain rich insights. The analysis and collection of data was performed simultaneously and after the first five cases I realised that a common pattern could already be identified. Therefore, I chose to only conduct one interview per subsequent case organisation to be able to see whether differences would appear across industries. The knowledge gained in the first ten interviews allowed me to properly identify who would be most helpful to interview in subsequent case organisations.

Prior to data collection, a pilot interview was conducted to gain broad insights into the phenomenon under study. The informant for this pilot interview was a senior consultant specialised in the use of BA in finance, named David De Jong (+10 years of experience). Based on the pilot interview, an interview guide was constructed to narrow down the inquiry into the phenomenon. The questions were structured according to the two themes described in the theoretical background, namely design and use of MCS. The main questions were inspired by frameworks of both Ferreira and Otley (2009) as well as Raffoni et al. (2018). By asking questions based on Ferreira and Otley (2009), a broad view of the key aspects of both MCS design and use was gained. The questions based on Raffoni et al. (2018) allowed me to focus more specifically on the use of BA within MCS. Based on the think-aloud technique, the questions were tested with one master student with - and two people without - knowledge on BA and/or MC to make sure the wording of the questions was clear and questions were easy to follow (Patton, 2015, p. 486). Then, these questions were piloted with the same senior consultant to test whether the example questions would provide proper insights and would trigger people to tell interesting stories. The final interview guide is included in the appendix and includes example questions which were used as a checklist at the end of the interviews to see whether all of the main topics had been discussed. The interview guide was not used extensively, since the aim was to have very open conversations and let the informants speak. The themes were sent to the informants a week before the interview, so that they came with focus. The interviews were conducted anonymously and the concepts of the study were clearly explained. To protect anonymity of the informants, information from prior interviews was not shared with informants.

3.3 Data analysis

14

subsequent interview. Archival documents such as the PDCA cycle from TechNow were used only to gain a general insight into the situation. The goal of the data analysis is to identify patterns, rather than testing propositions or constructs (Edmondson & McManus, 2007).

15

4. F

This chapter is structured based on the process of MC as it was described by informants. It shows how BA is used in designing MCS and what the effect of its use is on the operation of MCS. The table below shows the data structure that was built from the analysis of the data gathered during the study.

Table 2. Data structure

First-Order Concepts Second-Order Themes

Aggregate Dimensions

- Prioritise KPI’s to focus on the most important ones - Use data to move towards leading indicators - No more focus on financials, they are only the result

_{Causality between}

indicators

Enlighten thinking for backing up plans

- Utilising data to get from strategy to targets - Forecasting to determine targets

- Combining experience with past performance data

Data-based performance levels

- Link data only on the basis of functional questions - Everyone still uses their own system (legacy) - Combining data from all systems has no meaning

Data does not speak for itself

- Managers have to sell their plans with data - Information asymmetry has decreased - I better have my story straight - All performance is visible, all the time

Managers are held accountable

- One easy overview allows focus on outliers - Proactive management of external environment - Combine outliers with trends to filter out incidents

Manage by exceptions

Proactive monitoring and acting on

performance

- Less time spent on plan versus actuals - On-the-spot problem solving - Everyone is aware of their performance

Looking forward

- People have focus and know what to discuss - Visualisation helps interpreting data

- Shift from cross-examination to solving problems

Focus on what matters

Effective reviewing

- Defend against and with BA when not on target - Managers come to the meeting prepared - No discussions on the numbers - One fact-based truth

No more smokescreens

16

The findings show that the general impact BA has on the case organisations is that it creates visibility and awareness of performance. This enables higher level managers to have a quick overview of how the organisation is doing and to hold middle managers accountable for their performances. The findings on the first MC activity, showing that managers are enlightening their thinking with data for backing up plans, show evidence concerning how the MCS is built up with the help of BA, which concerns MCS design. The second theme of this study, the use of MCS, is described by the ways of using BA for monitoring, acting and reviewing of performance. The involvement of BA in the three MC activities is described in the subsequent paragraphs of this chapter.

4.1 Enlighten thinking for backing up plans

Based on the interviews, it is clear that BA is essentially used by middle managers to back up the plans that they have to hand in every year to higher management. BA is used to find causalities between Key Performance Indicators (KPI’s) to make sure middle managers focus on the right measures and to identify a reasonable height of targets. This is the process of designing MCS. For all case organisations, this process starts with strategy. The board of directors formulates a strategy in line with the organisation’s vision, which consists of several strategic pillars, but is still vague and qualitative. Examples are improving employee satisfaction, keeping costs down and growing revenues. These broad goals define what the rest of the organisation focuses on during the year.

‘Going top to bottom, we move from the “what” and “why” to the “how”.’ (N, SourceD)

On this higher level of the organisation, goals receive a target (e.g. 4% revenue growth) based on external expectations and internal ambitions. This is not done based on BA, because the goals are very broad, which means managers would need to combine all of the organisation’s data to get insights into a feasible target. This is not (yet) done in the case organisations. However, at middle management level, these overarching goals are translated into department KPI’s with the use of BA.

‘If the board states externally that we want to grow with 4%, that is translated to the business: how much should our business units grow, and within a unit, how much should each product grow? […] How do we reach these targets and what programmes do we need for doing so? […] It is translated into real operational plans.’ (E, TechNow)

17

‘I need to support my plans with data, so my analyst looks at trends to see what is possible and I come up with the programmes needed to improve my business. […] I better have my story straight.’ (B, OneShop)

In essence, middle managers use BA for (1) identifying which KPI’s to focus on for achieving the overarching goal, (2) coming up with the targets to put on these KPI’s, and (3) using these data-based plans to back up and essentially sell their story to their higher level managers. In order to effectively use BA for these purposes, the data needs to conform to certain criteria. The most important criterion is that ‘everyone has to speak the same language’, as stated by informant A from OneShop. It is crucial that definitions are clear and the same definition for a certain KPI is used across the organisation. Another key aspect of data is its quality. ‘Garbage in, garbage out’, as stated by informant A from OneShop, covers the essence of this criterion. This means that BA cannot transform low-quality data into high-quality insights. Besides, informant F from TechNow described how one wrong outcome of a model means that people never trust analytics again. So, for using data as the key argument for strategic decisions, it is crucial that people within the organisation can and will trust the data. Furthermore, a certain urgency and usefulness, such as the need to back up plans, need to be present for BA to actually be used. This was illustrated by an informant:

‘My department has built several models that could have really been valuable, however people did not trust the data, saw using the model as only taking extra time, and the urgency for the model would fade. […] The models were never used.’ (C, Shipon)

Given that these criteria have been met so that BA can be used effectively, the first step in the process of moving from strategy to measurable targets for the annual plan is choosing which KPI subjects to focus on. For identifying KPI’s, middle managers aim to find the indicators that show the achievement of certain goals. In essence, it is about finding causalities between variables. For instance, based on logic or experience, one might think focusing on customer satisfaction as a KPI will lead to an increase in sales. With the same reasoning, customer satisfaction might be reached by any number of factors, such as high quality products, quick delivery times or excellent service. In the past, these causalities between variables were assumed causalities, determined based on experience and gut feeling. However, managers would end up with long lists of un-actionable KPI’s, causing them to lose track of what is really important. Besides, this allows a focus on KPI’s without being sure these are actually the ones organisations should focus on.

‘We thought student satisfaction relied heavily on the number of electrical outlets in the school, so we had several projects for improving this. […] Ten years later, it turned out this was not actually important to students. We had been using the wrong KPI’s all along.’ (M, DutchUni)

18

activities must be undertaken to achieve this desired output (e.g. having high-performing processes). BA helps in the sense that it integrates middle managers’ thinking in selecting these performance measures. They use a combination of data and experience, which allows them to really think through their plans and ideas and subsequently use data to back them up. Where the focus used to lie on outputs and managers were ‘bashing on the numbers’, BA has allowed organisations to shift their focus towards leading indicators. This is helpful according to the informants, because even though lagging indicators are easy to measure, they can sometimes paint the wrong picture when considered in isolation.

‘If a department has very low sales, you would say they are doing badly. However, looking at the leading indicators, namely the scheduled orders to come, you might see that there were troubles with a political situation in that country, causing orders to pile up, which means next month will be a great month in terms of sales.’ (F, TechNow)

Informants describe that they now see that financials are only the result, not the actual goal to focus on. BA is used to figure out which lagging indicators to focus on, but also which indicators lead to a certain lagging goal. Organisations are making sure that the right activities are in place to ensure the right outcomes.

‘Translating goals to KPI’s in a systematic way starts with lagging, so: how can you determine whether you have been successful in the last period? And the next question is: how can you predict whether you will reach your goal? Or whether issues will arise? Which concerns leading indicators.’ (E, TechNow)

The analysis for testing causality between variables is conducted by the analytics department. A manager that is working on his or her plan for the upcoming year, goes to the analytics department with specific questions. The analytics department then determines what data they will need. For example, if an organisation’s strategy is to grow in terms of customer numbers, what data do you need to analyse to figure out what the leading indicators are and what a feasible KPI target is? Because the case organisations have existed for over ten years, the number of systems they have has grown, hindering the analytics teams from combining all data sources and creating one space where data comes together. This is called “legacy” and was mentioned by informants from five organisations as a large issue. This causes analytics departments to take things “project per project”. They do not try to keep perfect records of every imaginable subject, but search for specific data when a project asks for it. For example, when a middle manager is wondering whether a certain relationship exists between two or more variables. As stated by informant H from TelCo, the collection of data can start with ‘I believe that if I see X, and also Y, then probably Z will occur’. This suggests that the meaningfulness of data is still given by people, since they are searching for specific data to use for backing up their plans. They use BA to enlighten their own thinking. This was illustrated by an informant:

19

To figure out what leading indicators are, five case organisations are using BA to experiment.

‘If I turn a button here, what happens elsewhere?’ (A, OneShop)

Informants describe that they make small adjustments to see what happens and what the effect is on the end result. DutchUni for instance experiments with putting extra money in the HR budget to see what the effect is on the yield of the students. The analysis of data is what makes the data intelligent and brings the eventual business insights. As an informant remarked:

‘Someone from HR, finance, legal, will all look at their own data. […] Analytics is the cement that brings everything together.’ (M, DutchUni)

Once middle managers have determined which KPI’s to focus on, the second step is determining the performance level to be achieved for these KPI’s, so the targets to put on the measures to eventually get to the higher-level goal. In the annual plan, middle managers need to come up with their own expectations of what they think they can do for the coming year. Previously, middle managers would be tempted to protect themselves by handing in low targets. Managers were able to do so, because their superiors would not be able to judge these targets in a profound way due to information asymmetry. This has changed now that BA is used.

‘Managers need to defend against what analytics is saying, so they cannot just stand back. No more sandbagging.’ (F, TechNow).

The process of determining performance levels was illustrated by informant M from DutchUni to be the following:

‘What was the target last year? Do we want to improve? If yes, by how much? And is that realistic?’ (M, DutchUni).

For determining whether a certain target is realistic, both forecasts as well as experience and knowledge are used as a cross-check on the historic data. Forecasts can be conducted by normalising the data, so getting rid of the outliers, however that does not provide a proper forecast. BA is now used to recognise correlations and trends in the outliers and include these in the forecast.

‘The most important thing is that analytics enables us to look forward for more than 12 months. That is difficult to do for business people, because they think in terms of customers and do not know who their customers will be a year from now. Analytics can really say; if this happens as you plan it, this year, you will have trouble in the beginning of 2019. So then we can still fix it!’ (F, TechNow)

20

levels to be reached. Therefore, a middle manager cannot hand in a target that is too low or say they will reach a high target with a certain improvement programme without proper evidence.

‘If I want to get my personnel costs down, I need to lower my absenteeism. If I say I will lower that KPI by 3%, my boss and finance will ask me how I will manage to do that.’ (B, OneShop)

In the same vein, a higher level manager is not able to stretch a target too much, because the data will show that the target is unachievable. If a middle manager expects to only sell X euros of a certain product, even though that is worse than previous trends, and can back that up with data, directors will accept the number even though it is below expectations.

To be able to communicate and sell their plan to higher level management, middle managers combine the historic data and forecasts with their business experience. This is essential, because the numbers do not speak for themselves and therefore a point of discussion can be how to interpret them. This was illustrated by an informant:

‘If the numbers are good in Quarter 1, someone looking at an overall trend might say Quarter 2 will be good too. However, someone in operations knows the details and will say: we had a lot of luck in Quarter 1 so Quarter 2 will not be as good.’ (E, TechNow)

It is clear that someone who is not aware of daily business, such as a director, is vulnerable to making wrong conclusions based on data alone. Middle managers are close to the business, aware of the situation in operational teams and can therefore interpret the data in a sharper way. Therefore it is important to merge the overall trend that directors see with the more detailed view of people in the field, to get real insights into what a feasible target will be. This shows that even though BA can bring exciting new insights, knowledge and experience are still key for both identifying leading and lagging indicators, as well as their targets. It is the combination of BA and human intelligence that provides value.

4.2 Proactively monitoring and acting on performance

While the above paragraph describes how BA is used by middle management for strengthening their story for the annual plan, the interviews and observations also showed that BA is used throughout the year, for monitoring and taking action on performance. For monitoring performance, BA is used in the form of real-time dashboards, to keep track of the established KPI’s and to look at trends on a daily basis. The use of dashboards creates a sense of responsibility because it makes people aware of their performance. Besides, dashboards allows managers to easily spot and prioritise issues, which allows for quickly taking action on these issues. This also influences the way performance review meetings are carried out, as will be discussed later in this chapter.

21

there was no (near) real-time overview of the organisation’s current performance to monitor during the week. Performance data was only collected before meetings to bring along a print-out to discuss. Besides, managers focused on financial performance because financial data was easiest to collect and analyse. Nowadays, dashboards are used in several layers of the case organisations. People in operations of Shipon use a dashboard to get an overview of their area with one push of a button, and the board of directors of TelCo does the same with a dashboard of the whole organisation. The key is that there is one overview of the organisation, whether it is a unit, department or an area, that shows financials, non-financials and external information.

‘Everything is in that overview, from financials to FTE’s and absence percentages.’ (L, Finis)

Dashboards are applied in different forms, such as TV screens in an open office, where whole departments can continuously see their performance. The dashboards continuously show how the department is factually doing. Dashboards are used for several aims, such as showing the number of defaults in the network for TelCo, how much was sold per customer for WeTrade, and waiting times for customers of SourceD. The dashboards also show a comparison of current performance with the plan, budget, forecast, a month ago and/or the same day a year ago. Apart from looking at internal performance, dashboards used by case organisations also show data on the external environment. For instance, during my visit at TelCo’s office I saw their dashboard, which included an ‘outside in’ view. Several case organisations use this feature to map opportunities and threats, to be discussed in performance meetings. This allows managers to easily keep track of not only internal performance but also the external environment of the organisation.

‘This screen shows everything that is on the internet about our industry and our company, it shows a sentiment analysis of all competitor brands compared to ours and shows a map of the Netherlands with all network defaults.’ (G, TelCo)

Something repeatedly stated by informants was that a main effect of BA is therefore awareness of peoples’ performances. This influences how managers monitor their teams, as was stated by an informant:

‘It is not my role to continuously measure people’s individual performance, they are more than capable of doing that themselves.’ (K, FlyUnited)

22

‘The truth is laid on the table for everyone to see, but it is not used as a stick to hit someone with.’ (N, SourceD)

The awareness of performance establishes a feeling of responsibility for employees. Because the overview is already showing people how they are doing, it allows managers to quickly move towards problem solving when a KPI is off target. When a problem occurs, meaning a KPI goes red during a certain period or a target has not been reached at the end of the month, BA is used to take action. Managers have a dashboard in their office, to be able to go deeper into the data on their own. The drilldown option these dashboards come with, allow managers to click on a certain KPI to go deeper into the data and discover why something has happened. By looking at trends, one can tell the difference between major and minor issues. All case organisations combine focusing on red and green lights with looking at trends in the data, since a KPI can be in the red for any number of reasons. Informant A from OneShop describes:

‘We noticed that we had a problem with inventory almost every Saturday at a certain hour, so realised we had to do something and that it was not just an incident.’ (A, OneShop)

When a problem occurs, managers ask the analytics department to help them solve their problem by collecting and analysing data. This problem solving includes ad hoc on-the-spot meetings within a department to discuss the issue and potential solutions.

‘For these problem solvings, I notice people are coming to analytics more and more: do you have data for this? And for this?’ (F, TechNow)

Forecasts can help in solving problems by predicting what is to be expected in the foreseeable future. An example of how BA is used in the form of forecasts for solving problems on a daily basis was given by an informant:

‘I have to get a 50% service level every day. If my service level is 45% at 16:00, the question is whether I will still make the 50%. I can let more people come to work in the evening to try and raise the service level, but that only helps if I will still get a high enough number of callers to be able to improve to 50%. I then use data to analyse the issue and solve it.’ (B, OneShop)

23

4.3 Effective reviewing

The increased performance awareness and the allowed focus on problem solving rather than looking back, has caused a change in performance review meetings. Several performance meetings are held in the case organisations. First of all, managers from four case organisations start the day or week with a short meeting with the help of a stand-up screen. Just to quickly discuss what is going on, what people expect for that day, how the team is doing, what projects are running, what the issues are and what the actions are to solve those issues. Besides, as described before, spontaneous meetings are held to try to solve small problems on the spot with the help of BA. Furthermore, a more official weekly or monthly review meeting is held to go through the performance of an entire department. This meeting is held by the directors and the middle managers responsible for the different departments. The implementation of BA in review meetings is realised by conducting trials in the top layers of the organisation. Organisations are testing whether it works there, and if it does, it is cascaded down into the organisation. For example, TelCo and TechNow redesigned their board of directors meetings by implementing the use of dashboards. The content of these meetings has changed dramatically with the use of BA, as illustrated by an informant:

‘The monthly meetings used to be a cross-examination. Managers would send a management letter, we [directors and management accountants] would read it, and ask questions on all 6 cm of printed paper. Analytics enables focus and serves as a point of reference. […] There is one truth only and that is clear for everyone.’ (G, TelCo)

The dashboards are used in the meetings to visualise performance, in the form of graphs, but also in the form of keywords that come out of text analysis. This brings focus, because people know what they are talking about and what they should be talking about. A main consequence of using BA is that during the performance meeting, managers have to explain themselves when they are not on target. As illustrated by an informant, this was not always the case in the past:

‘Before we used this dashboard, managers would simply say they did not recognise the numbers that were given to them by finance, to get away with bad performance. That is no longer possible.’ (G, TelCo)

These smokescreens used to mean that nothing really happened when actually performance was bad and needed attention. As described by all informants, BA helps in the sense that it creates one truth and, as stated earlier, facilitates performance awareness. Organisations make sure that what is in the dashboard is correct, and as stated by informant G from TelCo, someone who says they do not recognise or trust the data is not taken seriously.

‘People obviously get to have an opinion, but there are no alternative facts.’ (M, DutchUni)

24

feeling, which significantly decreases the amount of discussions held on the numbers. This makes it easier to manage the organisation and allows for effective meetings. As stated before, BA creates a sense of responsibility, because everyone knows what they can expect, what they have to do, and know their performance will be measured. Managers take the time to really prepare before going into the review meeting, so that they are able to have a fruitful meeting. Informant N from SourceD for example stated that when he is not on target, he uses correlation analysis to show that an incident has had an influence on a KPI, to argue why he did not meet his target. If the reason for not being on target is not just an incident, but a proper issue for management, this is addressed during the meeting. Because there is no need to elaborately discuss good past performance, reviewees are aware of their performance and they have identified counter measures, the meeting focuses on solving problems.

‘During the meeting, managers talk about their trend, about what really causes the problem, and counter measures, the impact the counter measures are expected to have and how it closes the gap. This includes owners, timing, and so on. We monitor every month by how much the gap has closed.’ (F, TechNow)

Overall , any type of performance meeting is moving away from looking backwards, which is already done during the month with monitoring, towards discussing problems, how to solve them and forecasting. Besides, the subject of the review meeting tends to change due to newly (real-time) incoming data.

‘If a competitor is suddenly doing really badly, we discuss whether we should do something with that information.’ (G, TelCo).

Based on external information such as the earlier described outside-in view from TelCo, managers discuss where opportunities might lie or where threats have to be tackled. BA facilitates data-driven review of both internal and external performance, but above all helps in bringing the organisation towards its goals.

25

5. D

The findings show that BA enables visibility and awareness of performance, thereby allowing managers to use data to enlighten their thinking for backing up plans, proactively monitor and act on performance and have effective review meetings. In this section, I will discuss these findings in light of the current body of literature on the subject.

Due to information asymmetry, middle managers used to have an information advantage over their superiors, limiting the ability of these superiors to adequately assess plans and decisions of their subordinates and carry out MC. This study shows that the availability, visualisation and analysis of data with the use of BA increases visibility of people’s performances across the organisation, thereby significantly decreasing information asymmetry between management levels. This enables higher level managers to perform MC in a more effective way, because more accurate and detailed performance information is available (Fanning & Grant, 2013). Higher level managers can now hold middle managers accountable for their performance. They expect middle managers to use BA to explain their performances and back up their plans for reaching goals and solving problems. When making decisions under uncertainty, managers are subject to cognitive biases that limit their ability to make high quality decisions (Bazerman & Moore, 2008). Middle managers are now using data to enlighten their own thinking, thereby lowering potential biases when making plans and decisions. However, even though performance is now visible, the findings do not support statements saying that numbers speak for themselves when enough data is present (Mayer-Schönberger & Cukier, 2013). Middle managers use their experience to think about what might be interesting to research and then collect and analyse data based on a functional question, thereby giving meaning to the data. Data alone cannot identify indicators, targets, or solutions to problems. Hence, even though BA helps in building evidence-based plans and solutions, human interpretation is still needed, so biases are not entirely eliminated. However, this more evidence- and fact-driven way of decision making and MC can provide guidance in designing effective and unbiased decision making and management processes.

26

their indicators and targets, with the aim of backing up their annual plans to higher level managers. Since BA allows data-driven decision making, higher level managers expect middle managers to take that opportunity. This means middle managers cannot hand in numbers that are too low, but in the same vein higher level managers cannot set targets that are too high either. This is interesting given the positive performance effect of correctly balancing what is desired and what is thought to be feasible in setting targets (Ferreira & Otley, 2009). On the other hand, while some authors suggest that BA can be used for re-formulating the organisation’s strategy (Mello, Leite, & Martins, 2014), this was not supported by the findings. Even though tactics are based on BA as described before, the overarching strategy and subsequent goals and high-level targets are based on external expectations and internal ambitions, rather than BA. This might however be explained by the fact that the case organisations are not yet fully data-driven and use many different systems (legacy), limiting their ability to combine and analyse data from the entire organisation.

In the process of management control, or using MCS, information is used to enable corrective actions (feedback) and to enable the generation of new ideas (feed-forward). The findings show that BA facilitates both of these uses of control. In terms of diagnostic control, the findings show that BA can be used to help measure and analyse results and take action on performance (Raffoni et al., 2018; Visani, 2017; Warren, Moffit, & Byrnes, 2015). Case organisations are comparing actual performance with planned KPI targets and BA helps by visualising performance data in dashboards (Peyrot, Childs, Doren, & Allen, 2002) and allowing easy analysis with drill-down options. However, less time is spent on monitoring actuals versus plan, because everyone is aware of their performance (Appelbaum et al., 2017; International Federation of Accountants, 2017). This is an interesting finding, since a focus on the past is a common weakness of MCS (Visani, 2017). BA allows for looking forward and solving problems by evaluating different options for remedial action (Appelbaum et al., 2017; Pauwels et al., 2009). BA thereby facilitates an interactive way of control by managers. Managers have frequent face-to-face meetings with their teams for problem solving to discuss and challenge data, which promotes the discussion of current tactics’ effectiveness (Raffoni et al., 2018). Besides, BA facilitates the detection of strategic uncertainties, for instance by enabling real-time information about the external environment of the organisation (Raffoni et al., 2018; Visani, 2017; Warren, Moffit, & Byrnes, 2015).

27

review meetings was realised in the top layers of the organisation, which according to Meekings (2005) is an effective way of doing so since the people at the top can serve as role-models.

28

6. C

The data available for analysis is exploding, while at the same time organisations face increasing competition and rising customer expectations. Technological innovations like BA can help managers by providing evidence-based insights for decision making and control due to more fine-grained and updated data (Raffoni et al., 2018). For organisations to get value from their data, MCS must be able to gather, absorb and leverage this new information (Elbashir, Collier, & Sutton, 2011). However, only two empirical studies have been conducted to explore how BA is involved in MCS. Therefore, this study attempts to contribute to research on technological innovations in MC, by exploring the BA phenomenon in the design and use of MCS. Based on a multiple case study conducted in ten organisations selected for their variation in industry and size, this study brings forth several interesting findings.

29

As in any study, this study has several limitations, which also provide opportunities for future research. First of all, the findings show that BA enables visibility of performance and can thereby potentially lower possible biases in decision making. Right now, the data is used in combination with manager’s own thoughts and experience, so biases are not eliminated. However, this is partly caused by organisations having many systems (legacy), which is a major issue for organisations in applying BA on their data. Start-up organisations were not included in the sample because their MCS are said to be limited (Sandelin, 2008; Sandino, 2007). However, because these organisations are still small and started their business knowing the advantages of creating a proper data landscape, their use of BA might prove to be more advanced. Future research could shed more light on the role of data when legacy is not an issue, which might allow the data to give meaning on its own. Therefore, future studies should explore the role BA plays in MCS of start-up organisations.

This study was cross-sectional, meaning no long-term effects of using BA for MCS were studied. To help understand these effects, a longitudinal study would be an interesting avenue for future research. Several studies have claimed that using BA increases organisational performance (Brynjolfsson, Hammerbacher, & Stevens, 2011; Chang, Kauffman, & Kwon, 2014; Warren, Moffit, & Byrnes, 2015). However, it takes two years to see improvements in efficiency and financial performance of using ERP systems (Wier, Hunton, & HassabElnaby, 2007). It would be interesting to see whether BA’s advantages lead to profitability in the long run.

Furthermore, even though this study focused on several aspects of MC, the subject of reward systems was not discussed. These are typically the outcome of performance evaluations and can be a next aspect to be considered with regards to BA (Ferreira & Otley, 2009).

Even though a maximum variation strategy was chosen to study the phenomenon in different contexts, no explicit differences were found in the ways case organisations are using BA. A strong common pattern was derived from the interviews and observations, however only a small sample was studied and the scope is thus limited. Therefore, a statistical generalisation of the findings is not possible. While this is acceptable in light of the exploratory nature of the study, future research should further investigate the use of BA within MCS in other contexts.

30

31

R

Amani, F. A., & Fadlalla, A. M. (2017). Data mining applications in accounting: a review of the literature and organizing framework. International Journal of Accounting Information Systems,

24, 32-58.

Amigoni, F. (1978). Planning management control systems. Journal of Business Finance & Accounting,

5(3), 279-291.

Appelbaum, D., Kogan, A., Vasarhelyi, M., & Yan, Z. (2017). Impact of business analytics and enterprise systems on managerial accounting. International Journal of Accounting Information

Systems, 25, 29-44.

Atlas.ti. (2018, 1 23). Product: Atlas.ti 8 for Windows. Retrieved from Atlas.com: http://atlasti.com/product/

Bazerman, M. H., & Moore, D. A. (2008). Judgment in managerial decision making. Hoboken: Wiley. Belfo, F., & Trigo, A. (2013). Accounting Information Systems: Tradition and Future Directions.

Procedia Technology, 9, 536-546.

Benoot, C., Hannes, K., & Bilsen, J. (2016). The use of purposeful sampling in a qualitative evidence synthesis: A worked example on sexual adjustment to a cancer trajectory. BMC Medical

Research Methodology, 16(21), 16-21.

Berry, A. J., Coad, A. F., Harris, E. P., Otley, D. T., & Stringer, C. (2009). Emerging themes in management control: A review of recent literature. The British Accounting Review, 41, 2-20. Bhimani, A., & Willcocks, L. (2014). Digitisation, 'Big data' and the transformation of accounting

information. Accounting and Business Research, 44(4), 469-490.

Bisbe, J., Batista-Foguet, J., & Chenhall, R. (2007). Defining management accounting constructs: a methodological note on the risks of conceptual misspecification. Accounting, Organizations and

Society, 32(7-8), 789-820.

Bititci, U., Garengo, P., Dörfler, V., & Nudurupati, S. (2012). Performance measurement: Challenges for tomorrow. International Journal of Management Reviews, 14(3), 305-327.

Bloom, N., Garicano, L., Sadun, R., & Van Reenen, J. (2014). The distinct effects of information technology and communication technology on firm organization. Behavioral Economics,

60(12), 2859-2885.

Blumberg, B. F., Cooper, D. R., & Schindler, P. S. (2014). Business Research Methods (4 ed.). New York: McGraw-Hill Education.

Brands, K., & Holtzblatt, M. (2015). Business Analytics: Transforming the Role of Management Accountants. Management Accounting Quarterly, 16(3), 1-12.

Brignall, S. (2002). The Unbalanced Scorecard: A Social and Environmental Critique. Performance

32

Brynjolfsson, E., Hammerbacher, J., & Stevens, B. (2011). Competing through data: three experts offer their game plans. McKinsey Quarterly, 36-47.

Bughin, J., Livingston, J., & Marwaha, S. (2011). Seizing the potential of "Big Data". McKinsey

Quarterly, 103-109.

Buonanno, G., Faverio, P., Pigni, F., Ravarini, A., Sciuto, D., & Tagliavini, M. (2005). Factors affecting ERP system adoption: a comparative analysis between SMEs and large companies. Journal of

Enterprise Information Management, 18(4), 384-426.

Carlsen, B., & Glenton, C. (2011). What about N? A methodological study of sample-size reporting in focus group studies. BMC Medical Research Methodology, 11, 26.

Chae, H., Koh, C. E., & Park, K. O. (2018). Information technology capability and firm performance: Role of industry. Information & Management, 55(5), 525-546.

Chang, R. M., Kauffman, R. J., & Kwon, Y. (2014). Understanding the paradigm shift to computational social science in the presence of big data. Decision Support Systems, 63, 67-80.

Chapman, C. S., & Kihn, L. A. (2009). Information System Integration, Enabling Control and Performance. Accounting, Organizations and Society, 34(2), 151-169.

Chen, H., Chiang, R., & Storey, V. (2012). Business Intelligence and Analytics: From Big Data to Big Impact. MIS Quarterly, 36(4), 1165-1188.

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

Society, 28, 127-168.

Chenhall, R. H., & Moers, F. (2015). The role of innovation in the evolution of management accounting and its integration into management control. Accounting, Organizations and Society, 47, 1-13. Chenhall, R. H., & Morris, D. (1986). The impact of structure, environment, and interdependence on

the perceived usefulness of management accounting systems. The Accounting Review, 16-35. Chiasson, M. W., & Davidson, E. (2005). Taking industry seriously in information systems research.

MIS Q, 29(4), 591-605.

Chow-White, P. A., & Green, S. E. (2013). Data Mining Difference in the Age of Big Data: Communication and the social shaping of genome technologies from 1998 to 2007.

International Journal of Communication, 7, 556-583.

Côrte-real, N., Ruivo, P., & Oliveira, T. (2014). The diffusion stages of business intelligence & analytics (BI&A): A systematic mapping study. Procedia Technology, 16, 172-179.

Davenport, T. H., & Harris, J. G. (2007). Competing on Analytics. The new science of winning. Boston: Harvard Business Press.

Davenport, T. H., Harris, J. G., & Morison, R. (2010). Analytics at work: smarter decisions. Better

results. Boston: Harvard Business Press.