Insights about Business Intelligence and Decision-Making

(1)

Insights about Business Intelligence and Decision-Making

A case study at Systembolaget

Master’s Thesis 15 credits

Department of Business Studies Uppsala University

Spring Semester of 2020

Date of Submission: 2020-06-03

Elisabeth Hugner Viktor Sjöberg

Supervisor: Shruti Kashyap

(2)

Abstract

In today’s constantly evolving technological environment, businesses have more tools to support decision-making and these can be categorized as Decision Support Systems (DSS).

One of the tools is Business Intelligence (BI), which is regarded as a high-priority investment in organizations nowadays. Even though there exists a vast amount of research in the DSS area, most of the influential work is conducted in time incomparable to today’s technological environment. In addition, most of the research focuses on profit-seeking organizations, as BI has been regarded as a tool to increase profits. However, non-profit organizations also use BI, but are not portrayed in the BI research area. The aim with this study is to explore how BI is used in relation to decision-making in a non-profit organization and to investigate the crucial factors in the usage of BI in relation to decision-making.

A qualitative case study approach is applied where the Swedish non-profit organization Systembolaget AB is the case company. The main findings indicate that interaction between the two decision-making types is needed when using BI in a non-profit context. Moreover, having data literacy, data reliability, and data accessibility is found crucial in order to achieve BI success in relation to decision-making, especially when more and more decisions are made at the operational level. Finally, the results of this study amplify the need for an update in the DSS framework.

Keywords: Business Intelligence, Decision Support System, Decision-making levels, System 1 and System 2 decisions, Non-profit organization

(3)

Acknowledgments

We wish to express our deepest gratitude to our supervisor Shruti Kashyap for the support throughout the process. Shruti showed great engagement and challenged us constantly, which has been invaluable and has helped us to carry through this thesis. Also, we would like to direct a special appreciation to our case company Systembolaget and the respondents who dedicated their time to be interviewed, which enabled us to conduct this study. Finally, we wish to show our gratefulness to the opponents for providing us with constructive feedback during the seminars.

(4)

1. Introduction

The chapter begins by introducing decision-making and Decision Support Systems. The phenomena of Business Intelligence in relation to decision-making is problematized and the research questions are thereafter presented. Followed by a brief presentation of the methodology and the main findings of the study. Finally, the remaining part of the thesis is outlined.

Decision-making, whether complex or simple, is part of our everyday life. For both individuals and organizations, decision-making is important in order to ensure survival and growth.

Historically, research on decision-making has focused on the rational behavior of actors as founded on neoclassical economic assumptions of full rationality (Ackert & Deaves, 2010;

Bazerman & Moore, 2015). In recent years, a growing awareness of the bounded rationality of actors has grown and the basis of neoclassical economic approaches to decision-making have come into question (Kahneman, 2011). The role of biases and heuristics are increasingly recognized in decision-making research as well as in organizational approaches towards decision-making (Bazerman & Moore, 2015; Ackert & Deaves, 2010). In today’s business environment, technological innovations are recurrent, and businesses therefore have more tools to reduce the cognitive mistakes in the decision-making, in order to increase rationality (Brynjolfsson & McAfee, 2014). One increasingly acknowledged tool is Business Intelligence (BI), that falls in the category of Decision Support System (DSS) (Arnott, Lizama & Song, 2017). In this study, BI is defined as a tool used to support organizations in making more efficient decisions by providing business information, which is closely related to the definition made by Lönnqvist and Pirttimäki (2006).

Research in the BI area has so far focused on how to achieve BI success by identifying the critical success factors, mainly in the implementation phase of a BI system (IşıK, Jones, &

Sidorova, 2013; Yeoh & Koronios, 2010; Popovič et al., 2012). For instance, IşıK, Jones and Sidorova (2013) identifies user access, system integration and data quality as critical success factors, whilst Popovič et al. (2012) identifies the analytical decision-making culture as a critical success factor. Prior research proposes that the use of BI has a positive impact on decision-making (Ranjal, 2009; Watson & Wixom, 2007), however, several studies agree that certain prerequisites are required (Ackert & Deaves, 2010; IşıK, Jones & Sidorova, 2013; Shah,

(6)

Horne, & Capellá, 2012; Provost & Fawcett, 2013; Wieder & Ossimitz, 2015). Nevertheless, there are more to discern regarding factors affecting decision-making in relation to BI.

Acting as a decision-making lens, Gorry and Scott Morton (1989) have created an influential framework of DSS, which divides the decisions into different decision types and categorizes based on which organizational level the decision is made on. The organizational levels build on the taxonomy presented by Anthony (1965) and are strategic planning, management control and operational control. This was later updated by Anthony et al. (2014) who outlined the concept of strategic, tactical, and operational decision levels. For instance, strategic decisions can refer to an organization's strategic objectives, tactical decisions concern how resources are obtained and handled to meet the objectives, and operational decisions concern how to carry through specific tasks effectively (Anthony et al., 2014). The different decision types were initially referred to as structured, semi-structured, and unstructured decisions (Gorry & Scott Morton, 1989), but in a more recent application of the DSS framework, it is referred to as System 1 and System 2 decisions, and the interaction between them (Arnott, Lizama & Song 2017). System 1 decisions are characterized to be based on intuition and gut feeling, whereas System 2 decisions are more analytical and logical (Bazerman & Moore, 2015). For this study, the updated DSS framework by Arnott, Lizama and Song (2017) has been used, however the old definitions by Anthony (1965) used in their framework have been replaced with the updated definitions by Anthony et al. (2014). Decision-making is regarded as a key factor to company success (Bazerman & Moore, 2015), and there exists a vast body of research that addresses DSS (Gorry & Scott Morton, 1989; Gray, 1987; Sprague, 1980). Most of the extant research on DSS has been conducted in times where technological capabilities have been severely limited in comparison to the technological progress of the current times (Power, Heavin &

Keenan, 2019).

One of the main incentives to use BI in an organizational context is to increase profits (Davenport, 2006), by for example identifying profitable products, customer segments and finding ways to increase sales and reduce costs (Ranjan, 2009). Accordingly, most of the research about BI is centered around profit-seeking companies, and the absence of studies of BI in a non-profit organizational context is apparent. However, LeRoux and Wright (2010) have found that reliance on performance information, which is under the scope of BI (Lönnqvist & Pirttimäki, 2006), is positively related to effective decision-making in non-profit organizations. Although, LeRoux and Wright (2010) stresses the need for more qualitative

(7)

research regarding decision-making in non-profit organizations to fully discern how it relates to performance information. Furthermore, decision-making in non-profit organizations is interesting to study because it can be distinguished from decision-making in profit-seeking organizations. For instance, Khan and Khandaker (2016) approximate that decision makers in non-profit organizations or public organizations typically make decisions that affect a whole population, whilst decision makers in profit-seeking or private organizations typically affect their specific customers. In addition, non-profit organizations normally operate with a social or political purpose, whilst profit-seeking organizations normally operate to create excess profit (Khan & Khandaker, 2016). Hence, there is a need to investigate decision-making in non-profit organizations and how BI is used in relation to it.

In addition to the above, it is of relevance to note that the DSS framework has previously been applied in a BI context by Arnott, Lizama and Song (2017), who contextualized different use patterns. However, an existing gap that remains is the question of how BI relates to the different types of decisions and how it is used depending on organizational levels. A motive to address why organizational levels are interesting for research inquiry is that few studies have included how BI is used in relation to operational decision-making. Although in practice, an increasing number of decentralized companies has led to more and more decisions being taken at the operational level, in the core of the business (Nilsson & Rapp, 2005: Petersen, Plenborg &

Kinserdal, 2017). Hence, there is a need to contextualize BI in relation to decision-making by using the DSS framework as a lens.

Building on the above, there is an identified need to further discern how BI is used in relation to decision-making, and also to investigate how non-profit organizations make decisions.

Consequently, the aim with this study is to explore how BI is used in relation to decision- making in a non-profit organization and to investigate the crucial factors in the usage of BI in relation to decision-making. Accordingly, the research questions are:

1. How is Business Intelligence used in relation to decision-making in a non-profit organization?

2. What are the crucial factors in the usage of Business Intelligence in relation to decision- making?

Additionally, in order to understand how BI is used in relation to decision-making in a non- profit organization, a prerequisite is to investigate the purpose behind the usage of BI.

(8)

Furthermore, the DSS framework presented by Gorry and Scott Morton (1989) and later updated by Arnott, Lizama and Song (2017) will be used as a lens to investigate decision- making, in order to incorporate different decision types and different organizational levels into the research inquiry.

This paper extends current knowledge about how BI is used in relation to the decision-making, by building on the research made by Arnott, Lizama and Song (2017) and contextualizing BI through the lens of the DSS framework by Gorry and Scott Morton (1989). Additionally, this paper contributes to previous research by including the perspective of a non-profit organizational decision-making. This research will hence add to both decision support theory and organizational theory.

To enable depth in the research inquiry, a qualitative case study approach has been applied (Yin, 2011), where the non-profit context is the case company Systembolaget. Systembolaget is owned by the Swedish state and has a monopoly of the distribution and retail of alcoholic beverages in Sweden (Systembolaget, 2020). In order to answer the research questions, semi- structured interviews were chosen as primary data collection method, due to the level of details needed (Bryman & Bell, 2011). A total of eight interviews were conducted with employees at Systembolaget, most of them in managerial positions. In addition, a data triangulation attempt was made in order to reach data saturation (Fusch & Ness, 2015). However, due to the lack of other data sources, interviews were concluded to be the superior data collection method for this type of research inquiry.

The main findings in this paper indicate that several benefits related to BI and decision-making have been identified, in addition to motives why a non-profit organization uses BI. Moreover, the findings indicate that more decisions are made with interaction between the different decision types, therefore it is not optimal to become completely data-driven, especially in non- profit organizations with a specific purpose. Furthermore, the findings evidence that the shift towards becoming more decentralized requires organization-wide data literacy, since it leads to BI being used to a higher extent at the operative level. Besides data literacy, data reliability and information accessibility are also identified as crucial factors in the usage of BI in relation to decision-making. Finally, a need to update the DSS framework to be more suitable to today’s technological decision environment is found, where both parameters regarding decision types and organizational levels demand adjustments.

(9)

The outline of the remainder of this thesis is as follows; first, the theoretical framework is presented and an illustrative model of the linkage between the theories is presented. Thereafter, the case company is introduced, followed by the methodology section. Then, the empirical findings are presented followed by an analysis where the theoretical framework is applied to the findings. Finally, the conclusions of this paper are outlined, limitations are addressed, and possible avenues for future research inquiries are presented.

(10)

2. Theoretical Framework

The following chapter presents the theoretical framework used in this paper. To begin with, the Decision Support System framework is presented, followed by decision-making levels and two ways of decision- making. Thereafter, BI and decision-making levels are presented followed by data-driven decision- making and Business Intelligence. The Business Intelligence section ends with theory about achieving BI success. The theory section ends with a figure showing the linkage between the different theories.

2.1 The Decision Support System framework

The Gorry and Scott Morton (1989) DSS framework was produced in order to support managerial activities and can be used as a lens to underpin decision-making. It is emphasized that information systems should only exist with the purpose to support decision-making, not to overtake the decision-making process. First, the framework is built on what kind of organizational level the decision is made on. Gorry and Scott Morton (1989) use the taxonomy by Anthony (1965) to describe the different levels of managerial decision-making: strategic planning, management control and operational control. These levels will be elaborated in section 2.1.1.

Second, the framework is built on to what extent a certain decision is handled with a lot of human interaction, or if it can be replaced by an information system. Gorry and Scott Morton (1989) use the terms structured, semi-structured, and unstructured decisions to explain these different states. Structured decisions refer to processes that can be fully or largely automated or replaced by an information system, whilst unstructured decisions refer to processes where the human judgement stands for the decision basis. Semi-structured decisions can be placed in between the two, because both information systems and human judgment are used in the decision-making process.

The DSS framework by Gorry and Scott (1989) is extremely well-established and highly cited in existing literature within the field (Abbasi, Sarker & Chiang, 2016). Arnott, Lizama and Song (2017) have updated the framework in order to better align it with more current business environments. More specifically, they have acknowledged the type of discussion held by Ackert and Deaves (2010) and Bazerman and Moore (2015) about bounded rationality. Thus, they have replaced structured, semi-structured, and unstructured decisions with the type of

(11)

decisions Stanovich and West (2000) describe as System 1 and System 2, and decisions that have interaction between the two. These systems will be elaborated in section 2.1.2.

2.1.1 Decision-making levels

As mentioned above, Gorry and Scott Morton (1989) base their framework on the organizational levels by Anthony (1965). Anthony et al. (2014) present an updated version of the different levels, which can be named strategic decisions, tactical decisions, and operational decisions. These levels are most commonly presented like a pyramid with the strategic level at the top and the operational level at the bottom. Anthony et al. (2014) and Nilsson and Rapp (2005) exemplifies the levels by dividing them into different time-perspectives: the strategic level concern long term actions and year-to-year decisions such as entering a new market, the tactical level concern month-to-month decisions such as handling volatile foreign exchanges, and the operational level concern day-to-day decisions such as promotion of products in a retail store. If the strategic decision-making is about setting the long-term strategy and objectives for a company, the tactical level concerns that resources are obtained and used effectively, in order to meet the objectives. Lastly, the operational decision-making is about how to carry through the strategic goals, by carrying through specific tasks effectively and efficiently (Anthony et al., 2014; Nilsson & Rapp, 2005). Following model has been created with inspiration from Anthony et al. (2014), to illuminate the different decision-making levels:

Figure 1. Decision-making levels (Anthony et al., 2014)

Even if the distinction between the different levels seems apparent, there is an ongoing discussion about the actual differences between them. Nilsson and Rapp (2005) elucidate that

(12)

the biggest reason behind the diminishing of the different levels is the vast shift from centralized to decentralized companies. When firms remove its hierarchical structures and move decisions further down in the organization, the different levels lose its relevance.

Langfield and Smith (1997) supports this and stresses that the boundaries between strategic, tactical and operational decision-making may no longer hold. However, it is emphasized that the different levels in itself are not disappearing, it is just the boundaries between them.

Furthermore, Nilsson and Rapp (2005) and Petersen, Plenborg and Kinserdal (2017) addresses that more and more organizations are becoming decentralized, which has led to more decisions being taken at the operational level. Usually, the operational level constitutes the core of a business, because it is the place where the real business appears, which is the main reason for the growing amount of decisions taken at the lower levels in organizations (Nilsson & Rapp, 2005; Petersen, Plenborg & Kinserdal, 2017).

2.1.2 Two ways of decision-making

As previously presented, the study conducted by Arnott, Lizama and Song (2017) has given a more contemporary version of the framework by Gorry and Scott Morton (1989), by using the different kinds of decision systems explained by Stanovich and West (2000). With the rise of behavioral economic theories, the decision-making theories have broken new ground. In general, traditional economic theories based on rationality have been into question lately because the human behavior simply is not rational (Ackert & Deaves, 2010). With the knowledge that people are not rational in their decision-making, two kinds of decisions have emerged: System 1 and System 2. Bazerman and Moore (2015), Arnott, Lizama and Song (2017) and Stanovich and West (2000) elaborate this and characterize System 1 as decisions that are fast, automatic, effortless, implicit, emotional and based on intuition. System 2 on the other hand, refers to decisions that are slower, conscious, effortful, explicit and logical (Bazerman & Moore, 2015; Arnott, Lizama & Song, 2017; Stanovich & West, 2000). Because these decisions are thoroughly grounded and based on logical reasoning, Bazerman and Moore (2015) argue that the more important decisions should be made by System 2. Furthermore, Bazerman and Moore (2015) illuminate that people should try to move away from System 1 decisions based on gut feelings and intuition into more System 2 decisions, based on logic and thorough analysis. However, System 1 should not be removed completely, rather it should be

(13)

replaced in those situations where intuition and expertise is not enough to foster decision- making (Bazerman & Moore, 2015).

With the basis of what Bazerman and Moore (2015), Stanovich and West (2000) and Arnott, Lizama and Song (2017) define as System 1 and System 2, following table has been produced.

Table 1. Two different types of decisions

System 1 System 2

Fast Slow

High capacity Low capacity

Automatic Conscious

Skilled Rule following

Effortless Effortful

Implicit Explicit

Emotional Logical

Highly contextualized Decontextualized

Personalized Personalized

Intuition and gut feeling Analytical

Despite being described as two separate systems, they often operate with interaction. Arnott, Lizama and Song (2017) describe the two systems as two minds in the same body. System 1 can be intuitive answers to sudden judgment problems, whereas System 2 verifies the quality of the intuitions. In other words, the System 2 works as a justification for the System 1, if the intuition should be endorsed or overridden (Arnott, Lizama & Song, 2017). Also, Bazerman and Moore (2015) stress the importance of the timing of the two ways of making decisions. For instance, they emphasize that it is not required to fully rely on System 2 in every decision- making. Instead, a good decision maker should know when to move from the intuition and expertise-based System 1 to the more analytical System 2. Also, it can be beneficial to move away from System 2 in order to be more contextualized (Arnott, Lizama & Song, 2017).

The decision-making lens in this paper is based on the DSS framework presented by Gorry and Scott Morton (1989), however, the decision types are referred to as System 1, System 2, and

(14)

the interaction between them, in line with Arnott, Lizama and Song (2017). Moreover, the decision levels in this paper are referred to as strategic, tactical, and operational levels (Anthony et al., 2014; Nilsson & Rapp, 2005). Hence, an updated version of the DSS framework originally presented by Gorry and Scott Morton (1989) is applied in this paper. The framework is illustrated in figure 2.

Figure 2. Decision Support System Framework

2.2 BI and decision-making levels

Golfarelli, Rizzi and Cella (2004) discuss a second era of BI and how a new approach of BI is adjusted in order to better fit into the different decision levels. The new approach of BI is called Business Performance Management (BPM) (Golfarelli, Rizzi & Cella, 2004). Instead of mainly focusing on quantifying business information and to make it available throughout an organization, this BI approach seeks to quantify the business strategy and targets, with the purpose to decentralize decision-making (Golfarelli, Rizzi & Cella, 2004).

In order to facilitate the absorbance of the strategy on the tactical and operational level, the main users of this BI approach are at the tactical and operational levels in the organizations.

The BI approach refers to information systems that foster the users at tactical and operational levels to absorb the strategy in their day-to-day work. Instead of working towards the overall

(15)

strategy, workers at the tactical and operational levels should break down the strategy into multiple goals. The information system captures these goals and supports the decision maker to work towards the strategy (Golfarelli, Rizzi & Cella, 2004).

Decisions at the tactical and operational levels have to be faster than strategic decisions, because these decision makers mainly deal with a limited number of tasks (Golfarelli, Rizzi &

Cella, 2004). Furthermore, Golfarelli, Rizzi and Cella (2004) stress that the decision makers at the tactical and operational levels do not have time or skills to dig into deep analytical issues, thus the information will mainly be translated into reports and interactive dashboards, that is used to facilitate the decision-making (Golfarelli, Rizzi & Cella, 2004).

2.3 Data-driven decision-making

The structured decisions by Gorry and Scott Morton (1989) and System 2 decisions by Stanovich and West (2000) can be related to what Anderson (2015) refers to as data-driven decision-making. Data-driven decision-making is also connected to the relation of BI and decision-making (Anderson, 2015). Today, being data-driven is something many organizations strive to be. There are a number of advantages with becoming data-driven, however the overall idea is that making decisions based on facts and evidence are considered more reliable than decisions based on intuition and gut feeling (McAfee & Brynjolfsson, 2014). For instance, being data-driven enables a better understanding in cause of problems, it enables more informed decisions, it can create new business opportunities, and it also saves costs and time (Anderson, 2015; Barton & Court, 2012; Bazerman & Moore, 2015; Watson & Wixom, 2007).

However, Provost and Fawcett (2013) are also discussing the interaction between data-driven decision-making and intuition. They highlight that data-driven decisions should not be seen as an all-or-nothing practice, because the intuitive judgements play an important role in the decision-making.

In order to become data-driven in the decision-making processes, Anderson (2015) brings up two prerequisites. First, the organization has to collect the right data. The data has to be clean and unbiased, but most importantly trustworthy. Second, the data has to be accessible and queryable. Anderson (2015) stress that the data has to be accessible throughout an organization,

(16)

otherwise the risk of relying on few experts increases. Barton and Court (2012) illuminates a similar theme and stress that everyone in an organization should have access to the data, and that everyone should be able to make decisions based on the data. Furthermore, Anderson (2015) produced a model called Analytics Value Chain, that describes the process of how data ends up being value generating for an organization. This model is presented in figure 3 below.

Figure 3. Analytics Value Chain (Anderson, 2015)

If the prerequisites are fulfilled and the steps in the model are followed, an organization can consider them self as being data-driven (Anderson, 2015). The data has to be of the right type and trustworthy, which in turn feed the reports. The reports lay the ground for deeper analysis in which the decision maker can take direct actions on. Finally, these actions serve as the value- creation for the organization (Anderson, 2015). Anderson (2015) further emphasizes that in order to be a data-driven decision maker the analysis has to be deeper and be able to answer questions such as why, when, where and who, instead of only what.

2.4 Business Intelligence

As previously mentioned, the DSS described by Gorry and Scott Morton (1989) is a system with the purpose of supporting decision-making. A system that can be placed in that category is BI, which is an umbrella term for systems related to decision support (Arnott, Lizama &

Song 2017). Shortly explained, BI is a tool to support decision-making by collecting, integrating, analyzing and presenting business information (IBM, 2020). Lönnqvist and Pirttimäki (2006) define BI as a “managerial philosophy and a tool used to help organizations manage and refine business information with the objective of making more effective business decisions” (p.32). However, there are several other definitions of BI (Davenport, 2006; IşıK, Jones & Sidorova, 2013; Watson & Wixom, 2007). In this paper, BI is defined as a tool used

(17)

to support organizations in making more efficient decisions by providing business information, which is closely related to the definition made by Lönnqvist and Pirttimäki (2006).

According to previous research, BI enables improved decision-making (Ranjan, 2009; Watson

& Wixom, 2007). For instance, Ranjan (2009) argues that BI enhances the overall performance of the organization using it and presents several related benefits. First, since information is valued to be the second most important asset in an organization, having accurate and timely information in the decision-making enables the organization to perform better. Hence, by using BI and gathering trends in the market, the organization can respond quicker to changes than competitors and thereby gain a competitive advantage, and the organization can also improve customer satisfaction by constantly being up to date about customer’s changing demands regarding new innovations and services. Second, using BI can erase guesswork by providing facts, and it can also improve communication between departments. In summary, BI aids the decision maker to make better decisions by providing timely, accurate, relevant, and accessible information (Ranjal, 2009).

Moreover, a motive to work with BI can be to improve the competitive advantage. Ranjal (2009) stresses the importance of staying competitive in a world that is rapidly changing. In order to meet customer’s changing demands and expectations, the organization must be adaptable to changes and also be able to respond quickly. BI facilitates this kind of information.

Ranjal (2009) further argues that it is vital for organizations to rely more and more on the information from the BI systems in order to increase competitiveness by predicting future events and identifying trends. To conclude, BI facilitates well-informed decisions and can thereby contribute to increased competitiveness.

2.4.1 Achieving BI success

As previously mentioned, research within the BI area has focused on BI success and what factors influencing it. For instance, IşıK, Jones and Sidorova (2013) investigated in their study the relationship between five BI capabilities, and the influence of the decision environment on the relationship between BI capabilities and BI success. The results showed that the user access, the integration of other systems in BI, and the data quality are vital in order for BI to be

(18)

successful, despite the decision environment. Furthermore, Popovič et al. (2012) examined the relationship between information quality, maturity, analytical decision-making culture, and the use of information for decision-making as significant elements of the success of BI systems.

The two information quality dimensions, information content quality and information access quality, illustrated in the study how the focus in implementing BI systems is damaging the success of the system. This is because more focus is on the information access quality in the implementation, meanwhile the users of the system need information content quality to increase the use of the information in business processes (Popovič et al., 2012).

Popovič et al. (2012) further concluded that business organizations with a high analytical decision-making culture use information available to more extent than those with less analytical decision-making culture, regardless of the content quality. This is in line with IşıK, Jones and Sidorova (2013) argument that in order to achieve maximum benefits from the BI investment, an organization must apply the right BI capabilities to the right decision environment. Even though organizations with a high analytical decision-making culture are using BI to the largest extent, the BI systems have to be constructed in a way that the general populace are able to use it. Barton and Court (2012) addresses that analytical models should balance between complexity and ease of use. The organizations should ask them self “what is the least complex model that would improve business performance” (Barton & Court, 2012, p.

82). The models need to be able to lay the ground for deep analysis, but the user friendliness cannot be forgotten (Barton & Court, 2012). In accordance, Barton and Court (2012) and Shah, Horne and Capellá (2012) stress that analytical models often end up being designed for experts, and not for the employees at the front lines. Thus, a BI system should be simple and understandable also for decision-making at the operational levels (Barton & Court, 2012; Shah, Horne & Capellá, 2012).

2.4.1.1 Achieving BI success in relation to decision-making

There are several factors affecting the successfulness of data-driven decision-making, which can be grouped in the following six ways. To begin with, Shah, Horne and Capellá (2012) explain that in many of the leading data-driven organizations, it is usual to have predefined decision-making processes and standardized methods, in order to enable employees to use the most suitable data. Provost and Fawcett (2013) also argues that it is vital to have frameworks

(19)

to structure the analytical thinking within an organization in order to improve data-driven decision-making. Another important aspect in improving data-driven decision-making is according to Provost and Fawcett (2013) to reach a general knowledge about the fundamental concepts among the employees.

Additionally, in a world where data and technology are becoming more and more important (McAfee & Brynjolfsson, 2014), having data-analytical skills within an organization is a crucial prerequisite (Provost & Fawcett, 2013). Provost and Fawcett (2013) stresses that understanding the essential principles of the organization’s data science enables employees to get the best out of the systems. Shah, Horne and Capellá (2012) even argues that “investments in analytics can be useless, even harmful, unless employees can incorporate that data into complex decision-making” (p. 23). According to Shah, Horne and Capellá (2012), it is common that organizations rely on a few highly analytical skilled individuals, and thus, fail to spread the competence throughout the organization.

Furthermore, Shah, Horne and Capellá (2012) argues that many executives are not engaged in the data, often with the motives that they do not possess enough knowledge about it, or totally rely on the IT department to handle it, and thereby risk not having insights in how the data is shared throughout the organization. The organization will hence only achieve limited benefits with the investment in data, due to the underinvestment made by the executive in information knowledge.

As a further point of consideration, information overload can affect data-driven decision- making. With the increased awareness of data-driven decisions and System 2 thinking, decision makers strive to collect as much data and information as possible to support the decision- making (Bazerman & Moore, 2013). However, with the vast amount of data and information collected, a state of information overload can arise. Ackert and Deaves (2010) and Bazerman and Moore (2013) describes information overload as a situation when the decision maker has too much information on the table, which leads to confusion. The large amount of information can backfire and lead to a sense of disorientation, which has a direct negative effect on the decision-making. If the user is not capable of processing the information, a data-driven or

System 2 decision cannot be made (Ackert & Deaves, 2010).

(20)

In addition to the above, IşıK, Jones and Sidorova (2013) found user access quality to be crucial to BI success. The access to the BI systems can be limited or it can be unlimited for all users in the organization. When the access to the BI systems are limited, the organization applies some kind of access control or authorization process, and when the access is unlimited, the organization usually applies a web-centric approach. However, IşıK, Jones and Sidorova (2013) stresses that it is vital for the BI user to have access to the information related to the decisions they make, and Wieder and Ossimitz (2015) argues that “proper tools are required to easily access only relevant and current information” (p. 1167). Furthermore, having the appropriate access can be related to the different organizational levels, since the needs vary depending on which organizational level you operate on (Eckerson, 2010). Eckerson (2010) explains that on the operational level, access to real-time data and information about the core processes are needed, whilst on the strategic level, the managers will have to have access that enables them to monitor the strategic targets.

Finally, the reliability of the information affects data-driven decision-making. IşıK, Jones and Sidorova (2013) argues that having reliable information is an important factor in order for BI to support decision-making successfully. However, Shah, Horne and Capellá (2012) states that many organizations have a data structure that can be compared with “libraries with no card catalog and no covers on their books” (p. 24). Hence, the reliable information is hard to find.

In order for the decision maker to rely on the data, it should be of high quality, be accessible, and interact with related systems seamlessly (IşıK, Jones & Sidorova, 2013). The data also have to be transparent, consistent, and complete, in order to be trustworthy (Wieder & Ossimitz, 2015).

2.5 Link between theories

The theories applied in this paper encompass DSS, decision-making levels, BI and decision- making levels, two ways of decision-making, data-driven decision-making, Business Intelligence, and achieving BI success. To provide an overview of the theoretical framework and to illustrate how the different theories are connected, Figure 4 has been produced. Recall that the DSS framework presented by Gorry and Scott Morton (1989) and the updated model by Arnott, Lizama and Song (2017) is used as a decision-making lens (Figure 2). The framework partly consists of the different decision-making levels, which is divided into

(21)

strategic, tactical, and operational levels by Anthony et al. (2014), and partly of two ways of decision-making, which in a modernized version provided by Arnott, Lizama and Song (2017) is referred to as System 1 and System 2 decisions.

Decision-making levels can be linked to the usage of BI in different decision-making levels.

This is presented in section 2.2 of BI and decision-making levels. Moreover, in the two ways of decision-making, System 2 decisions are closely related to data-driven decision-making which is presented in section 2.3. In turn, BI and decision-making levels and data-driven decision-making are related to BI, which is further discussed in section 2.4. Derived from BI is achieving BI success and its relation to decision-making, which is presented in section 2.4.1 and 2.4.1.1.

Figure 4. Link between theories

(22)

3. The Case Company: Systembolaget

In the next chapter, the reasons for the choice of case company are explained, and some overall information about the organization is outlined. Sections 3.1.1 to 3.1.4 can be seen as an introduction to chapter 5, because it provides the reader with useful background information before the presentation of the empirical findings.

3.1 Selection of case company

Systembolaget was selected as the case company because the authors knew from contacts that they worked actively with BI. In addition, the authors strived to fill the research gap by investigating how BI is used in relation to decision-making in a non-profit organization, because most of the previous research in the field concerns profit-seeking organizations.

Systembolaget is owned by the Swedish state and is not profit-seeking. Rather, they exist with the purpose to foster public health, by limiting the alcohol consumption in Sweden (Systembolaget, 2020). According to Cambridge Dictionary (2020), a non-profit organization is “not intended to make a profit, but to make money for a social or political purpose or to provide a service that people need”.

The main purpose of Systembolaget is to make a positive change for Sweden’s public health.

To limit the availability of alcohol and to sell alcohol without maximizing profit in order to reduce the harms of alcohol is central. What makes Systembolaget unique is that their customers are the whole population of Sweden, regardless if they can or are allowed to shop at Systembolaget. (Systembolaget Responsibility Report, 2019)

Like every other company, Systembolaget is in the middle of a digital development. The customers’ expectations and purchasing behavior has changed due to new technological possibilities (Systembolaget Responsibility Report, 2019). Additionally, in order to be resource efficient, Systembolaget needs to be adapt to the digitalization. The importance of being resource efficient is highlighted below:

“Systembolaget is not a profit maximizing company. To run Systembolaget resource efficiently, namely both cost efficient and sustainable, is although an important part of our work. It gives us bigger margin to create the best prerequisites” (Responsibility Report, 2019, p. 58, translated from Swedish)

(23)

3.1.1 Regulations and Laws

The Swedish government states that public owned organizations operate within the same laws as private owned organizations (Regeringen, 2020). However, Systembolaget must adjust to the alcohol law. The alcohol law says that a 20-year-old age limit applies at Systembolaget. As a precaution to not break the law, Systembolaget demands that everyone that looks to be under 25 years old shall be age-checked with valid ID (Systembolaget Responsibility Report, 2019).

3.1.2 Strategic Plan and Ratios

There are three strategic movements in the strategic plan of Systembolaget up to year 2020; to keep impressing the customers, to increase the knowledge about the risks with alcohol and the purpose of the company, and to have the best prerequisites. Systembolaget’s targets are both of financial and non-financial character. The goals related to the purpose of the company are primary, but the secondary economic targets are also regarded as important. Strategic ratios are, among other things, satisfied customer index, opinion index, age check rate, and rate of absence in the personnel force due to sickness. (Systembolaget Responsibility Report, 2019)

3.1.3 Organizational Structure

According to the Swedish government, public owned organizations in Sweden are run as private owned organizations (Regeringen, 2020). Furthermore, Systembolaget consists of 446 stores and around 490 delegates across Sweden. The headquarter consists of the Human Resources department, Information Technology department, Strategy and Offerings department, Assortment, Purchasing and Supply department, Sales department, Economy and Administration department, Company and Society department, Communication department, and the E-commerce department (Systembolaget Responsibility Report, 2019).

(24)

3.1.4 BI at Systembolaget

The purpose of implementing BI at Systembolaget was mainly to create a better decision support system. They replaced old tools, applications and reports systems, and implemented a system that enabled visualization and tools to measure Key Performance Indicators (KPIs).

Before Systembolaget implemented BI, a lot of work was done in Excel. Now, working with BI and having data-driven decision-making clearly formulated in the strategy, the use of BI has increased vastly in Systembolaget during the last year. It is estimated that the active users who use BI in their decision-making has increased from 150 to 300 during the last year.

The usage of BI differs across the organization, both the extent and purpose. For instance, the team in the Analytical department mainly works with BI to visualize and make the data available for others in the organization. Moreover, different kinds of BI systems exist at Systembolaget, such as PowerBI, Mercur and an enterprise cube by Microsoft. However, because the research question in this study concerns investigating how BI is used in decision- making, and what the crucial factors in the usage of BI in relation to decision-making are, the authors do not aim to explore how different kinds of BI systems are used in relation to decision- making. Accordingly, the overall designation “BI” will be used henceforward.

Many types of different data are collected into the BI systems and a number of reports can be produced. The data collected to the BI system often end up in some kind of measure or KPI.

The BI tool is thus powerful because it can be applicable to a lot of different areas. For instance, you can analyze and measure how many have quitted, how many youths have decided to quit, how many elders are working and so on. Thus, figure 5 has been produced to explain different types of data used in the BI system:

Figure 5. Examples of data types at Systembolaget

(25)

4. Methodology

The following chapter will address the methodological choices made in the study, how data was collected and the selection of respondents. Furthermore, this chapter will elaborate how the interviews have been conducted, and how a data triangulation attempt was made. Finally, the authors explain how the analysis of the empirical material was performed.

4.1 Methodological choices

The research questions in this study are as follows:

1. How is Business Intelligence used in relation to decision-making in a non-profit organization?

2. What are the crucial factors in the usage of Business Intelligence in relation to decision- making?

In order to answer the aforementioned questions, a qualitative approach has been selected.

Qualitative studies aim at describing, exploring and understanding a certain phenomenon (Bryman & Bell, 2011), and because the nature of the research question in this study is centered around investigating how a phenomenon works, a qualitative approach is thereby suitable. In addition, qualitative studies can be characterized by having a micro perspective with rich and ample data, which is a prerequisite to answer this paper's research questions. The nature and purpose of the study is exploratory, because the aim is to explain and understand a relatively unknown area (Bryman & Bell, 2011), and it is thereby necessary to work in parallel with theoretical approaches and the empirical findings during the inquiry process (Van de Ven, 2007). Hence, an abductive research approach is applied. By setting an initial theoretical framework and adjusting it and adding applicable theories during the empirical collection, the knowledge about the research area will successively grow (Van de Ven, 2007), which is in line with the purpose of this study.

As presented in section 3.1, the organizational context of the research inquiry of this study is Systembolaget, a non-profit organization operating in Sweden. As the aim with this study is to explore how BI is used in relation to decision-making in a non-profit organization and to investigate the crucial factors in the usage of BI in relation to decision-making, the methodological approach was formed in accordance with answering these questions. A case

(26)

study design is therefore applied, in line with Yin (2011) who argues that it is adequate to carry through a case study if the purpose is to investigate a specific case, in a detailed way so that research questions addressing how and what can be answered. In this study, the authors seek to contextualize the phenomenon BI and decision-making, thus a case study design is appropriate (Yin, 2011). Moreover, case studies are common in qualitative studies (Bryman & Bell, 2011).

One common misunderstanding presented by Flyvberg (2006) is that a case study is most useful in the beginning of a research process by generating hypotheses. However, Flyvberg (2006) argues that case studies can also be suitable for theory building and hypotheses testing.

4.2 Data collection

In order to answer the research questions, the data was collected by semi-structured interviews.

Because the aim of the study is of explorative nature, a data collection method that enabled contextualized and rigorous data was needed (Bryman & Bell, 2011). Furthermore, Bryman and Bell (2011) explains that semi-structured interviews usually consist of an interview guide with interview questions. However, the questions are most often very broad and where the respondents have a great deal of leeway in how to reply. The strengths of semi-structured interviews are according to Bryman and Bell (2011) that you can go in depth on a specific topic, which is in line with the purpose of this study. Section 4.2.2 and 4.2.3 will disclose more information about the interviews.

4.2.1. Selection of respondents

In the selection of respondents, the authors had the research questions as reference point.

Because the aim with this study is to explore how BI is used in relation to decision-making in a non-profit organization and to investigate the crucial factors in the usage of BI in relation to decision-making, the authors first and foremost had to find respondents who used BI as a tool in their decision-making. Thanks to contacts at Systembolaget, the authors got support in how to get in touch with appropriate respondents. Most of the respondents had a managerial position and their tasks involved decisions across different organizational levels. Also, a prerequisite to be involved in the study was that each respondent had to work with the BI systems on a daily

(27)

basis. In order to bring more nuance and depth to the thesis, the authors searched for respondents working at different departments in the organization. Bryman and Bell (2011) stress that diversity among the respondents leads to a more thorough analysis. Accordingly, the interviewed persons worked within the department of IT, financial controlling, sales, analytics and e-commerce.

Because the objectives of the thesis are to examine how BI is used in relation to decision- making and what the crucial factors in the usage of BI in relation to decision-making are, focus was mainly to find respondents using the end-product of the BI system. In other words, persons using the produced outcome from the BI systems to facilitate their decision-making were searched for. However, in order to get a general understanding about the BI system itself and how it is used to facilitate decision-making, it was necessary to conduct interviews with respondents who had insight and knowledge of the structure in the BI systems. After three interviews with persons working behind the BI systems, the authors had reached data saturation regarding the structure of the systems. Hence, one of the interviews with a System Specialist was cancelled. Consequently, the rest of the interviews were conducted with respondents that mostly use the end-product of the BI system, to facilitate their decision-making. General information about the interviews is listed in table 2 below.

Table 2. Compilation of the interviews

Name Profession Interview

time Interview mode Date

Ulrika Tjerneld BI specialist 52:21:00 Virtual meeting 2020-04-03 Sara Kristiansson Business Controller 50:50:00 Virtual meeting 2020-04-06 Erik Fender Head of Analytics 52:51:00 Virtual meeting 2020-04-06 Mikael Andersson Area manager, Sales 36:49:00 Virtual meeting 2020-04-08 Johan Oinonen Head of Decision

Support

46:35:00 Virtual meeting 2020-04-08 Louise Eggoy System Specialist 46:21:00 Virtual meeting 2020-04-14 Jan Gunnarsson Head of Sales 49:06:00 Virtual meeting 2020-04-15 Linda Carlberg Analyst, E-commerce 42:19:00 Virtual meeting 2020-04-29

4.2.2 Interview guide

In accordance with Bryman and Bell (2011), an interview guide was created (Appendix 1). The questions were carefully formulated to not be hard to understand, misleading or result in short

(28)

answers. However, it occurred that some questions had to be re-formulated depending on the receiver. Furthermore, the interview guide was divided into different themes to ensure a red thread during the interview. Questions which the authors believed would result in answers of similar themes were placed next to each other, in order to avoid shifting themes too much (Kallio et al., 2016). The interview guide was translated into Swedish in order to minimize the possibility of unclarities during the interviews. Because the native language for both the authors and respondents was Swedish, the interviews were held in Swedish. The potential translation problems will be discussed in section 4.4.

Before each interview, the interview questions were sent to the respondent. Although sending the interview questions in advance was requested by some respondents at Systembolaget, the authors believed that it would be beneficial for the respondents to know the terminology in the interview questions before the interview (Kallio et al., 2016). Thus, the interview questions were sent out by email to the respondents the day before each interview, with an encouragement to look through the questions and be prepared to sort out eventual terminology unclarities before the interview started. According to Thomas (2004), there are some disadvantages with sending out the interview questions in advance. The respondents may not answer spontaneously and may have prepared the answers to what they think the interviewer wants to hear. However, the authors estimate that this effect was fairly light and did not affect the answers significantly.

4.2.3 Execution of interviews

When conducting an interview, it is preferable to do it in an environment where the respondent feels safe and comfortable in order to receive as open answers as possible. A physical meeting is most recurring in qualitative research (Bryman & Bell, 2011). Due to a global pandemic affecting the possibility to meet in person, online-interviews were conducted instead of physical meetings. The platform Microsoft Teams was used which enabled interaction via both video and audio.

Both of the authors were present during each interview, where one of the authors conducted the interview, and the other acted as a passive interviewer. This ensured that the questions were correctly asked, and also enabled that one of the authors could interrupt if a follow up question came to mind (Kallio et al., 2016). It occurred that the passive interviewer had to interrupt to

(29)

ask follow-up questions, and in one of the interviews, technical difficulties led to the passive interviewer having to take over the remaining part of the interview.

In order to ensure no ethical harm in the study, the authors have considered the four ethical principles by Bryman and Bell (2011); harm to participants, lack of informed consent, invasion of privacy, and whether deception is involved. First, in consideration of harm to participants, each respondent had the possibility to be anonymous in the paper. However, none of the respondents asked for anonymity. In order to lay the ground for a thorough analysis, all interviews were recorded. In accordance with the ethical consideration of informed consent, the recording was performed with consent from the respondents. With respect to the respondents, all recordings were deleted when the thesis was finished. Consideration was also taken regarding invasion of privacy, since the authors did not ask for any sensitive personal information. In order to avoid deception, the authors described and informed about the purpose and research question to the respondents ahead of every interview. Before each interview, it was assured that each respondent completely understood the interview questions in itself, but also the purpose with the study. The authors also stressed before each interview that the respondents should not be afraid to interrupt, if there was any sense of ambiguity (Bryman &

Bell, 2011).

4.3 Data Triangulation

Data triangulation, to use multiple methods and data sources, is important for studies of qualitative nature in order to increase validity and reliability (Van de Ven, 2007). Data triangulation ensures data saturation, according to Fusch and Ness (2015), who further argues that if the research fails at reaching data saturation, the validity and the quality of the study will be affected. An attempt of collecting additional data was therefore made. During the interviews, the respondents were asked if they knew about any material or report that would help the authors gain a better understanding about how BI is used in the organization or about their decision-making. What emerged was a lack of relevant additional material that could prove useful to gain additional insight through internal documents or other sources. It was not that these materials were difficult to obtain, but rather that such materials did not exist. A modest amount of secondary data was nonetheless collected from Systembolaget’s website and from their Responsibility Report from 2019. In addition, news articles related to Systembolaget were

(30)

searched for. Thus, the results of the article search were not applicable to the research objectives of this study. Given the lack of additional data sources that emerged during the attempt to triangulate data, the interview heavy approach of this study is justified.

Critics on case studies argue that the external validity or generalizability are problematic in these kinds of studies (LeCompte & Goetz, 1982; Van de Van, 2007), with the underlying argument that it is only applicable to the specific case. However, as Flyvbjerg 2006 identifies, case studies nonetheless offer important insights into research and may offer generalizability of certain conceptual understandings that apply beyond the specific case setting. The aim of our particular study is to provide insight into a relatively unresearched area, and to achieve what Lee, Collier and Cullen (2017) describe as particularization, to explain a phenomenon in a detailed way, in a specific case. This, as Lee, Collier and Cullen (2007) argue, is the main strength in case studies.

One potential limitation of this study is that the internal validity in this study may have suffered, because the interviews were not conducted in person (LeCompte & Goetz, 1982), however, the authors estimate that the internal validity can still be considered as high, due to the visual meeting solution and relaxed small talk with the respondent before each interview. The internal reliability of this study is also considered to be high, because in a situation of disagreements during the construction of themes (Bryman & Bell, 2011; Gioia, Corley & Hamilton, 2013), the material was revisited and discussions were made, until consensus was achieved (Gioia, Corley & Hamilton, 2013; LeCompte & Goetz, 1982). The external reliability of this study, to what degree a study can be replicated or not (LeCompte & Goetz, 1982), is regarded as fairly high because the interview guide is included in the thesis (Appendix 1).

4.4 Analysis of the empirical material

After each interview, the recordings were transcribed. As mentioned previously, the interviews were held in Swedish, which led to a translation of both the interview guide and the transcribed material. Xian (2008) emphasizes that it might arise a problem in the translation process, because of the linguistic differences. The interviewee might use words that have no equivalent translation, or a grammatical structure that is difficult to translate. However, the authors sought to make the translation as accurate as possible, without manipulating the empirical material.

(31)

In order to facilitate the analyzing process, the transcription and translation took place within one day after each interview, which is in line with Bryman and Bell (2011) argument that it is appropriate to transcribe the interview as soon as possible, in order to not lose context.

Furthermore, the transcribed material of this study resulted in approximately 90 pages. To analyze large amounts of data from transcribed interviews, which is common when conducting qualitative interviews, can sometimes be challenging (Bryman & Bell, 2011). A common approach is to break down the transcribed material and search for themes and patterns, also referred to as thematic analysis (Bryman & Bell, 2011). In this paper, a similar approach will be undertaken, however it is mostly inspired by the Gioia Methodology.

The analysis of the transcribed material is divided into three different stages: 1st order themes, 2nd order themes and aggregate dimensions (Gioia, Corley & Hamilton, 2013). An example of how the authors have generated different level of themes from the empirical material is illustrated in figure 6 below:

Figure 6. Example to describe the process of analyzing the empirical material

The 1st order themes have a broader scope, meaning that a lot of themes and categories are generated. Consequently, with the 1st order themes as reference, similarities and differences will be generated and the initially pretty vague themes get labels and phrasal descriptors.

Hence, the 2nd order themes are generated. This level of analysis is also meant to include theoretical connections to the empirical findings (Gioia, Corley & Hamilton, 2013).

The abductive approach of the study enabled the authors to get a direction or indication on what kind of theory that should be added or revised, based on the themes and concepts generated from the analysis (Gioia, Corley & Hamilton, 2013). Once theoretical saturation was achieved (Bryman & Bell, 2011; Gioia, Corley & Hamilton, 2013), the analysis moved on to the last stage. The aggregated dimensions serve as the building blocks in the analytical part of this

(32)

paper, because the analysis of the empirical findings will use these dimensions as the reference point. For structural reasons, dimensions produced from analyzing the empirical material will be the headings in the section of the empirical findings as in the analysis and discussion section.

(33)

5. Empirical Findings

In the following chapter the empirical findings are outlined. The purpose of using BI in the decision- making at Systembolaget is elaborated, followed by the relation between decision types. Finally, the empirical findings on BI and decision-making are explained. The last section includes three factors affecting BI and decision-making.

5.1 The purpose of using BI in decision-making

Because Systembolaget is a non-profit organization, they are not working with additional sales.

Instead their mission concerns fostering public health and to limit the harm of alcohol. Thus, they cannot request for money to drive sales, rather they can request money to find a better and smarter way of working. For example, the use of BI in the logistics department can concern flow optimization, which in the end leads to satisfied customers thanks to the efficient way of getting products out of the store and customers. Therefore, the focus is not on the financial ratios. Furthermore, being a public owned company, Systembolaget has a different purpose than commercial companies. The driving force of Systembolaget is not to make extreme amounts of money, but to minimize the harms of alcohol. Therefore, other aspects than only increased earnings are considered.

“(...) there are so many aspects that are at least as important as the bottom-line.

We do not exist with the purpose to have extremely high earnings, we exist with the purpose to limit the harm of alcohol” Jan, Head of Sales

Systembolaget always weighs in both “hard” and “soft” aspects into each decision. The “hard”

data is not the most important. For example, a decision can be bad for the financial results but be good for the public health. Considering all consequences of the decision is important, not only are the financial aspects important but also the societal aspects. Hence, the directions that the data points at may not be the way Systembolaget decides to go, Sara, Business Controller, explains.

There are some aspects that could be regarded as limitations that derive from the purpose of Systembolaget. First, the importance of considering the purpose of Systembolaget is being stressed when choosing which information to bring into the BI systems. There is some

Insights about Business Intelligence and Decision-Making