Degree project in informatics at master level
Trust in Data
Prerequisite for Self-Service Business Intelligence Adoption by Business Users
Author: GUAN Zhong Lai Supervisor: Imad Bani Hani Examiner: Jan Aidemark
Abstract
As data becomes a ubiquitous part of today’s business, trust in data is recognised as a crucial factor for organisations on the data-driven journey to stay competitive in the fast-evolving marketplace. To support the journey, Self-Service Business Intelligence (SSBI) has emerged as a popular approach for organisations to empower business users and gain actionable insight from data faster and better. Despite its importance and relevance at the organisational level, SSBI has suffered sluggish adoption rates at the user level. The purpose of this thesis is to explore the importance of trust in data and how it influences SSBI adoption by business users.
Through seven semi-structured interviews, this thesis is able to establish that: 1) trust in data is a prerequisite for SSBI adoption by business users; 2) business users trust the people behind the data; 3) trust in SSBI tools is essential; and 4) trust in data is necessary for user adoption.
Furthermore, these findings lead to a descriptive model of how trust in data influences SSBI adoption by business users as well as how business users can transition between a vicious cycle of SSBI resistance and a benign cycle of SSBI adoption.
Keywords
Self-Service Business Intelligence (SSBI), Trust in data, Business User, User Adoption
Acknowledgements
The learning and thought process for this thesis would not have been possible without the patience and support from my professors and family. To my professors, thank you for being selfless and wise in your guidance. To my family near and far, I am the luckiest person in the whole world for having you.
Many special thanks to Company X for trusting me, the stars were indeed aligned leading me to your organisation. I most sincerely hope that the findings will be useful on your continued journey towards becoming more data-driven.
Forever grateful!
Stockholm, June 2021.
Table of Contents
1 Introduction ________________________________________________ 5
1.1 Background ___________________________________________________________ 5 1.2 Previous Research ______________________________________________________ 6 1.3 Problem ______________________________________________________________ 7 1.4 Research Question ______________________________________________________ 7 1.5 Purpose ______________________________________________________________ 8 1.6 Delimitation ___________________________________________________________ 8 1.7 Thesis Organisation _____________________________________________________ 9 2 Theoretical Background _____________________________________ 10
2.1 Business Intelligence ___________________________________________________ 10 2.2 Self-Service Business Intelligence ________________________________________ 12 2.3 User Adoption ________________________________________________________ 14 2.4 Trust _______________________________________________________________ 16 2.5 Trust in Data _________________________________________________________ 17 2.6 Summary ____________________________________________________________ 18 3 Methodology _______________________________________________ 19
3.1 Methodological Approach _______________________________________________ 19 3.2 Research Strategy _____________________________________________________ 20 3.3 Techniques for Data Collection ___________________________________________ 21 3.3.1 Company X and Participants _________________________________________ 22 3.3.2 Interview Guide ___________________________________________________ 23 3.4 Techniques for Data Analysis ____________________________________________ 24 3.5 Quality and Validity ___________________________________________________ 25 3.6 Ethical Considerations __________________________________________________ 27 4 Empirical Findings __________________________________________ 29
4.1 Influence of Trust in Data _______________________________________________ 29 4.2 Establishing Trust in Data _______________________________________________ 30 4.3 Understanding SSBI Adoption ___________________________________________ 31 4.3.1 Trust in SSBI Tools ________________________________________________ 32 4.3.2 Trust in Data During User Adoption ___________________________________ 33 4.4 Other Findings ________________________________________________________ 33 5 Discussion _________________________________________________ 35
5.1 Trust in Data Is a Prerequisite ____________________________________________ 35 5.2 Trust in People Behind Data _____________________________________________ 35 5.3 Trust During User Adoption _____________________________________________ 36 5.3.1 Trust in SSBI Tools ________________________________________________ 37 5.3.2 Trust in Data During User Adoption ___________________________________ 38 5.5 Descriptive Model of How Trust in Data Influences SSBI Adoption by Business Users _______________________________________________________________________ 39 5.6 Practical Contributions _________________________________________________ 40 5.7 Reflections on Methodology _____________________________________________ 40
6 Conclusion _________________________________________________ 42
6.1 Concluding Remarks ___________________________________________________ 42 6.2 Future Research _______________________________________________________ 42 References __________________________________________________ 43
Appendix ___________________________________________________ 47
Appendix 1: Informed Consent Form _________________________________________ 47 Appendix 2: Pre-interview Guide ____________________________________________ 48 Appendix 3: Interview Guide _______________________________________________ 49
1 Introduction
1.1 Background
Trust is recognised as a crucial factor in dealing with uncertain, uncontrollable, and risky situations, a description that fits the data practices of organisations in the digital age (Steedman, et al., 2020). As the world continues to digitalise, the collection, possession, and analysis of data will offer unparalleled potential for organisations to improve efficiency and develop better products and services (Cappa, et al., 2020).
The availability of zettabytes of data presents both unprecedented opportunities and challenges for organisations (Cappa, et al., 2020). On the one hand, organisations can outmanoeuvre competitors with data-driven decisions; on the other hand, organisations are struggling to keep up and to find new value from their data (LaValle, et al., 2011; McAfee & Brynjolfsson, 2012;
Howson, 2014). To know what happened and why it happened are no longer adequate, the race is to quickly process data to gain insight into what is happening now, what might be around the corner, and which actions to take to achieve optimal results (LaValle, et al., 2011). Herein lies the “double-edged sword”, organisations must balance adoption of new technology with ensuring a high, stable level of trust in the ever-changing data landscape (KPMG, 2016).
In response to the opportunities and challenges presented by data, organisations have been investing substantially in digital infrastructure during the last two decades (Provost & Fawcett, 2013) and are increasingly turning to software systems for assistance (LaValle, et al., 2011).
One of the central software systems that organisations implement for faster business insight is Business Intelligence (BI) (Howson, 2014). Research results from Logi (2017) confirm this view and show that 66% of organisations have either started to or already fully implemented BI. The same research also indicates that only 7% of organisations have never considered BI.
Since its introduction to the business arena as mainframes during the 1980s, BI has undergone many stages of technological evolution (Howson, 2014). Alpar & Schulz (2016) points to two fundamental changes in recent years that are fuelling BI development: abundance of new types of data that are different from traditional operational data; shift of scope from strategic questions to operational tasks. In addition, organisations recognise the need for making data analytics available to everyone (Bani-Hani, 2020) so that organisations can gain access to and extract value from data at the speed of thought (Howson, 2014). To address those evolving business needs, Self-Service Business Intelligence (SSBI) emerges as a new approach of BI (Lennerholt, et al., 2018).
In broad terms, SSBI empowers business users to be self-reliant in data access and analysis as well as report creation and sharing (Michalczyk, et al., 2020). IT departments and BI teams no longer serve as intermediaries between users and the data, users can indeed generate reports at any time (Eckerson, 2009). Despite its popularity and importance at the organisational level, the adoption rate of SSBI is stagnant at the user level, remaining in the range of 50-55% since 2014 (BARC, 2017). This sluggish adoption rate is evidential of the many challenges that SSBI deployments encounter in practice (Johannessen & Fuglseth, 2016; Lennerholt, et al., 2018;
Weiler, et al., 2019).
Ain, et al. (2019) points to individual-level acceptance and use to be one of the key challenges of SSBI adoption. An organisation’s goal of making better, faster decisions based on insight through data is highly dependent on the effective utilisation of BI systems, which in turn depends on the end-users (Ain, et al., 2019). This sentiment is shared by McReynolds (2018) who poses the question “What if the final gap, the issue holding us back from data-driven answers, is not a technology barrier at all? What if it is something very human?”. In response, this thesis attempts to bring the human perspective into the spotlight and explores the relationship between trust in data and SSBI adoption by business users.
1.2 Previous Research
In spite of its recent rise to popularity, there is plenty of literature on SSBI. A systematic review of 60 articles published between 2009 and 2019 on this topic reveal three perspectives in literature (Michalczyk, et al., 2020). Michalczyk, et al. (2020) find that those perspectives are not mutually exclusive but have different focus areas:
- Artefact-centric view
Emphasises the technical solutions that support users navigate data and make well-informed decisions
- User-centric view
Emphasises the drivers for user acceptance and the potential negative effects of not putting those drivers in place
- Governance-centric view
Emphasises the importance of data management principles and their trade-off with flexibility
While acknowledging that technical solutions and data management principles are essential, this thesis leans more towards the user-centric view as users are ultimately the ones who adopt the SSBI tools. Findings by Michalczyk, et al. (2020) also confirm the relevance of the user- centric view: 82% of the articles reviewed take the user-centric view when discussing SSBI adoption. Furthermore, the user-centric view has the highest relevance amongst the 16 classifications presented in the concept matrix by Michalczyk, et al. (2020). A closer examination of some of the most quoted articles written from the user-centric view reveals that there are, however, some finer differences within this view.
A somewhat earlier study by Imhoff & White (2011) argues to make BI tools easy to use, source data easy to access, BI reports easy to consume and enhance, and data warehouse solutions fast to deploy and easy to manage. Those user-friendly functionalities and features are especially important when there is ambiguity in data knowledge amongst different users. Significant relationship is identified between users’ knowledge of data modelling and their ability to support decision making with SSBI tools (Johannessen & Fuglseth, 2016). The authors also suggest training for novice users as a potential topic for future research. Not surprisingly, SSBI education is one of the categories of challenges identified by Lennerholt, et al. (2020). Research on user uncertainty also contributes to the user-centric view. Such uncertainties arise as user tasks, workflows, and environment change during the implementation of SSBI, leading users to distance themselves from SSBI (Weiler, et al., 2019). Passlick, et al. (2020) offer yet another approach and emphasise the importance of flexibility, expected time savings and data quality as factors leading to the intention of using SSBI applications.
1.3 Problem
The global Business Application Research Centre (BARC) BI survey reveals that 58% of organisations base at least half of their regular business decisions on gut feel or experience rather than data and information (BARC, 2017). In addition, on average, only 50% of all available information in organisations is actually used for decision making (Bange, 2014). With the abundance of data and the availability of user-centric software systems such as SSBI to analyse it, what is preventing organisations from becoming more data-driven? What is holding business users back from SSBI adoption? McReynolds (2018) believes that users are held back because they do not trust the data and the analysis. Trust in data is comparable to the default hygiene factor in the laboratory environment in order for users to use the insight derived from the data to make decisions and ask new questions (MIT, 2019). However, this trust in data may be lacking.
The sentiment above is echoed in the business world. One of the largest global consulting firms, KPMG (2016), sees a trust gap between the vast majority (≈ 70%) who acknowledges how critical data is to their organisations, and the small minority (<30%) who seems to trust the analytics they generate. Organisations need trusted data to operate successfully and to create long-term value in the business ecosystem (Vernocchi, 2021). Another consulting giant, PwC (2020), goes a step further to state that “Yet data is not enough, because it is worse than useless – it is a source of risk – unless you can trust it”.
The researcher agrees with the above views and sees trust in data as fundamentally important in the SSBI context, especially from the perspective of business users. Organisations and software vendors have prioritised product improvement and data management while skirting the topic of trust in data. This is partially because of the natural progression for technology adoption, but also largely because tackling trust in data would involve the messy situation of dealing with humans. For organisations to be truly data-driven and realise the value of their investments in SSBI tools, they will have to take on the topic of trust in data in order to nudge people towards change. These thoughts lead to the research question in the next section.
1.4 Research Question
Since empowering business users to analyse data and derive actionable insight from it independently is at the core of SSBI (Alpar & Schulz, 2016), it is highly relevant to understand the adoption challenges this particular user group faces. Business users by definition do not have technical BI skills (Lennerholt, et al., 2020). Not only does their lack of knowledge perturb business users, but also the lack of training programs, which used to be the rule when software tools were previously introduced (Weiler, et al., 2019). Without adequate skills and proper understanding in place, business users find it difficult to trust SSBI tools. When there are introductory courses, dummy data is commonly used for exercises and drills. In other words, business users are unlikely to have practised SSBI tools using the data they will actually perform analysis on. After training, business users are expected to continue to use SSBI tools and be self-sufficient, albeit in a controlled environment. As a result, Weiler, et al. (2019) find that business users forget despite training, do not use SSBI tools after training, and in general lose interest in SSBI tools.
To stimulate adoption, some organisations actively make alternatives less attractive or simply unavailable, leaving business users with limited or no choice (Liu, 2012). Business users find themselves in the situation where SSBI tools have become a part of their work, but there is inadequate communication from management within organisations on the purpose of the SSBI tools to establish the necessary trust. Business users are expected to use SSBI independent of IT professionals and BI specialists when there is actually a wide spectrum of user skills and technical understanding (Alpar & Schulz, 2016; Weiler, et al., 2019; Lennerholt, et al., 2020).
With little or no formal training to guide them, business users find it difficult to access data, manipulate content, and assure quality, let alone derive actionable insights. Furthermore, business users may perceive SSBI tools as a disturbance to the existing social dynamics based on teams and teamwork. There could even be concerns that SSBI will eventually make the business users’ skillset redundant.
The important topic of trust in data is not addressed during SSBI adoption. With minimal involvement from IT professionals and BI specialists as intermediaries between data and users, business users in the SSBI environment are largely on their own to establish trust in data, continue to use the SSBI tools, and make decisions based on data analysis. This thorny situation deserves some exploration and leads to the research question below:
“How does trust in data influence SSBI adoption by business users?”
In answering this research question, organisations and software vendors will gain a better understanding of the relationship between trust in data and SSBI adoption by business users.
Without trust in data, business users will continue to resist SSBI adoption and organisations will be unable to realise the full potential of data. Organisations and vendors need to collaborate and find ways to build trust in data so that business users integrate data into decision-making and become more data-driven.
1.5 Purpose
The purpose of this thesis is to explore how trust in data influences SSBI adoption by business users. The hope is to contribute to academic research with the following: 1) confirm the importance of trust in data in the context of SSBI adoption by business users and 2) identify how trust in data influences SSBI adoption by business users . It is also hoped that this thesis will make a positive impact in the business world in two ways: 1) provide insight to practitioners so as to improve the success rate of SSBI projects and 2) encourage business user adoption so as to increase the number of decisions taken based on data.
1.6 Delimitation
To start with, this thesis excludes organisations that do not currently have SSBI or are still in the implementation stage. As the name suggests, self-service becomes relevant only when the software applications are already in place. Otherwise, the applications are not ready for use and there is nothing for users to perform self-service on. The types of organisation that fit the research interest of this thesis are those that have just transitioned from implementation to adoption and those that have adopted SSBI for a number of years.
A second distinction to make is the level of SSBI adoption. When there is widespread usage within an organisation and SSBI is an integral part in a firm’s value chain activities, this organisation has reached the other side of the SSBI adoption tipping point and should be studied for best-practice purposes. This thesis looks at organisations that are actively working with increasing user adoption.
Thirdly, this thesis does not study SSBI adoption by power users. The reasoning is that the ultimate goal of SSBI is to empower business users in accessing and analysing data to derive actionable information without involvement from technical specialists (Alpar & Schultz, 2016).
In other words, business users are the intended target audience for SSBI. The definitions of those two user groups and the distinction between them are presented in the section on Theoretical Background.
1.7 Thesis Organisation
The remainder of the thesis is organised as following: the next section unpacks the research question into its constituent parts and provides definitions as well as reflections on each component. Next, research methodology is introduced with descriptions and justifications for data collection, data analysis, as well as quality and validity. A short description of the participating company, Company X, is also made available in this section. This is followed by the empirical findings and analyses on how trust in data influences SSBI adoption by business users. In the section on discussion, theoretical contributions and practical limitations on the findings are presented. Finally, the thesis closes with conclusions and recommendations for future research.
2 Theoretical Background
In this chapter, the thesis question is unpacked into constituent parts. Literature on and definitions of BI, SSBI, user adoption, trust, and trust in data are presented as guidance for data collection and analysis in order to answer the thesis question.
2.1 Business Intelligence
Decision-support systems were introduced during the 1970s to support decision making within organisations (Watson & Wixom, 2007a). Such systems were vastly different from the then existing transaction-processing or operational applications and underwent multiple transformations to evolve into what is known as BI today (Watson & Wixom, 2007a). BI can take on different shades of meaning to different people, it is often associated with concepts such as competitive intelligence, reporting, analysis, business analytics, big data, visual data discovery, and many more. Although there is no unified definition of BI, this thesis uses the description put forth by Howson (2014, p. 1-2):
BI is a set of technologies and processes that…allow people at all levels of an organization to access, interact with, and analyse data to manage the business, improve performance, discover opportunities, and operate efficiently.
BI deployments often start with constructing an underlying BI architecture and this architecture consists primarily of two parts: the BI front-end and the BI back-end (Howson 2014). Figure 1 is an extremely simplified depiction of a BI architecture.
Figure 1: BI architecture to demonstrate concept of front-end and back-end (created by author)
The BI back-end is commonly considered the responsibility of IT departments. Users are rarely exposed to the BI back-end and hence have limited understanding of as well as appreciation for what takes place there (Howson, 2014). As Figure 1 illustrates, everything begins with data.
Data is collected from a variety of internal and external sources (Gibson & Arnott, 2005), with operational systems being the starting point for most quantitative data in an organisation (Howson, 2014). Operational systems collect data from operational tasks such as manufacturing, sales, supply chain, and accounting. Enterprise Resource Planning (ERP)
systems are then implemented by organisations to reduce duplicate data entry and ensure adherence to standard processes (Howson, 2014). Additional data sources can vary from legacy systems, click-streams, social media, and wearables to advertisers and third-party data providers.
As the sheer number of data sources makes apparent, the data that organisations collect is diverse in aspects such as content, format, and location. Therefore, data needs to be extracted from source systems and transformed so that it is consistent, meaningful, and optimised for decision support (Watson & Wixom, 2007a). Some typical tasks of data transform will handle missing value, inconsistent coding, incomplete data, and something trivial yet important like spelling mistakes. The “clean” data is then loaded into a data warehouse for analysis. This is in essence the extract, transform, and load (ETL) process during the data transform stage. ETL is often viewed as the most critical, challenging, and time-consuming part of BI (Watson &
Wixom, 2007a).
A data warehouse can be seen as a central repository that is subject-oriented, integrated, time- variant, and non-volatile (Watson & Wixom, 2007a). From here, users and applications can access data to perform enterprise reporting, Online Analytic Processing (OLAP), querying, and predictive analytics (Watson & Wixom, 2007a). Of course, each BI architecture is different and BI architectures continue to evolve with advancements in technology. Figure 2 provides a simplified visual of the BI back-end.
Figure 2: BI architecture to demonstrate concept of BI back-end (Adapted from Howson 2014)
When users interact with BI, they most certainly interact with the front-end tools. Those front- end tools are provided by software vendors and connect seamlessly to the back-end. With intuitive features and endless functionalities, those sleek interfaces are designed to make the front-end tools appealing to the users. Business query and reporting, dashboards and scorecards, production reporting, and visual data discovery are some of the tools available at the BI front- end (Howson, 2014). The four vendors in the Leader Quadrant identified by Gartner (2020) are Microsoft, Tableau, Qlik, and ThoughtSpot. Figure 3 is a screenshot of Microsoft’s Power BI Desktop which exemplifies what the BI front-end could look like.
Figure 3: Screenshot of Microsoft Power BI Desktop
(Available online: https://powerbi.microsoft.com, Accessed February 9, 2021)
In the traditional BI environment, IT professionals and BI specialists act as the intermediaries between data (the BI back-end) and users (the BI front-end). Those professionals and specialists are the curators and gatekeepers of trusted data and knowledge, they are also the trusted sources of information and solution. While seen as a bottleneck for accessing data, this arrangement actually helps with building trust. Organisations typically have a small group of specialists who perform data analysis and make it available to the rest of the organisation. Knowledge is passed from specialist to specialist, and those one-to-one relationships are a sufficient foundation for trust (McReynolds, 2018). In the SSBI environment, this foundation for trust is severely eroded.
2.2 Self-Service Business Intelligence
While Weiler, et al. (2019) see SSBI as new possibilities to utilise BI by providing a universal and accessible platform, Alpar & Schulz (2016) think SSBI should enable business users to engage with multifaceted data without having to involved BI specialists. Bani-Hani (2020) considers SSBI more as the technology readiness within the organisational environment and the willingness of a user to engage in self-service activities using the resources available for the ultimate aim of solving an analytical task independently. There is therefore still vagueness in
the exact meaning of SSBI. This thesis concurs with the view that SSBI is an approach of BI to address evolving business needs and uses the definition from Lennerholt, et al. (2020, p. 188):
SSBI is an upcoming trend allowing non-technical casual users to use BI in a self-reliant manner without the support of technical power users.
This definition of SSBI can be visualised by curve A in Figure 4 (the business employees and techno-oriented employees in this figure are the equivalent of business users and power users in this thesis). Business users represented by curve A are in the “no dependence mode” and are expected to accomplish complex tasks such as data preparation and data gathering by themselves (Bani-Hani, et al., 2019). As business users in this mode can solve an analytical task fully independently from power users, this is the mode of SSBI organisations aspire to achieve (Bani-Hani, et al., 2019). All curves under A, inclusive of curve B (low dependence mode) and C (high dependence mode), will require varying degrees of support from power users.
Figure 4: Three modes of engagement in an SSBI environment (Bani-Hani, et al., 2019)
With the aim of empowering business users make faster and better decisions based on data, organisations strive after shifting from curve C and B towards curve A. As such, at the same time as business users are empowered to explore data independently, the coordination effort between business users and BI specialists is minimised or even eliminated (Alpar & Schulz, 2016). If done too hastily and without taking trust into proper consideration, this decrease in coordination effort could have the undesirable effect of distancing business users further away from the technology they have no control over and the data they cannot verify (McReynolds, 2018).
Another important aspect of SSBI to understand is the nature of self-service technology (SST).
SSTs are technology interfaces that enable users to produce a service independent of direct service employee involvement (Meuter, et al., 2000). Considine & Cormican (2017) argue that SSTs benefit organisations and users differently: organisations are enthusiastic about SSTs as they can help increase service efficiency, reduce operational expenses, and enable service delivery anytime anywhere; user benefits centre around improved user experience. However, the shift towards SST undoubtedly requires increased effort on the part of the user in terms of participation, learning, and taking responsibility in the production of the service (Considine &
Cormican, 2017). Organisations need to therefore pinpoint the needs of users and consider their perceived benefits in order to encourage users to put in the extra effort required to accept SSTs (Considine & Cormican, 2017). Conversely, research has shown that forced SST adoption is perceived as a threat to user freedom and causes users to resist and/or switch (Feng, et al., 2019
& Liu, 2012).
It is also central here to recognise that there are different levels of self-service within SSBI.
Alpar & Schulz (2016) categorise self-service into three levels depending on the system support required: at the most basic level, users gain access to ready-made reports with the possibility to set parameters or “drill anywhere”; next up, users are allowed to create new information from data at the lowest disaggregated level as well as the opportunity to autonomously perform advanced analytics beyond historical data; at the highest level, users are permitted to autonomously capture new data sources and combine different functionalities using reusable components through mashup. Such categorisations help provide a better sense of control over SSBI tools for users, they do not, however, address the fundamental issue of lack of trust in data which is more challenging to establish in the SSBI environment.
2.3 User Adoption
Howson (2014) presents a spectrum of SSBI scenarios with varying degrees of user sophistication, IT involvement, and user capability. As shown in Figure 5, SSBI is a continuous spectrum ranging from minimal technical skills to high technical skills. The fit between users and the self-service level is therefore determined more appropriately by the users’ computer skills, analytic skills, information demands, and specific tasks (Eckerson, 2014) rather than their business function or seniority.
Figure 5: SSBI continuum (Adapted from Howson 2014)
While there are a few different approaches to categorise them, SSBI users are commonly divided into two groups: casual/business users and power/technical users. For simplicity and consistency, this thesis uses the terms business users and power users with the following definitions (Lennerholt, et al., 2020, p. 188):
Users who have technical skills to build and run BI efficiently are the power users; and
Casual users are the opposite to power users since they do not have technical BI skills. Instead, they are using different foundations such as pre-defined reports or dashboards to make decisions.
As stated under Delimitation, business users are in scope for this thesis. User adoption is therefore examined through the lens of business users and within the framework of technology acceptance model (TAM). TAM is believed to be the most suitable model for this thesis because it also takes a user-centric view and evaluates technology acceptance from an individual perspective. Two additional frequently cited Information System (IS) adoption models are DeLone & McLean’s (D&M) IS success model and the diffusion of innovation (DOI) theory. Whereas D&M proposes six dimensions (information quality, service quality, system quality, use, user satisfaction, and net benefits) to measure IS success (DeLone &
McLean, 2003), DOI combines a macro process (concerned with the spread of an innovation from its source to the public) and a micro process (concerned with stages individuals go through when deciding to accept or reject an innovation) (Jiang, 2009). Both D&M and DOI include influences from organisational and environmental attributes (Ain, et al., 2019; Jiang, 2009;
Rouhani, 2018).
TAM proposes two instrumental beliefs in explaining a user’s intention to use IS, namely perceived usefulness (PU) and perceived ease of use (PEOU) (Jiang, 2009). The latter of the two, PEOU, is found to be the stronger predictor of users’ behavioural intention in the BI system context (Ain, et al., 2020). Findings from two separate qualitative studies by Weiler, et al.
(2019) and Lennerholt, et al. (2020) reveal, unfortunately, that some users perceive SSBI to be neither useful nor easy to use. Instead of adopting SSBI, users are resisting and distancing from SSBI tools due to the many uncertainties and challenges they face. Figure 6 and Figure 7 are thematic maps of the uncertainties and challenges identified.
Figure 6: User uncertainties during the implementation of SSBI (Weiler, et al., 2019)
Figure 7: User uncertainties during the implementation of SSBI (Lennerholt, et al., 2020)
Why do some business users perceive SSBI tools to be neither useful nor easy to use? Do the findings on uncertainties and challenges offer a complete understanding of user resistance in SSBI adoption? Or is there a more fundamental piece missing from discussions, the utterly human aspect of trust.
2.4 Trust
Trust is a complex concept and has been studied through the viewpoints of philosophy, sociology, psychology, management, marketing, ergonomics, and more recently human- computer interaction and e-commerce (Paliszkiewicz, 2011). In sweeping strokes, the modern conceptualisation of trust begins with the social perception of gentility, integrity, and credibility (qualities embodied by gentlemen of noble birth) and evolves to include objectivity (endorsed by scientists, mathematicians, engineers and the like) with advancements in science and technology during the 19th century. Today, objectivity goes hand in hand with quantification through numbers (Passi & Jackson, 2018).
In complex organisational settings, which is the environment most relevant for this thesis, trust involves employees’ willingness to be vulnerable to their organisation’s actions (Paliszkiewicz, 2011). White & McCarter (2013, p. 57) share a similar line of thought and identify two fundamental components regarding the object of one’s trust for information sharing within organisations:
1) freedom from fear that the person, group, or organisation will cause you, your group, or your organisation physical, material, or psychological harm (benevolence based trust); and
2) confidence in the veracity and accuracy of information provided by the person, group, or organisation (competence based trust).
In her literature review on trust management, Paliszkiewicz (2011) identifies three general factors that make it difficult to build trust: it is an interactive process that involves (at least) two individuals learning about each other’s trustworthiness; it is a dynamic process based on
continuous feedback and reinforcement; it is built gradually and incrementally but is lost quickly and difficult to regain. The Dutch proverb summaries the concept of trust neatly: trust comes on foot and leaves on horseback (KPMG, 2016).
Bearing in mind the uncertainties and challenges described under User Adoption, it can be therefore argued that the environment in which SSBI business users are in is not conducive to building trust. To begin with, there are narrow grounds for benevolence based trust. In addition to technology opacity, users can become uncertain due to a lack of understanding for the intentions with SSBI implementation and its effect on their future career prospects. With the decreased presence of IT professionals and BI specialists as intermediaries between data and users, competence based trust is also trickier to build. How to assure quality of reports generated by other users? How to verify the insights generated by users themselves before basing decisions on them? Is robust data governance in place to ensure data quality from source through ETL?
If broad trust is too elusive to establish in the SSBI adoption context for business users, what about the narrower concept of trust in data?
2.5 Trust in Data
As the study by Steedman, et al. (2020) shows, a complex range of factors come into play in relation to trust in data: who is trusting (or not); who or what is being trusted; contexts of trust;
and degrees of trustworthiness. Trust in data must be built, not only through robust data governance structures to ensure data security, quality, and lineage, but also by addressing the social aspect of foundations, boundaries, and norms of trust (McReynolds, 2018). McReynolds (2018) stipulates that in the self-service context, organisations need to establish a system that captures the people, processes, and data involved in analytics discovery in order to support trust in data and trust in analytics.
In the absence of such a system as described by McReynolds, it is challenging for users to establish and maintain their trust in data throughout its lifecycle. A study from KPMG (2016) shows that while trust is strongest at the data sourcing stage, it falls apart as data is moved along the different stages of its lifecycle. Another insight from the same study shows that despite different levels of investment and maturity amongst the organisations surveyed, more sophisticated data and analytics tools do little to enhance trust. In other words, the trust gap cannot be closed by simply investing in better technology (KPMG, 2016).
Zooming in, the complex factors influencing trust in data are considered to gain insight on SSBI adoption by business users. First, who or what is being trusted (Steedman, et al., 2020). In the SSBI context, data “changes hands” numerous times before ending up in front of a business user. Organisations are overwhelmed by data and struggle with providing documentation for data provenance, the ability to record all data derivations from its origin until the final data product (Sacha, et al., 2016). In addition, there is a lack of transparency and quality assurance on reports created by other users within the organisation. It is in fact quite infeasible for business users to pinpoint who or what to trust.
Secondly, contexts of trust (Steedman, et al., 2020). At the organisational level, business users are expected to derive actionable insight with operational, tactical, and strategic impact from the SSBI tools. At the personal level, business users are also likely to be evaluated by the quality
of their insight and consequence of their decisions. There is a lot at stake and trust in data, on which business decisions are made, is paramount but lacking.
Thirdly, degrees of trustworthiness (Steedman, et al., 2020). The study from Passi & Jackson (2018) shows that it is hard even for data scientists to articulate and ascribe trustworthiness to approaches and insights under the circumstances of large-scale data coupled with the complex and opaque models. Business users who lack specialised knowledge have little, if any, chance to judge or establish trustworthiness of data.
2.6 Summary
Based on the discussions in the previous sections, this thesis has uncovered the complexity of SSBI, the challenges business users face, and the lack of trust in data in the SSBI context. The understanding so far is trust in data is important in encouraging business users to adopt SSBI tools with the ultimate goal of becoming more data-driven and competitive in the marketplace.
The following section will describe the methods used to validate this understanding as well as research conducted to uncover on how trust in data influences SSBI adoption by business users.
3 Methodology
In this chapter, the methodological choices taken to conduct this thesis are presented. This section includes a description of the research strategy, the participating company, the respondents, and ethical considerations.
3.1 Methodological Approach
The question this thesis sets out to answer is how trust in data influences SSBI adoption by business users. As covered earlier, there are established theories on the constituent parts of the research question, but not specifically on how all the components connect together. In order to specify this relationship, the deductive approach is used in the capacity of a theory testing process which commences with an established theory or generalisation and seeks to find out if the theory applies to specific instances (Hyde, 2000). The inductive approach is not applicable here as this thesis does not start with observations of specific instances to establish generalisations about the phenomenon under investigation (Hyde, 2000).
Given the level of technical knowledge that defines this user group, the potential impact of their actions based on the insights garnered, and the central role they play in the process, adoption by business users ultimately determines the outcome of SSBI projects. This thesis explores the human aspect of trust in data and how it influences business users in their decision to adopt or resist SSBI tools. As this thesis attempts to understand human thought and action in social and organisational contexts (Klein & Myers, 1999), it lends itself particularly well to qualitative methods of research.
Unlike the quantitative approach, which seeks to obtain accurate and reliable numerical data that enables statistical analysis, the qualitative methodology intends to understand a complex reality and the meaning of actions in a given context (Queirós, et al., 2017). Since the mid- 1990s, qualitative research has been accepted as a legitimate enterprise by the wider IS research community (Sarker, et al., 2013). One of the qualitative research techniques, the case study strategy, allows IS to be studied in a natural setting in order to understand the nature and complexity of the processes taking place (Benbasat, et al., 1987). Benbasat, et al. (1987) recommend researchers to judge the appropriateness of the case study strategy by asking the following questions:
- Can the phenomenon of interest be studies outside its natural setting?
No, SSBI adoption by business users cannot be studied outside its organisational setting as such a study would not make any sense or have any value. It is the messiness of the phenomenon in its natural setting that adds richness and depth to the understanding of human behaviour.
- Must the study focus on contemporary events?
Yes, SSBI is a relatively new trend that is evolving all the time. Studying non-contemporary events would date the study and render it of significantly less valuable.
- Is control or manipulation of subjects or events necessary?
No, it would be counterproductive to control or manipulate subjects or events, the goal is to find out the “how”, not “what if”.
- Does the phenomenon of interest enjoy an established theoretical base?
No, given its relative novelty, there remains many areas of SSBI that deserve academic attention. More specifically, no prior research is found on the topic of exploring the relationship between trust in data and SSBI adoption by business users.
Based on the answers above, it can be argued that the case study strategy is suitable. The case study strategy can be broken down further into positivist (formal propositions, quantifiable measures of variables, hypothesis, and inferences), critical (social critique to eliminate unwarranted alienation and domination), and interpretive (understanding phenomena through context and meaning) (Klein & Myers, 1999). This thesis has neither formal propositions and inferences nor the intention to emancipate restrictive and alienating conditions of the status quo.
Instead, this research focuses on understanding the dynamics present within single settings (Eisenhardt, 2013). The interpretive case study strategy is therefore selected as the research methodology for this thesis.
3.2 Research Strategy
Once the research question was formulated, some flexibility was retained to allow for influence from study results. The intuition was strong relationships existed between trust in data and SSBI adoption by business users, but the exact nature of this relationship was previously unclear and undefined. This relationship could be that the former was a foundation for the latter, or a catalyst, prerequisite, enabler, or as McReynolds (2018) put it, the final mile of the analytics journey and the key to data-driven decision making.
In designing the case study to explore this relationship, recommendations from Benbasat, et al.
(1987) were employed as guidelines. First, business users were identified as the unit of analysis.
As stated earlier, the ultimate goal of SSBI is to empower business users in accessing and analysing data to derive actionable information without involvement from technical specialists (Alpar & Schultz, 2016). Business users were the natural subjects of this research. It was also hoped that results from this research could be generalised to business users in other organisations under similar circumstances. Such circumstances would dictate that organisations either have just transitioned from implementation to adoption or have adopted SSBI for a number of years without reaching widespread use. In other words, the relationship between trust in data and SSBI adoption by business users within one organisation would be expected to be the same as within other organisations.
Next, two aspects were taken into consideration when choosing between single-case versus multiple-case designs. As a start, this thesis was exploratory with no known established theory on the same topic. The terminologies for stages of case research programs offered by Benbasat, et al. (1987) indicated that single-case design was well-suited for the exploration stage. Another advantage with the single-case design was the possibility to reduce unnecessary variables influencing the results. With business users from a single organisation, variances such as formal training, exposure to SSBI, and the SSBI tools deployed could be kept at a minimum. The studies could then zero in on the role of trust in data in SSBI adoption by business users.
The type of organisation that was included in the site selection process was aligned with Delimitation. It needed to be an organisation that had just transitioned from implementation to adoption or had adopted SSBI for a number of years without having reached widespread use.
Criteria like industry, size, nature, or geographical presence were not important. However, given the nature of SSBI, organisations that could qualify for site selection would tend to be
larger entities with a leader or leading market position and international if not global operations.
The organisation participating in this research might wish to understand and avoid SSBI resistance, or to understand and resolve adoption resistance from business users. Either way, this thesis hoped to add-value to the participating organisation through gaining deeper understandings of its business users and contributing to widening SSBI adoption.
3.3 Techniques for Data Collection
A number of data collection methods can be employed in case studies, Benbasat, et al. (1987) list five categories: documentation, archival records, interviews, direct observation, and physical artifacts. While no single method is absolutely right or wrong, interviews are viewed as a suitable method in situations where probing, open-ended questions need to be asked (Adams, 2015). Interviews are conducted conversationally with one respondent at a time and can include a mixture of focused and open-ended questions to explore the “how” and “why” of the research question (Adams, 2015). Although dialogues can meander, interviews are not entirely without structure. Blandford (2013) describes semi-structured qualitative studies as qualitative approaches that have explicit theoretical or methodological structure but are not completely structured. Such studies typically involve systematic, iterative coding of verbal data (Blandford, 2013) with the ultimate goal of obtaining a rich set of data surrounding the specific research issues, as well as capturing the contextual complexity (Benbasat, 1987). Since their characteristics were aligned with the approach of this thesis, semi-structured interviews were chosen as the technique for data collection.
The ongoing pandemic has profoundly changed our lives in more ways than imaginable, one of which is social distancing. Under the guidelines of health officials, physical meetings deemed noncritical and nonessential in nature are best to be avoided. In light of those recommendations and with respect to each other’s wellbeing, all interviews in this thesis were conducted one-on- one in the virtual setting. With the development in technology and rapid uptake in social media, video conferencing was becoming increasingly accepted as a means to conduct interviews by the qualitative research community (Gray, et al., 2020) even prior to the pandemic. Video conferencing offers unique opportunities to gain access to larger and more diverse populations, eliminate time wasted on travel, reduce unpredictable circumstances that would deter a physical meeting, and reduce total cost (Gray, et al., 2020). In addition to not compromising the overall interview quality, Gray, et al. (2020) find video conferencing to be a convenient and user- friendly way to collect data. More importantly, during the time of Covid-19, video conferencing has the added benefit of being compliant with social distancing measures.
This is not to say that there are no drawbacks to conducting interviews via video conferencing.
One of the biggest challenges is distractions caused by technical difficulties such as sound and/or video disruptions due to unreliable internet connections (Gray, 2020). Another challenge comes hand-in-hand with the flexibility video conferencing offers, interference from the environment (Gray, 2020). Participants in interviews may be at the office, at home, in the car, or at a public place during the video conference, exposing the session to unwanted noises or activities competing for attention. And lastly, the new phenomenon of “virtual meeting fatigue”
as practically all meetings are moved online due to the pandemic. The extra effort required to focus and absorb information, the constant gaze to show interest, and the sheer number of online meetings contribute to draining people’s energy and tiring them out (Fosslien & Duffy, 2020).
To help alleviate those challenges, a short checklist was sent to participants ahead of the interview. Appendix 2 shows a generic sample of such a pre-interview guide.
3.3.1 Company X and Participants
After reaching out to a network of contacts for suggestions, Company X was identified as a potential candidate. Company X is a leading financial services provider with a global presence.
Some 10 years ago, Company X embarked on a transformation journey to become a more digital organisation. During the more recent years, its transformative endeavours began to focus more on data and analytics. As a result, Company X implemented a number of software tools, including BI/SSBI aimed at reaching broader levels of data-driven decision-making. Due to legacy systems and budget constraints, not all employees have access to all BI/SSBI tools.
Despite sincere efforts from relevant teams, internal survey results consistently revealed a lack of enthusiasm amongst users towards BI/SSBI tools. Since users explicitly identified trust in data as the major blocker for not working in a more data-driven manner in the latest survey, Company X would like to dig deeper so as to understand and remedy this issue. Those were the perfect circumstances for this thesis, Company X was approached and graciously agreed to be involved.
As for the selection of participants for interviews, it was a decision made jointly with Company X. First, the mutual understanding was to have a handful of deep and insightful discussions that would shed light on the question at hand rather than numerous meetings that barely skimmed the surface. Secondly, some calibration in definition was necessary as Company X did not distinguish between BI and SSBI strictly and had different terminology for its user groups.
Once aligned, only SSBI business users as defined in this thesis were qualified to be considered as potential participants. Thirdly, potential participants were selected with diversity in mind, covering different nationalities, departments, seniority level, and gender.
Company X acted as the intermediary and approached 30 business users via email with the opportunity to participate in this research. This approach was more practical because Company X had information on the roles of its employees and had access to employee calendars. It was not possible to verify whether the business users contacted also participated in the recent company survey, the reason being this survey was anonymous. However, all 30 business users had clicked on the survey link. The message communicated was Company X would like to understand business users’ experience of SSBI, specially in relation to trust in data, so as to identify areas of improvement. After a period of two weeks, eight business users responded positively. One of the eight people unfortunately did not show up for the scheduled interview, seven discussions were therefore successfully conducted for analysis. Table 1 provides an overview of the participants interviewed for this research.
Table 1: Overview of interview participants
ID Role SSBI Usage Interview Duration
P1 Sales Infrequent March 29 50 min
P2 Sales Frequent March 30 40 min
P3 Finance Infrequent March 31 50 min
P4 Content Indirect April 6 50 min
P5 NA NA NA NA
P6 Risk Frequent April 7 60 min
P7 Customer Frequent April 8 40 min
P8 Transformation Indirect April 9 50 min
3.3.2 Interview Guide
A crucial stage in semi-structured interviews for data collection is the interview guide, which functions as an outline of planned topics and questions to be addressed (Adams, 2015). This thesis integrated two major components in its interview guide: 1) recommendations from Adams (2015) for the overall design and flow of interviews and 2) questions rooted in Theoretical Background so as to understand and answer: “How does trust in data influence SSBI adoption by business users?”.
1. Recommendations from Adams (2015) for the overall design and flow of interviews:
- Find a balance between the planned questions and the allotted time;
- Mix close-ended and open-ended questions to create a natural flow of probing and exploration;
- Be aware of and sensitive to social stigma when formulating questions to minimise defensiveness and receiving “the correct answer”;
- Keep the interview guide flexible and modify when necessary.
2. Questions rooted in Theoretical Background so as to understand and answer: “How does trust in data influence SSBI adoption by business users?”
- User Adoption
Questions based on TAM to understand PU and PEOU.
- Trust
Questions to explore benevolence based trust and competence based trust.
- Trust in Data
Questions to understand who or what is being trusted, context of trust, and the degree of trustworthiness.
- Connecting the Parts
Questions to connect the three parts above to answer the thesis question.
Once a draft of the interview guide was ready, it was revised and improved upon in three stages.
First, the questions were shared with a senior consultant with more than 10 years of work experience in the BI/SSBI field. The goal was to make sure that the questions were relevant in the SSBI context. Secondly, a mock interview was conducted to check the clarity of questions, the flow of dialogue, the total amount of time necessary, and most importantly, if responses to questions would lead to deeper understandings of the research question. The participant of this mock interview did not work at Company X but had been a business user of one of the SSBI tools implemented at Company X. Lastly, the majority of interview questions were tested on the contact person at Company X in a quick run-through. Thoughts and feedback were collected and integrated to improve the draft interview guide. Appendix 3 shows the final interview guide used.
During the course of interviews with participants from Company X, the interview guide was followed in principle with allowances for adjustments in time, order of questions, and the exact wording of questions. In addition, interviews were conducted via the video conferencing tool preferred by Company X to ensure compliance with corporate policies and maximise familiarity of meeting environment for the participants.
3.4 Techniques for Data Analysis
Analysing data is the heart of qualitative research (Eisenhardt, 2013) and data can exist in many different forms (Blandford, 2013). Interview notes, audio files, photographs, videos, and more recently, interactions on social media, are just some sources of data. While the dominant data source for this thesis was recordings from semi-structured interviews, two minor data sources were also included: field notes from those semi-structured interviews and answers to follow-up questions via email. After having gathered all the relevant data, the first step in data analysis was to transform the interview recordings into text. Transcribing interview recordings made them easier to work with (Blandford, 2013) and three aspects were taken into consideration during this process:
- Direct involvement
Transcription was done by the researcher herself as the very act of transcribing, and making notes at the same time, was a useful step in becoming familiar with the data and getting immersed in it (Blandford, 2013). Any signs of inadequacy during the interview could be picked up and analysed to improve the quality and experience of the next meeting. Should some questions feel awkward, the Interview Guide could be modified to address those issues.
- Timeliness
Interview recordings were transcribed as soon as possible while details were still fresh in mind both for the research and the participant. If there were incomplete answers or if further clarifications would be needed on certain points, those could be included in the post- interview thank-you note (Adams, 2015). This timeliness would also enable inadequacies to be taken care of and necessary modifications in the Interview Guide to be carried out before the next interview.
- Selective transcription
The purpose of this thesis is to explore how trust in data influences SSBI adoption by business users, phatic utterances, pauses, intonations, and the like were therefore not deemed necessary for this understanding. In other words, interview transcriptions only captured words relevant to the research question (Blandford, 2013).
Software tools such as ATLASti and NVivo are readily available to assist researchers in data management and analysis. Regardless of the tool, its purpose is to create mediating representations between the researcher and the data, allowing the researcher to look at data in news ways (Blandford, 2013). Inherently more interpretive, coding in qualitative research is the analytical process of organising raw data into themes in order to better understand data and discover patterns in it (Baralt, 2012). For this thesis, interview recordings were transcribed manually using Microsoft Word first and then transferred to Microsoft Excel line by line for analysis. Thereafter, transcriptions were coded based on the recommended steps of thematic analysis from Blandford (2013):
- Familiarising with data through multiple cycles of reading and examining the transcriptions;
- Generating initial codes by collating data. Each code represents a group of similar items, ideas, or phenomena;
- Searching for themes and gathering codes into candidate themes for further analysis;
- Reviewing themes and creating a thematic map of the analysis;
- Refining themes and the overall narrative iteratively;
- Producing a report with analysis and thematic map in relation to the research question.
Before being able to produce the final report, qualitative coding needs to reach saturation, which is the final stage where all concepts are well established and defined, and further analysis does not lead to any new codes or patterns (Baralt, 2012). At the end of the data analysis process, an Excel file containing coded interview transcriptions emerged. Figure 8 shows a screenshot of how answers were coded with nodes.
Figure 8: Example of coding using Excel
3.5 Quality and Validity
Unlike quantitative research, which has widely agreed criteria for quality such as internal validity (experiments conducted without confounding variables) and external validity (generalisability of results), there are some challenges to reaching consensus on criteria for assessing and ensuring quality for qualitative research (Blandford, 2013). As qualitative research is primarily concerned with textual analysis (Myers, 1997) and while software provides powerful assistance, the human mind is still what drives coding decisions and the analysis (Baralt, 2012). The distinction between data gathering and data analysis can be problematic for some qualitative researchers, Myers (1997) brings to attention that the researcher’s presuppositions can affect the gathering of data, the analysis can affect the data, and the data can affect the analysis.
Being aware of the challenges mentioned above, the seven principles for interpretive field research put forth by Klein & Myers (1999) were employed to ensure quality for this thesis.
This set of principles was selected as they were developed to guide and evaluate interpretive research in IS, which matched the research method and context of this thesis. Each of the seven principles was defined (Klein & Myers, 1999) and reflected upon:
1. The Fundamental Principle of the Hermeneutic Circle
- This principle suggests that all human understanding is achieved by iterating between considering the interdependent meaning of parts and the whole that they form. This principle of human understanding is fundamental to all the other principles.
- In order to answer, “How does trust in data influence SSBI adoption by business users”, the research question was first unpacked into smaller and independent components of BI/SSBI, user adoption, and trust/trust in data. After reaching deeper understandings of each component with reference to academic research and business articles, the parts were put together to explore the relationship amongst them. This reiterative process was in turn supplemented by data gathered from semi-structured interviews.
2. The Principle of Contextualisation
- Requires critical reflection of the social and historical background of the research setting, so that the intended audience can see how the current situation under investigation emerged.
- With SSBI being relevant only in an organisational setting, organisation was the context for this thesis. Detailed descriptions of the current situation at Company X in relation to SSBI along with criteria for Delimitation strived to further contextualise this research so as to provide a clear narrative for the intended audience.
3. The Principle of Interaction Between the Researcher and the Subjects
- Requires critical reflection on how the research materials (or “data”) were socially constructed through the interaction between the researchers and participants.
- With respect to the ongoing pandemic and everyone’s wellbeing, all interactions between the researcher and participants were carried out online. While some communication was handled through email, all the individual interviews were conducted via video conferencing.
The pros and cons of virtual interviews were examined and measures in the form of pre- interview guideline were taken to help tackle the drawbacks of this medium. During the interview process, it was not evident that participants were less willing to share due to video conferencing. In fact, participants were comfortable with the virtual setting as they had had one year to adjust to the world Covid-19 had propelled everyone in. All interactions between the researcher and the participants were kept on a professional level as demonstrated by Appendix 1, 2, and 3.
4. The Principle of Abstraction and Generalisation
- Requires relating the idiographic details revealed by the data interpretation through the application of principles one and two to theoretical., general concepts that describe the nature of human understanding and social action.
- This thesis started with an intuition that trust in data could be an important factor for SSBI adoption by business users but did not define this relationship at the onset. Instead, data analysis based on semi-structured interviews were used to guide this definition. Participants were asked a series of questions, as outlined in Appendix 3, which were centred around the three components of the research question. The answers to those questions, both from the meeting and from the follow-up emails, were coded and analysed to reach insight for answering the research question.
5. The Principle of Dialogical Reasoning
- Requires sensitivity to possible contradictions between the theoretical preconceptions guiding the research design and actual findings (“the story which the data tell”) with subsequent cycles of revision.
- Previous academic research on a topic similar to the research question in this thesis could not be found. There was some research that explored user uncertainties and challenges without zeroing in on trust in data and other business articles that talked about trust in data but not specifically in the SSBI context. In other words, there were no established theoretical preconceptions available. This thesis was therefore explorative and tried to gain insight through the iterative process of data analysis. The techniques recommended by Blandford (2013) were applied during data analysis to fully represent “the story which data tell”.
