How does Business intelligence 
& analytics create business value 
and facilitate ambidexterity?  

A case study in the Swedish online retailing industry 
Master’s thesis in Management and Economics of Innovation 
 

 
 

OSCAR THORNANDER 
ADAM WALMAN 
 
 
 
 
 
DEPARTMENT OF TECHNOLOGY MANAGEMENT AND ECONOMICS  
DIVISION OF ENTREPRENEURSHIP AND STRATEGY 

CHALMERS UNIVERSITY OF TECHNOLOGY 
Gothenburg, Sweden 2021 
www.chalmers.se 
Report No. E2021:107 





 
 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 



 
 

 
REPORT NO. E2021:107 

 
 
 
 
 
 
 
 

 

How does Business intelligence & 
analytics create business value and 

facilitate ambidexterity?  
 

A case study in the Swedish online retailing industry 

 
OSCAR THORNANDER 

ADAM WALMAN 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Department of Technology Management and Economics 
Division of Entrepreneurship and Strategy 

CHALMERS UNIVERSITY OF TECHNOLOGY 
Gothenburg, Sweden 2021



 
 

How does Business intelligence & analytics create business value and facilitate 
ambidexterity?  
A case study in the Swedish online retailing industry 
OSCAR THORNANDER 
ADAM WALMAN  
 
 
© Oscar Thornander, 2021.  
© Adam Walman, 2021.  
 
 
Report no. E2021:107 
Department of Technology Management and Economics  
Chalmers University of Technology 
SE-412 96 Göteborg 
Sweden 
Telephone + 46 (0)31-772 1000  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Gothenburg, Sweden 2021 





 
 

How does Business intelligence & analytics create business value and facilitate 
ambidexterity? 
A case study in the Swedish online retailing industry 
 
OSCAR THORNANDER 
ADAM WALMAN  
 
Department of Technology Management and Economics  
Chalmers University of Technology 
 
 
 

Abstract 
Several tools and methods have arisen in the last decade to leverage data, where 
business intelligence and analytics have become commonly used by companies 
and significantly in the online retailing industry. The online retailing industry stands 
out with its immense growth and access to data on customers, competitors, and 
products, making it possible to find and create value from data with a relatively low 
initial investment. Business analytics and intelligence is a fast-evolving field in 
which earlier studies on value appropriation have primarily been conducted in 
Spain and the US, outside the online retailing segment. Thus, the Swedish online 
retailing market is previously unexplored and could be especially interesting to 
study, as the country is viewed as an IT frontier. 
 
A business is bound to include activities to create value for the present and the 
future by innovation and continuous improvement to survive, and these processes 
can be reinforced with business intelligence and analytics. However, how value is 
appropriated through business intelligence and analytics is sparsely covered and 
needs further examination. This paper aims to expand the knowledge on how 
value is appropriated through BI&A. Further, the paper aims to investigate whether 
and how BI&A can assist continuous improvement and innovation by supporting 
ambidexterity and its exploitation and exploration processes by studying the online 
retail market in Sweden. 
 
A qualitative methodology was adopted in which 14 different online retailing 
companies operating in Sweden have been interviewed, along with two experts in 
online retailing. The study identified improvements in operational efficiency, 
increasing sales, and competitiveness of the business as three main areas of 
value appropriations gained from business intelligence and analytics. The paper 
presents vast examples of how value is created and appropriated through 
business intelligence and analytics in these three main areas and are highlighted 
in the results. Our findings and discussion present how value is created, not only 
in the online retailers’ specific functions but gives a holistic view of how business 
intelligence and analytics support value creation throughout the organization. 
Further, our findings suggest that business analytics and intelligence improve the 
ability to achieve ambidexterity by supporting explorative and exploitive activities 
with, for example, methods and tools such as designing and aligning KPIs, and 
A/B testing, influencing the mediating factors of interconnectedness and cross-
functional interfaces.  
 
 
 
Keywords: Business intelligence and analytics, online retailing, ambidexterity, competitive 
advantage, dynamic capabilities 



 
 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

  



Acknowledgments 
 
The authors would like to offer our appreciation and gratitude to our supervisor at 
Chalmers University of Technology, Maria Kandaurova, for her guidance and 
engagement throughout the study. 
 
We would also like to thank the involved online retailers and respective interviewees 
who participated in the study. By providing their time and the openness to share their 
experiences, they have made this study possible. 
 
Lastly, we would like to thank Chalmers University of Technology, including its 
professors, staff, and students, for the support over the past five years. 
 
 
 
 
 
______________________________                 ______________________________                       
Oscar Thornander                                                Adam Walman 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 



 
 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 





 

i 
 

Table of Contents 
 

 

List of Figures .............................................................................................................. iii 

List of Tables ............................................................................................................... iii 

1 Introduction ............................................................................................................... 1 

1.1 Purpose and Research Questions ......................................................................... 3 
1.2 Delimitations ........................................................................................................ 3 
1.3 Outline of the Report ........................................................................................... 3 

2 Literature review ...................................................................................................... 4 

2.1 Value .................................................................................................................... 4 
2.2 Creating and sustaining a competitive advantage ................................................ 4 

2.2.1 Dynamic capabilities ..................................................................................... 5 
2.2.2 Ambidexterity ............................................................................................... 6 

2.3 Business intelligence and analytics ...................................................................... 8 
2.3.1 Data management and analytics .................................................................... 9 
2.3.2 Big data ....................................................................................................... 10 
2.3.3 Value creation process of business analytics .............................................. 10 

2.4 Online retail industry ......................................................................................... 11 
2.4.1 Business model of an online retailer ........................................................... 11 

2.5 BI&A and ambidexterity in the online retail industry ....................................... 14 

3 Methodology ............................................................................................................ 16 

3.1 Research Strategy and Approach ....................................................................... 16 
3.2 Data Collection .................................................................................................. 17 
3.3 Analysis of Results ............................................................................................ 21 
3.4 Research Quality ................................................................................................ 24 
3.5 Research Ethics .................................................................................................. 25 

4 Results ...................................................................................................................... 26 

4.1 Operational efficiency ........................................................................................ 26 
4.1.1 Digital marketing efficiency ....................................................................... 26 
4.1.2 Administration and customer service efficiency ......................................... 29 
4.1.3 Warehousing and logistics efficiency ......................................................... 30 
4.1.4 Decrease returns .......................................................................................... 31 
4.1.5 Optimization of purchases and stock .......................................................... 32 

4.2 Increase sales ..................................................................................................... 33 
4.2.1 Average order value .................................................................................... 33 



 

ii 
 

4.2.2 Conversion rate ........................................................................................... 36 
4.2.3 Customer retention ...................................................................................... 38 
4.2.4 Optimize traffic to the site .......................................................................... 39 
4.2.5 In-house marketing sales ............................................................................. 40 

4.3 Competitiveness of business .............................................................................. 41 
4.3.1 Value segmentation of customers ............................................................... 42 
4.3.2 Branding, positioning, and assortment ........................................................ 43 
4.3.3 Pricing strategy ........................................................................................... 44 

4.4 Online retailer ambidexterity ............................................................................. 45 
4.4.1 Organizational structure of BI&A and explorative and exploitative activities 
in online retailing ................................................................................................. 45 
4.4.2 BI&A supporting exploitation .................................................................... 48 
4.4.3 BI&A supporting exploration ..................................................................... 50 

5 Discussion ................................................................................................................. 52 

5.1 Value creation of BI&A by B2C online retailers in Sweden ............................. 52 
5.2 BI&A influence on exploitative and explorative activities and ambidexterity .. 55 

6 Conclusion ............................................................................................................... 59 

6.1 Theoretical implications ..................................................................................... 59 
6.2 Practical implications ......................................................................................... 60 
6.3 Limitations and Future Research ....................................................................... 62 

7 Sources ..................................................................................................................... 64 

A Appendix ................................................................................................................... I 

A.1 Interview guide online retailers ............................................................................ I 
A.2 Interview guide consultants ............................................................................... II 

 

 
 
 
 
 
 
 
 
 
 
 
 



 

iii 
 

 
 

List of Figures 
 
Figure 2.1. The level of sophistication of analytics in relation to potential competitive 
advantage. ...................................................................................................................... 9 
Figure 2.2. Process model for value creation ............................................................... 11 
Figure 3.1. Thematic coding of interviews .................................................................. 24 
 

List of Tables 
Table 2.1. Broad business models of online retailer B2C ............................................ 12 
Table 3.1. Interviews held ............................................................................................ 20 

 

 

 

 

 
 
 
 
 
 
 



 

iv 
 

 

 

 

 

 

 

 

 

 
 



 

1 
 

1 

Introduction 
 
 
The B2C online retailing market in Sweden is experiencing substantial growth (Statista, 
2020), and many traditional retailers are refining their business model to include the 
online segment with an omnichannel structure. Thus, the growth is coupled with 
increased competition domestically and by new market entry from platform-based 
marketplace giants, such as Amazon (Piotrowicz & Cuthbertson, 2014). To survive in 
a market that experiences substantial competition, firms need to innovate in their 
existing business areas and have the ability to catch new market opportunities early 
(Dereli, 2015). Grewal et al. (2017) describe the importance for retailers of using big 
data and analytics to increase their competitiveness to handle the trends and increased 
competition. 
 
Using data has had an increasing interest in almost all industries for the last decades for 
their potential of extracting insights into actions to gain a competitive advantage 
(Conboy et al., 2020). This phenomenon is largely coupled with the fact that more data 
is available to businesses than ever before. Data availability now includes data from 
competitors, customers, social networks, user-generated content, machines, products, 
and internal business (Mortenson et al., 2015). These data sources are often too 
complex, too large, and generated at such a velocity that traditional data processing 
techniques are insufficient (Delen & Zolbanin, 2018). Following these issues, new 
methodologies and processing techniques have been introduced, giving rise to a new 
era in business decisions making, as per Mortenson et al. (2015), the business analytics 
period. The most widely adopted definition of analytics (Hindle et al., 2020) arises from 
Davenport and Harris (2007, p7) that states: “By analytics we mean the extensive use 
of data, statistical and quantitative analysis, explanatory and predictive models and 
fact-based management to drive decisions and actions.” Further, data analytics is a 
broad term applied to the analysis of any data, while business analytics is considered 
the general term for any data analytics in business problems (Duan & Xiong, 2015). In 
this report, business intelligence and analytics (BI&A) will be used as a combined term 
and is defined according to Lim et al. (2013) as technologies, systems, practices, and 
applications to analyze critical business data to gain new insights of the business and 
markets. BI&A is a transdisciplinary field where operations research, machine learning, 
and information systems are particularly relevant (Hindle et al., 2020). However, the 
academic focus on the field has mainly been on technical aspects, and there is a lack of 
business and operations research in the area (Delen & Zolbanin, 2018; Conboy et al., 
2020).  



 
 

1. Introduction 

2 
 

 
Further, dynamic capabilities theory has grown to be one of the most dominant theories 
in explaining what enables a firm to adapt to changing environments and sustain a 
competitive advantage (Schilke, 2014). Dynamic capabilities suggest that the ability of 
a firm to reconfigure its resources and capabilities explains long-term competitive 
advantage. Building on the dynamic capability perspective, also seeking to resolve what 
explains a firm’s ability to remain competitive is organizational ambidexterity. 
Organizational ambidexterity refers to a firm’s ability to both explore and exploit 
business opportunities (O’Reilly & Tushman, 2008), which is often described as the 
ability of a firm to pursue both efficiency and innovation. The organizational theory 
assumes that the capabilities for exploring and exploiting are widely different, which 
creates challenges and tensions in organizations in sustaining both (Raisch et al., 2009). 
Many studies have focused on how organizations should structure themselves to sustain 
exploration and exploitation and the linking of these activities (Papachroni et al., 2015). 
 
The literature addresses several interesting topics for further analysis in areas of 
business intelligence and analytics, online retailing, and organizational change in 
uncertain environments. For instance, Akter and Wamba (2016) highlight a need to 
explore how firms extract value from big data at their disposal and how the use of big 
data varies in adoption and implementation between different firm types in the e-
commerce industry. Since 2016 the area of extracting business value from BI&A has 
been extensively examined, and “competitive advantage” and “customers” are the two 
most frequently covered topics in the intersection between the fields of data science and 
business analytics (Hindle et al., 2020). However, as Hindle et al. (2020) further state, 
the use of big data and BI&A is quickly evolving and improving, which gives the need 
for further exploration of how the field is developing. Further, Benitez et al. (2018) call 
for additional research on how firms leverage information technology (IT), such as 
business intelligence and analytics, to create business value. Additionally, they ask for 
further investigations on the effect IT has on firms’ exploration and exploitation 
capabilities in other countries than Spain and the US, where former studies previously 
have been conducted.  
 
This study aims to address how value is created through the use of BI&A in the online 
retail market in Sweden. By investigating how online retailers in Sweden use BI&A to 
create value, we hope to provide rich new insights on the topic in a previously 
unexplored market. Further, as the online retailing market is mature in Sweden, and the 
country is viewed as a fore-frontier in IT, it could be especially interesting to study this 
market. Additionally, by addressing if, and if so, how business analytics are used to 
facilitate exploratory and exploitative activities, we hope to contribute to academia on 
how business analytics can assist organizations in creating and sustaining a competitive 
advantage. 
 



 
 

1. Introduction 
 

3 
 

1.1 Purpose and Research Questions  

The study aims to provide insights to the strategic management literature on the 
phenomenon of how BI&A are used to create value in the context of B2C online retail 
companies operating in Sweden. Further, the study aims to describe how BI&A 
supports online retailers’ processes of exploring new opportunities and the exploitation 
of their existing business. Particularly this approach will highlight if and how BI&A 
are used to create and sustain both a short and long-term competitive advantage by 
online retailers. 
 
The questions the study will seek to answer are the following: 
 

1. How are business intelligence and analytics creating value in the B2C online 
retailing industry in Sweden? 

2. How are business intelligence and analytics supporting online retailers’ 
exploitative and explorative activities, and whether and how it facilitates their 
ability to be ambidextrous? 
 

1.2 Delimitations 

The study looks at B2C online retailers operating in Sweden. Online retailers included 
in this study are those having more than 50% of total sales through online sales 
channels. Further, the selection is delimited to companies primarily in retailing, mainly 
selling products manufactured by others. The study does not intend to answer if B2C 
online retailers in Sweden can be considered ambidextrous organizations but to clarify 
how BI&A tools facilitate ambidextrous processes. An assumption is made that the 
organizations selected are ambidextrous to some degree. Further, the study does not 
aim to make recommendations on how online retailers should conduct their BI&A 
processes to create value but to understand how they are used to create value today.  
 

1.3 Outline of the Report 

The report is structured by introducing the phenomenon and research field with its 
delimitations and research questions in sections 1.1 to 1.2. Chapter two includes the 
literature review describing the underlying theory and literature on the subject of 
dynamic capabilities, ambidexterity, value in online retailing, business analytics, and 
business intelligence. The chosen method, analysis, and sample of the study and its 
implications for the results are discussed in chapter three. Chapter four describes the 
study results followed by chapters five and six, including discussion and conclusions 
answering the research questions and fulfilling the aim of the study.  



 

4 
 

2 

Literature review 
 
 
The following sections 2.1 to 2.5 depict the underlying theory on the current research 
of value, competitive advantage, ambidexterity, dynamic capabilities, business 
intelligence, and business analytics related to this study’s area.  
 

2.1 Value 

What organizations perceive as value depends on the firm’s strategic goals (Günther et 
al., 2017). However, according to Grant (2016), a firm’s sole purpose is to pursue profit 
over the firm’s lifetime, which will provide value maximization for all shareholders. In 
this paper, value will therefore refer to economic value for the company, which can 
represent itself as an organization’s increase in profit, business growth, and competitive 
advantage (Davenport & Harris, 2007).  
 

2.2 Creating and sustaining a competitive advantage  

Research has shown that less than 0,1% of companies founded in the US are likely to 
survive to the age of 40 (O’Reilly & Tushman, 2011). Despite their size, vast stable 
financials and human resources, well-established large corporations can only expect to 
live between 6 to 15 years (Ormerod, 2005).  
 
As competition is increasing in almost all industries, the primary goal of a firm’s 
strategy is to build a competitive advantage (Grant, 2016). A competitive advantage 
can be defined as a firm’s capacity to create value or simply as a firm’s potential to 
create higher profits than competitors (Grant, 2016). Gaining a competitive advantage 
can arise from external sources such as acting on changing customer demands, input 
prices or technological change, or internally from a firm’s capabilities to improve or 
innovate (Grant, 2016). Schumpeter (1942) describes competition as a dynamic process 
in which the industry is in constant change. The higher level of competition, the less 
stable industry structures become, and thus the less stable a competitive advantage 
becomes. Grant (2016) elaborates and describes a competitive advantage as a 
disequilibrium, and it is created by change and sets in motion the competitive process 
of discontinuity. So, the higher level of competition, the more extensive degree of 
change and innovation is needed to sustain a competitive advantage. Managing a 
competitive advantage involves two major dimensions: positioning and improving for 



 
 

2. Literature review 
 

5 
 

the present and adapting to the future (Grant, 2016). Jim March (1991, p. 105) expresses 
it as “the heart of an enterprise’s long-term survival was to engage in sufficient 
exploitation to ensure its current viability and, at the same time, to engage in sufficient 
exploration to ensure its future success.” 

2.2.1 Dynamic capabilities  

Dynamic capabilities theory has grown to be one of the most dominant theories in 
explaining what enables a firm to adapt to changing environments (Schilke, 2014). 
Dynamic capabilities enable firms to sense and seize emerging business opportunities 
and transform their operations accordingly (Teece, 2007). Dynamic capabilities suggest 
that the ability of a firm to reconfigure its resources and capabilities in ways that are 
valuable to the customer but difficult to imitate explains long-term competitive 
advantage (O’Reilly & Tushman, 2011). Dynamic capabilities intend to change a 
product, process, the market, or the scale served by a firm (Miles, 2018). They should 
not be confused with a firm’s organizational capabilities that refer to how an 
organization earns a living (Miles, 2018). Even though there are considerable variations 
in definition, there is increasing convergence on the idea that dynamic capabilities are 
a set of identifiable and specific routines centered around coordination, learning, and 
transformation (O’Reilly & Tushman, 2011). These routines enable the renewal of 
organizational capabilities and increased flexibility in response to market change 
(Pezeshkan et al., 2016).  
 
To make the process of analyzing dynamic capabilities manageable, Teece (2007) 
disaggregates the concept into three activities: sensing new opportunities and threats, 
seizing opportunities, and enhancing, combining, protecting, and, when necessary, 
reconfiguring the firm’s intangible and tangible assets. Teece (2007) explains that 
sensing involves analytics systems of scanning, search, and exploration activities in 
customer needs, latent demands, technologies, markets, and the evolution of industries 
and the responses from suppliers and competitors. He further explains that overcoming 
a narrow search horizon can become problematic for firms tied to problem-solving 
competencies. Seizing is equivalent to addressing the new sensed opportunity through 
new products, processes, or services. Lastly, reconfiguring involves maintaining and 
improving tangible and intangible assets.  
 
The theory has met some criticism regarding its definition. It is often described in 
ambiguous ways, and other concepts such as absorptive capacity, change management, 
organizational learning, and strategic fit address similar problems (Miles, 2018). 
Further, the theory has been critiqued in its lack of measurability and that dynamic 
capabilities are often assumed without specifying their exact components (Miles, 2018). 
 



 
 

2. Literature review 

6 
 

2.2.2 Ambidexterity  

Another vastly influential theory, building on the dynamic capability perspective, also 
seeking to resolve what explains a firm’s ability to remain competitive is organizational 
ambidexterity - a firm’s ability to both explore and exploit market opportunities 
(O’Reilly & Tushman, 2008). Exploratory and exploitative activities are often 
described as whether a firm can simultaneously pursue efficiency and innovation. 
Similar to the theory on dynamic capabilities, criticism for the ambidexterity theory 
revolves around ambiguity regarding its definitions and measurability. According to 
O’Reilly and Tushman (2013) the generic use of organizational ambidexterity refers to 
the ability of firms to do two things at once, e.g., competing in different markets or with 
different technologies, which leaves room for interpretation. However, O’Reilly and 
Tushman (2013) stress that ambidexterity is not only about whether a firm can pursue 
both exploration and exploitation, but more importantly if they can develop the 
capabilities needed to compete in new markets and technologies to survive in changing 
market conditions. Turner et al. (2013) further discuss that the distinction between 
exploitation and exploration is not always clear-cut, and definitions are vague. 
However, in this study, exploratory and exploitative activities are defined according to 
Marchs’ (1991) article, in which the literature on ambidexterity originated (O’Reilly & 
Tushman, 2008; Brix, 2019). March (1991) defines exploration to include terms such 
as search, variation, risk-taking, experimentation, play, flexibility, discovery and 
innovation, and exploitation to include refinement, choice, production, efficiency, 
selection, implementation, and execution. O’Reilly and Tushman (2011) argue that the 
ability of a firm to be ambidextrous is at the very core of dynamic capabilities, which 
gives managers two critical tasks. Firstly, they must accurately sense changes in the 
environment, such as technology, competition, customers, and regulation. Secondly, 
have the capacity to act on the opportunities and threats by reconfiguring both tangible 
and intangible assets (O’Reilly & Tushman, 2011).  
 
While both exploratory and exploitative processes are viewed as necessary for a firm 
to survive in a changing or dynamic environment, no consensus has been reached on 
how these processes should be balanced (Brix, 2019). Many studies have focused on 
how organizations should structure themselves to sustain exploration and exploitation 
activities and how an organization should link these activities (Papachroni et al., 2015), 
which has led to two different views on how exploration and exploitation should be 
structured (Brix, 2019). The differentiation view assumes that the capabilities for 
exploring and exploiting are widely different, and therefore, exploitation and 
exploration must occur across units or organizations, defined as structural 
ambidexterity (O’Reilly & Tushman, 2013). On the opposing side, the integration view 
argues that the processes can exist within the same unit, with so-called contextual 
ambidexterity (Gibson & Birkinshaw, 2004). According to the integration view, both 
exploratory and exploitative activities can coexist and may be performed without trade-
offs (Turner et al., 2013).  
 



 
 

2. Literature review 
 

7 
 

Further, it is debated whether maintaining explorative and exploitative is always 
beneficial for firm performance. According to Luger et al. (2018), the long-term 
benefits of balancing exploration and exploitation are highly dependent on the 
company’s environment. They state that in environments of incremental change, 
maintaining a static balancing of exploration and exploitation might lead to superior 
performance, as incremental change often runs over long cycles. However, statically 
maintaining and balancing exploratory and exploitative activities will reinforce inertia, 
limiting a firm’s ability to adapt its balance between exploratory and exploitative 
activities in contexts of discontinuous change (Luger et al., 2018). This inertia will hurt 
a firm’s ability to change and thereby their long-term performance (Luger et al., 2018). 
From this fact, it can be derived that companies need to have the ability to perform 
exploratory and exploitative activities and have the ability to shift the balance of each 
when necessary.  
 
Even though there are opposing views on how exploration and exploitation should be 
managed, it is agreed that both the abilities to perform exploration and exploitation are 
positively associated with performance and increases the likelihood of sustaining a 
long-term competitive advantage, especially under changing market conditions 
(O’Reilly & Tushman, 2013). Further, efforts have been made to investigate how firms 
can positively influence ambidexterity. Jansen et al. (2009) identify a set of 
organizational integration mechanisms and investigate the mediating effect they have 
on firms’ ambidextrous ability. Their logic is that organizational integration facilitates 
value by linking knowledge sources and providing opportunities to use shared resources 
and gain synergies between units, which they hypothesize positively influences 
ambidexterity. The two mechanisms of organizational integration they analyze are 
cross-functional interfaces and connectedness. Cross-functional interfaces are defined 
as the use of personnel, task forces, or teams from different divisions with diverse 
expertise to enable knowledge sharing for exploitative and explorative processes 
(Jansen et al., 2009). The interfaces facilitate the cross-functional members to reach a 
common frame of reference and build understanding and agreement. Connectedness 
regards the patterns of a firm’s social network in terms of density and provides the base 
where organizational members can share experience, knowledge, and transfer and 
integrate new ideas (Jansen et al., 2009). Jansen et al. (2009) find that cross-functional 
interfaces mediate exploratory and exploitative processes as it provides linkages 
between units and on the hierarchical level and eases communication. Further, they find 
that connectedness has a direct contribution to achieving ambidexterity. Turner et al. 
(2013) also highlight the importance of similar mechanisms to accommodate formal 
and informal coordination and processes for creating social relationships and 
coordination for achieving ambidexterity. Further, Jansen et al. (2006) conducted a 
study to test how formalization affects exploitative and explorative processes. 
Formalization is the amount that rules, procedures, instructions, and communications 
are formalized or written down (Khandwalla, 1977). Formalization was found to have 
a positive relationship with exploitive outcomes, as existing knowledge and skills 
accelerate the diffusion of best practices within units. Additionally, it did not negatively 



 
 

2. Literature review 

8 
 

influence explorative activities, even though March (1991) stated that the reliance on 
rules and procedures hampers experimentation and ad hoc problem-solving efforts. 
 

2.3 Business intelligence and analytics  

The terms referred to as business intelligence (BI) and business analytics (BA) and their 
definitions are not agreed upon and are often used as subsets or special cases of each 
other (Mashingaidze & Backhouse, 2017). The most widely adopted definition of 
analytics (Hindle et al., 2020) arises from Davenport and Harris (2007, p7) that states: 
“By analytics we mean the extensive use of data, statistical and quantitative analysis, 
explanatory and predictive models, and fact-based management to drive decisions and 
actions.” Stubbs (2011) describes business analytics to expand analytics by using 
descriptive, predictive, and prescriptive analytics to create valuable information to 
improve business performance. These perspectives align with Mashingaidze and 
Backhouse’s (2017) definition, that describe business intelligence as a set of tools and 
techniques to use data for decision-making, and business analytics to be a more 
advanced form of business intelligence. In this report, we will use business intelligence 
and analytics (BI&A) as a combined term and define it according to Lim et al. (2013) 
as technologies, systems, practices, and applications to analyze critical business data to 
gain new insights about business and markets.  
 
BI&A, together with big data adoption, has recently in academic literature been 
discussed to create opportunities for companies to achieve growth and to improve their 
overall operations (Ajah & Nweke, 2019). Companies invest in business analytics to 
achieve a competitive advantage over their competitors (Seddon et al., 2017), and 
improving business processes through BI&A has great potential to create value and 
achieve a competitive advantage (Davenport & Harris, 2017). However, to be 
competitive in general, an organization must have attributes that they perform better 
than anyone else in their industry, for which BI&A can help improve performance 
(Davenport & Harris, 2017). Certain industries which generate or have appropriate data 
available are more suitable to leverage value from using BI&A, which has led some 
industries to have adopted BI&A to a greater extent (Davenport & Harris, 2017). In 
some industries, so-called analytical competitors have established themselves. 
Analytical competitors are organizations that use analytics extensively and 
systematically to outperform the competition (Davenport & Harris, 2017). Examples of 
these are Nike and Tesla in the consumer product industry, or Amazon, Tesco, and 
Walmart in the retail industry. However, Davenport and Harris (2017) state that the 
likelihood of obtaining a competitive advantage from a specific analytic resource 
increases with the level of sophistication (figure 2.1), seemingly unrelated to industry-
specific progress.  
 
 
 



 
 

2. Literature review 
 

9 
 

 

 
Figure 2.1. The level of sophistication of analytics in relation to potential competitive 
advantage. Based on Davenport and Harris (2017). 
 

2.3.1 Data management and analytics 

Data acquisition and management is the first step of being able to generate insights 
from data. The acquired data can either be fully unstructured, structured or between 
these extremes, so-called semi-structured (Gandomi & Haider, 2015). One of the 
differentiating factors of big data to data is the variety of data types, which commonly 
can handle unstructured data and increase data management complexity. Data 
management involves collecting, storing, processing, and transferring data (Duan & 
Xiong, 2015, which can involve internal, external, or open sources (Hindle et al., 2020). 
Data storage is commonly realized through data warehousing, which consolidates 
information of different kinds to be analyzed (Ajah & Nweke, 2019). Transfer and 
process data include the exchange of data within networks (Duan & Xiong, 2015). 
 
Data analytics can be used to draw insights from data when it has been gathered and 
stored correctly. There are three methods of analyzing data: descriptive analytics, 



 
 

2. Literature review 

10 
 

predictive analytics, and prescriptive analytics (Duan & Xiong, 2015). The methods 
can be seen to build on one another where descriptive analytics is the first stage of 
analysis. The primary goal of using descriptive analysis  
is to understand the data and its indicators of underlying success or failure (Duan & 
Xiong, 2015). The second stage of predictive analysis aims to predict the future with 
past data, using statistical models, assuming that what happened in the past will happen 
again in the future. Prescriptive analytics is the last stage of analysis involving heuristic 
search, mathematical programming, and simulation to find the most advantageous 
actions or decisions to be taken ahead (Duan & Xiong, 2015). The last stage of analytics 
tries to support decision-making by finding the optimal action using different 
parameters in mathematical modeling. Davenport and Harris (2017) define descriptive, 
predictive, and prescriptive analytics in similar ways. They correlate descriptive 
analytics with business intelligence, as reportings of historical and current data, 
predictive analytics as quantitative techniques that use past data to predict the future. 
Lastly, prescriptive analytics is assigned to methods and techniques that specify optimal 
behavior and action, such as recommendation engines (Davenport & Harris, 2017).  
 

2.3.2 Big data  

Big data has in recent years emerged as a concept used in academia and in companies 
using analytics to improve their business. Gandomi and Haider (2015) define big data 
similarly to Laney (2001) by three V’s: volume, variety, and velocity. Volume refers to 
the magnitude of data, variety to structural heterogeneity of the data, and velocity to 
how much data is supposed to be analyzed up to a certain speed (Gandomi & Haider, 
2015). Some sources include additional V’s to the definition of big data with veracity, 
variability, and value (Gandomi & Haider, 2015). Data needs to fulfill all three V’s to 
be classified as big data, but there is no specified numerical threshold value for 
classification. Limits of data classification as” big data” instead depend on the firm 
using the data’s size, industry belonging, and location, which also changes over time 
(Gandomi & Haider, 2015). The threshold for classification of big data is situational-
based and evolves over time, making it challenging to define a relevant universal 
threshold.   
 

2.3.3 Value creation process of business analytics 

Seddon et al. (2017) describe how business analytics contributes to business value 
through a process model. The model includes organizational factors to succeed with 
business analytics value creation and a high-level conceptualization of the process (See 
Figure 2.1). Seddon et al. (2017) argue that the value created by business analytics is 
formed by decision-making and actions taken based on insights from the use of business 
analytics. The value created by the action from business analytics can be aimed at better 
using current resources or by changing the organization’s existing resources.  



 
 

2. Literature review 
 

11 
 

 
Figure 2.2. Process model for value creation. Adopted from Seddon et al. (2017) 
 
Hindle et al. (2020) describe that business analytics can create value by using different 
types of analytics and data. Relatively small amounts of data can with analytics support 
better business decisions (Hindle et al., 2020). Examples of big data and analytics value 
include better and faster decision-making, cost reductions, and new types of product 
offerings (Ajah & Nweke, 2019).  
 

2.4 Online retail industry 

Retailers can no longer be described as intermediaries that buy from suppliers and sell 
to customers. Instead, they are orchestrators that serve an ecosystem where value is 
created and delivered to customers and appropriated by the retailers and business 
partners (Sorescu et al., 2011). Value creation can include product development or 
flexible pricing, and value appropriation can include improved inventory management 
or governance structures that increase customer switching costs (Sorescu et al., 2011).  
 

2.4.1 Business model of an online retailer 

Where a firm’s strategy articulates how a firm aims to achieve a competitive advantage 
(Grant, 2016), the business model focuses on how the firm creates and appropriates 
value to achieve a competitive advantage (Sorescu et al., 2011). The business model 
details the structures, activities, and processes that connect the firm’s internal areas such 
as marketing, sales, and finance to external parties such as customers, suppliers, and 
partners (Teece, 2010). There is no standard definition of a business model. However, 
it is agreed that a business model describes a firm’s value proposition (Sorescu et al., 
2011). Sorescu et al. (2011) define a business model as the structures, activities, and 



 
 

2. Literature review 

12 
 

processes that serve as a firm’s organizing logic for value creation (for customers) and 
value appropriation (for itself and its partners). 
 
Three broad B2C business models for selling products to end-consumers online are 
defined in table 2.1. 
 

Table 2.1. Broad business models of online retailer B2C 
 
While each online retailer falls into one of the above-specified categories in table 1, 
there are differences in the business models commonly used within each category. 
Virtual merchants are often divided into virtual merchants that handle the logistics for 
the products or only handle the economic transaction, and whether they sell their own 
branded products or not. One example of a virtual merchant business model is the 
platform strategy adopted by Amazon (Hagiu & Wright, 2015a). Virtual merchants 
typically buy products from wholesalers or manufacturers, store the product, and then 
sell it to consumers; platforms instead typically only handle the transaction letting third-
party sellers handle logistics around the product (Hagiu & Wright, 2015a). In the case 
of Amazon, a combination of a regular virtual merchant business model and a platform 
with third-party sellers has been adopted (Hagiu & Wright, 2015b). Further, does 
Amazon’s choice of business model differentiate itself in that third-party sellers 
compete on the Amazon site (Hagiu & Wright, 2015b). Another emerging trend within 
online B2C e-commerce is subscription models (Chen et al., 2018). The subscription 
model is possible to be adopted by all online retailers regardless of the category 
belonging above. The subscription model is based on the online retailer selling the 
service of delivering products on a recurring basis to customers (Chen et al., 2018). 
 
However, similarities, independent of the business model, can also be found. Firstly, 
retailers primarily sell products manufactured by others. Secondly, they engage and 
interact directly with the end-consumer. According to Sorescu et al. (2011), these two 
rationales can affect the potential for innovation in retail. Due to selling other firms’ 

Name Comment Companies Source 

Virtual merchant 

A retailer that sells products or services solely through 
an online channel. They can either be handling products 
themself or let third-party actors store, handle, and ship 

products. 

Amazon, 
Adlibris, 
CDON 

(aiHello, 
2019) 

Omni-channel 
merchant 

A retailer that possesses both a physical store and an 
online sales channel. 

Intersport, Clas 
Ohlson 

(aiHello, 
2019) 

Manufacturer-
direct merchant 

A manufacturing company that sells directly to 
consumers through their own online sales channel. No 
wholesale or retail involvement when it comes to their 

specific online sales channel. 

Apple, Ralph 
Lauren, Dell 

(aiHello, 
2019) 



 
 

2. Literature review 
 

13 
 

products, a focus on only product assortment is unlikely to sustain a competitive 
advantage since products are likely found elsewhere, which stresses the importance of 
how retailers sell rather than what they sell (Sorescu et al., 2011). The direct interaction 
with the consumer gives importance to the customer interface and how the retailer will 
optimize interaction and strengthen the customer relationship (Sorescu et al., 2011). 
These two facts have led to a moving focus from transactions to enhancing customer 
experience (Sorescu et al., 2011).  
 
Given that the retailing business models now require enhancing customer experience 
beyond traditional functions of moving, procuring, and stocking products, Sorescu et 
al. (2011) propose a few themes that define retailers’ value creation or appropriation. 
The themes of value creation they propose are customer efficiency, customer 
effectiveness, and customer engagement. Where operational efficiency, operational 
effectiveness, and customer lock-in represent ways of value appropriation for a retailer, 
and customer efficiency and customer effectiveness embody the process of creating 
value for the customer. Customer efficiency refers to the process of making customer’s 
access to products as available as possible (Sorescu et al., 2011). A recent trend towards 
increasing customer efficiency has been to implement an omnichannel business model, 
where customers can access products both online and in stores. Customer effectiveness 
is to which degree the retailer can facilitate the consumer’s shopping goals (Sorescu et 
al., 2011). Increasing customer effectiveness can be done by matching assortment with 
demand or creating a user-friendly and frictionless experience. Traditionally increasing 
customer effectiveness has meant increasing the depth of assortment of popular and 
quick-selling products while leaving the demand for niche products unmet. However, 
the internet has decreased consumer’s searching costs, allowing many online retailers 
to focus their business on niche market segments instead and selling hard-to-find 
products to many customers (Sorescu et al., 2011). Customer engagement involves to 
which extent the retailer can evoke emotional involvement beyond the purchase 
(Sorescu et al., 2011). Operational efficiency refers to doing things faster, cheaper, and 
simpler, which can manifest itself in, e.g., better inventory management and keeping 
shelf levels optimal for faster turnaround or through cost reductions by improving the 
store environment (Sorescu et al., 2011). Operational effectiveness refers to doing the 
right things, in contrast to efficiency, which entails doing things right. Operational 
effectiveness means getting the desired results by operating in a maximally effective 
way for reaching the organization’s objectives, such as target market reach or profits 
(Sorescu et al., 2011). Increasing operational effectiveness in the retail industry can 
manifest itself as implementing flexible pricing to extract maximum profits from 
different market segments or matching assortment with demand. Since retailers are not 
bound to a set product portfolio like manufacturers, they have an advantage in high 
flexibility in determining product assortment. Customer lock-in is activities intended to 
lower customers’ incentives to search for other firms and switch after the initial 
purchase (Sorescu et al., 2011). Traditionally in retail, this has been done through 
memberships or extended warranties. Today more retailers are trying to create loyalty 
that reflects enduring customer relationships, e.g., contract or subscription-based 



 
 

2. Literature review 

14 
 

models (Sorescu et al., 2011). Lock-in can also be created by creating an assortment 
that is unique, inimitable, and with a clear value proposition. 

2.5 BI&A and ambidexterity in the online retail industry 

Both academia and online retail companies have recently caught an interest in the use 
and benefit of using big data and business analytics (Akter & Wamba, 2016), much due 
to that research have shown that there is a positive relationship between business 
analytics and firm performance and that it can enable business process change (Torres 
et al., 2018). Akter and Wamba (2016) investigated in a literature review of 121 papers 
the progression of research in the intersection of big data and business value in the e-
commerce industry. Their review shows that research has shown big data analytics to 
create benefits in a range of areas. Big data analytics can create transaction value in the 
form of cost savings, increased productivity and efficiency, and strategic value such as 
in competitive advantage and firm performance, by transforming parts such as 
production, inventory, innovation, and finance (Akter & Wamba, 2016). 
 
Further, they state that the use of big data is skyrocketing in e-commerce due to social 
networking, the internet, mobile telephone, and other kinds of technology that create 
and capture data, which is influencing the e-commerce industry to handle the unique 
nature of big data, that of high volume, variety, and velocity (Akter & Wamba, 2016). 
However, Vidgen et al. (2019) show that while data may be classified as big, it is not 
required to create business value. Organizations can create substantial business value 
from relatively small amounts of data, which may not have been exploited previously, 
giving new insights on customers, processes, and the competitive environment (Vidgen 
et al., 2019). Thereby, this study will not focus on separating big data analytics from 
traditional data analytics but rather focus on the value created by BI&A independently 
of it can be classified as big data according to volume, variety, and velocity. 
 
In the context of ambidexterity, prior research on exploration and exploitation has 
mainly focused on balancing the activities are the trade-offs and problems that arise 
from pursuing both, but it remains unclear how IT technologies facilitate ambidexterity 
(Benitez et al., 2018). In their study, Benitez et al. (2018) identify that IT infrastructure 
provides a foundation for building business experimentation and helps to sense and 
explore business opportunities. Additionally, IT helps develop operational proficiency 
and to exploit opportunities (Benitez et al., 2018). Their report separates an 
organization’s exploratory capability into the organizational capabilities of business 
experimentation and business flexibility. Business experimentation is defined as the 
firm’s ability to foster experimentation, creativity, and innovation of new business 
opportunities (Benitez et al., 2018). Business flexibility is defined as the firm’s ability 
to sense and seize opportunities for competitive action by changing the operational 
processes, organizational structure, and business strategies. The exploitive capabilities 
of the firm are disaggregated into operational competence and refer to the firm’s ability 
to exploit its portfolio of operational capabilities for business benefits (Benitez et al., 



 
 

2. Literature review 
 

15 
 

2018). These definitions of exploratory and exploitative activities align with our 
definition, based on March’s (1991) definitions of exploration (search, variation, risk-
taking, experimentation, play, flexibility, discovery, and innovation) and exploitation 
(refinement, choice, production, efficiency, selection, implementation, and execution). 
Benitez et al.’s (2018) study investigated the relationship between IT capabilities and 
exploitative and explorative capabilities and the relationship between explorative and 
exploitative capabilities. Their results show that IT has the most significant effect of 
enabling business experimentation, which is argued to result from IT providing the tools 
to organize information, handle collaboration, provide ideas, and enable the pursuit of 
creating endeavors. IT was also found to affect business flexibility, however, with a 
smaller impact than for experimentation. Thus, Benitez et al. (2018) state that IT affects 
creativity more than the ability to change. In regard to operational competency, or a 
firm’s exploitive capabilities, IT was also found to facilitate the development of 
operational competency to exploit opportunities. Regarding the influence of how 
explorative capabilities affect exploitive capabilities, business experimentation was 
found only to mediate business exploitation, linked through business flexibility. This, 
they argue, shows that a firm needs to have the ability to change when an opportunity 
is found to have an impact on a firm’s ability to exploit.  
 
Benitez et al. (2018) contribute to how information technology influences a firm’s 
exploratory and exploitative capabilities. Their definition of IT is broad and regards 
computer-aided technological, managerial, and technical resources that provide the 
basis of using IT applications. Benitez et al.’s (2018) definition of IT, e.g., includes 
hardware, software, and the skills needed to use, develop, and improve applications. 
Further, Boaden and Lockett (1991) state that many common terms are used as subsets 
of IT, such as decision support. Thereby, we view business analytics to be a subset of 
IT. Benitez et al.’s (2018) study call for further research on how firms leverage IT to 
create business value and its effect on firms’ exploration and exploitation capabilities 
in other countries than Spain and the US, where former studies primarily have been 
conducted. 



 

16 
 

3 

Methodology 
 
 
This chapter entails the chosen methodology and theoretical reasoning for this study. 
The chapter includes a description of our chosen research design and approach, 
qualitative data collection, and adapted analytical method. Finally, the chapter ends 
with a discussion concerning the quality and ethics of the report.  
 

3.1 Research Strategy and Approach 

According to Bell et al. (2018), research strategy refers to the overall approach of a 
research project, including methods chosen to answer the research question. 
 
Blaikie (2011) states that the main approaches to business research are of an inductive, 
deductive, retroductive, or abductive kind. The approach refers to the role of theory in 
research, where research either tests a theory through a hypothesis or generates a new 
theory. The two extremes of the main approaches are the inductive and deductive 
approaches. An inductive approach refers to generating theory solely from data 
collected, and the deductive approach tests previous theory through data (Blaike, 2011).  
Braun and Clarke (2006) state that an inductive approach is enhanced by not engaging 
in prior literature before the analysis, as it likely influences the outcome. As the aim of 
the study was set, and readings were done on existing literature before conducting the 
analysis, an inductive approach was assessed to be inappropriate, as the theory 
generated with an inductive approach should emerge without influence. The chosen 
field to be investigated, of how online retailers operating in Sweden use BI&A and how 
it is used to support their explorative and exploitative processes, is scarcely researched, 
particularly in an e-commerce setting. As previous studies were limited and this study 
aspired to enrich and elaborate on existing knowledge, an abductive approach was 
assessed to be appropriate. As per Mantere and Ketokivi (2013), an abductive research 
approach entails that the researcher adopts the best fitting theory and bases the empirical 
findings on parts of that theory.  
 
Chosen methods in business research are commonly divided between quantitative and 
qualitative methods, meaning whether the methods are based on the collection of 
numerical or composed of spoken words and images, respectively (Bell et al., 2018). 
According to Easterby-Smith et al. (2018), qualitative methods are used to gain insights 
into organizational realities through discovering the views, perceptions, and opinions 
of individuals and groups, which tends to be of an explorative nature. As this study 
aims to understand how online retailers are using BI&A, in what areas they benefit from 



 
 

3. Methodology 
 

17 
 

it, and how they perceive that this creates value and supports exploration and 
exploitation, a qualitative approach was adopted. Further, a qualitative approach has 
the advantage over a quantitative method of giving access to information in context and 
provides a more comprehensive view of the topic (Easterby-Smith et al., 2018). In this 
study, the qualitative data was gathered through semi-structured interviews and is 
further explained and motivated in section 3.2.  
 
Through continuous iteration between empirical findings and existing theory, the aim 
was to find intersections and gaps between the collected data and theory and generate 
a holistic view of the value creation process of BI&A and how it supports both 
exploratory and exploitative activities. Identifying research gaps and fully covering the 
topic has been done by reading literature on business and data analytics, dynamic 
capabilities, ambidexterity, value creation, the retail and e-commerce industry, and 
through interviews with; CEOs and managers in retail firms; digital agencies 
specializing in data and business analytics in online retailing.  
 

3.2 Data Collection  

Qualitative data are gathered non-numerical data, typically collected through 
interviews, observations, videos, images, or documents (Easterby-Smith et al., 2018). 
Qualitative interviews are a powerful tool to capture the ways people make meaning of 
experiences (Rabionet, 2011). As value and its meaning are subject to individual 
perception and not consistently measurable, interviews were chosen. Further, as the 
study aimed to uncover and describe how firms work with BI&A and how this creates 
value for them, interviews were deemed appropriate as they are descriptive in nature. 
Interviews can be either structured, semi-structured, or unstructured. Structured 
involves having predefined questions asked in a specific order, allowing for no or little 
room for flexibility in responses (Easterby-Smith et al., 2018). Unstructured interviews 
involve informal conversation without an interview guide creating more variation in the 
answers but with no or minor guidance in the subject (Easterby-Smith et al., 2018). 
Semi-structured interviews include guided open questions, leaving room for some 
follow-up questions and variation in answers while still within the defined area of 
interest. Semi-structured interviews were chosen due to the explorative nature of the 
study, pressing on the importance of having questions within the area of interest. At the 
same time, it allows for flexibility to capture information outside the knowledge of the 
authors’ earlier academic research. Primary data collected in the form of semi-
structured interviews constituted the basis for the findings of the study.  
 
The crafting of an interview guide should, as per Rabionet (2011), be guided by reading 
previous literature and work on the subjects of interest. Consequently, the formation of 
the interview guide was directed by a continuous screening of literature on topics such 
as business value, online retailing, e-commerce, ambidexterity, exploration, 
exploitation, business analytics, data analytics, business intelligence, and big data. 



 
 

3. Methodology 

18 
 

Further, Google Scholar was used for finding articles, websites, and papers on business 
analytics to identify tools and terms used in the online retailing industry. The initial 
review of the literature allowed the interviewers to familiarize themselves with previous 
research and common concepts to craft the interview guide. The interview guide was 
designed with general and open questions at initiation and specific questions further 
into the interview, which allowed the interviewers to initially take a passive role and let 
the interviewee start the interview into areas that they viewed as more critical and in 
areas they likely were well-informed. The specific questions were used to help initiate 
new topics when the interviewees had nothing to add or if the conversation steered into 
irrelevant areas. Follow-up questions were frequently used to gain exemplifications and 
illustrations of discussed topics. The interview guides were continuously updated as 
previous interviews had been transcribed and analyzed, and new topics or ones in need 
of further exemplification were uncovered. See appendix A for the used interview 
guides.  
 
Purposive sampling was chosen as the strategy for sampling, followed by snowball 
sampling. Purposive sampling involves the researchers choosing the sample of 
interviewees in line with the aim of the study (Easterby-Smith et al., 2018). In this 
study, the sample was based on characteristics of the firm, industry, company size, 
business model, and role of the interviewee. Snowball sampling involves making use 
of selected interviewees’ recommendations of other people to interview. The use of 
snowball sampling in addition to purposive sampling was chosen as the interviewees 
could give recommendations of employees with additional insights to give. 
Additionally, them knowing the industry and their competitors yielded 
recommendations of new organizations and people to interview.   
 
Easterby-Smith et al. (2018) describe representativeness and precision as two principles 
guiding decisions around sampling. Representativeness involves the sample’s 
characteristics representing the larger population of companies, while precision 
involves the credibility of the results, representing the entire population by a large 
enough sample size in relation to the total population size (Easterby-Smith et al., 2018). 
Using purposive sampling with clearly defined characteristics of the population and 
targeting the B2C online retailers operating in Sweden, the demands to achieve 
representativeness and credibility were met.  Using snowball and purposive sampling 
to find new insights does, however, create potential bias issues. While all interviewee’s 
companies were within the defined characteristics of the sample, it cannot be concluded 
that the purposive and snowball sampling design has not affected the collected data. 
Although several different roles in different companies have been interviewed, higher-
level managers could potentially be biased to portray their company in a flattering way. 
Further, internal recommendations from the snowball sampling could have been steered 
towards employees with similar views as the person giving the recommendation.  
 
The interviews were held with 14 of the largest B2C online retailers having an online 
sales channel in Sweden, with yearly revenue between 100 million SEK and 5 Billion 



 
 

3. Methodology 
 

19 
 

SEK. Further, experienced experts within the online retail space were interviewed to 
broaden the insights from how business analytics are used. The interviewees chosen 
from the online retailers had titles such as CEOs, CIOs, CMO’s, Business controllers, 
and Business development managers. Interviewees primarily with managerial roles 
were chosen as it was deemed that they likely could give a holistic view of how business 
analytics are used in their respective firms. Further, these roles generally have strategic 
elements incorporated, and therefore information could be gained on how business 
analytics impact short and long-term value creation. These chosen roles were, in some 
cases, complemented by an analyst to cover more in detail the work with BI&A. 
Furthermore, two interviews were conducted with experts in online retailing to extend 
the view of value created by BI&A. The target segment of the largest online retailers 
operating in Sweden, in their respective market segment, was chosen due to them 
having the most considerable likelihood of having BI&A implemented in their 
processes. Some companies and the associated interviewees’ names are anonymized by 
their request, either due to being publicly traded companies, restricting what they can 
share, or due to internal company policies. Interviews were conducted until theoretical 
saturation was reached.  Bell et al. (2018) define theoretical saturation as a state in 
which further interviewees do not add additional data giving new insights to the found 
concepts in the analysis. In this study, this became apparent in the last held interviews 
where no new value-adding actions of using BI&A or in the process of exploitation and 
exploration arose. This insight was reached through continuous analysis of the 
interviews, where, from the thematic coding, it could be seen that while new 
interviewees provided detailed examples, no new areas of interest arose.  
 
The interviews were conducted during a period between 30 - 80 minutes, with an 
average of 50 minutes. The interviews continued until all questions in the interview 
guide had been covered. Due to the imposed recommendations related to the ongoing 
COVID-19 pandemic, interviewees were primarily conducted through video meetings 
via Zoom or Google Hangouts. In one case, the interview was held through telephone 
without a video connection. In close relation to a held interview, the recording was 
transcribed and analyzed manually. The analysis process is accounted for in section 3.3. 
In total, 18 interviews were held, resulting in approximately 15 hours of recorded 
material and 135 pages of transcripts. Table 3.1 displays the held interviews in this 
study.  
 

Company Company type Person Role Interview date 

Curamando 
Consultancy 

firm 
John Ekman Partner 26 January 

Engelsons 
Omni-channel 

merchant 
Stefan Engelson 

CEO (Chief 
Executive Officer) 

27 January 

Adlibris 
Omni-channel 

merchant 
Sofia Söderqvist 

CMO (Chief 
Marketing Officer) 

29 January 



 
 

3. Methodology 

20 
 

Table 3.1. Interviews held (See literature review for classification of company types) 
 

  

Company A 
Omni-channel 

merchant 
Person A CEO 1 March 

Company B 
Virtual 

merchant 
Person B Head of E-commerce 9 March 

Nelly 
Virtual 

merchant 
Kristina Lukes CEO 12 March 

Kids Brand 
Store 

Omni-channel 
merchant 

Adeline Sterner CEO 12 March 

Dahlquist 
Consultancy 

firm 
Niklas Dahlquist 

E-commerce 
consultant 

12 March 

Company C 
Omni-channel 

merchant 
Person C 

CFO (Chief Financial 
Officer) 

15 March 

CDON 
Virtual 

merchant 
Rickard Johansson 

Business Intelligence 
Manager 

15 March 

Adlibris 
Omni-channel 

merchant 
Magdalena Lindh Web analyst 17 March 

Pierce 
Virtual 

merchant 
Göran Sällvin CMO 17 March 

Bubbleroom 
Virtual 

merchant 
Esko Österbacka CFO 19 March 

Nelly 
Virtual 

merchant 
Adrien Mathieu Business Analyst 19 March 

Boozt 
Omni-channel 

merchant 
Peter Jørgensen CMO 22 March 

Company D 
Virtual 

merchant 
Person D Controller 24 March 

Company E 
Virtual 

merchant 
Person E 

CRM & Marketing 
Automation Manager 

30 March 

Jollyroom 
Virtual 

merchant 
Emil Thell CFO 30 March 



 
 

3. Methodology 
 

21 
 

3.3 Analysis of Results 

The analysis of the qualitative data was done by using thematic analysis as the adopted 
methodology. Thematic analysis involves searching for recurring features or patterns 
in the data (Braun & Clarke, 2006). The chosen methodology has its advantages in 
offering flexibility by minimally organizing data while describing the data in rich detail 
(Braun & Clarke, 2006). As the study aimed to provide rich insights and describe the 
phenomenon of how BI&A creates value and affects exploratory and exploitative 
processes, this methodology was seen as advantageous. Further, thematic analysis is 
appropriate as a method for interpreting experiences, meanings, and the reality of 
participants (Braun & Clarke, 2006), which was assessed suitable from the authors’ 
realist ontology. As per Braun and Clarke (2006), a thematic analysis should be initiated 
by familiarizing yourself with the data, which was reflected in the process by 
continuously reading and discussing the content of transcripts from interviews 
throughout the data collection and analysis process. During the data collection, this 
allowed new areas of interest to be discovered, or the realization on specific topics that 
needed further elaboration and gathering of data. Further, data regarding these areas 
could then be collected through additional contact with previous interviewees and 
additions to the interview guide. 
 
The transcripts were continuously coded by breaking down the data into parts and 
attributing names. Bell et al. (2018) state that the process of qualitative coding requires 
the researcher’s interpretation of the data to shape emerging codes and themes, unlike 
quantitative research that requires data to fit into preconceived codes. The coding was 
conducted by first reading through the transcripts, making comments and 
summarizations on citations and portions of the transcripts. Further, each transcript was 
shortly summarized, and the comments were categorized and coded into either value-
adding themes, processes, tools, or organizational functions, such as or A/B-testing, 
exploitation, Google Analytics, and marketing, respectively. When an initial coding had 
been conducted in Google Docs, the coding continued with the assistance of Microsoft 
Excel. Each citation and portion were initially shortly summarized. All summaries were 
then compared and analyzed to find common elements and then grouped into common 
codes. Additional remarks were made into what functions or units were inflicted by the 
codes. The codes were based on organizational actions done with BI&A, such as A/B-
testing or customer segmentation, as the use of BI&A most often was described through 
actions. As the aim was to capture how BI&A creates value, codes could emerge from 
frequent or unique mentions, as both instances were needed to capture the value created. 
Consequently, themes could emerge from codes with as few as one mention. 
Subsequently, the codes were analyzed and discussed to decide how they should be 
grouped into themes. According to Braun and Clarke (2006), it is essential to address 
what should constitute a theme when using thematic analysis, as it is not clear-cut. 
However, the key to making a theme should depend on whether it can capture 
something important in relation to the research question (Braun & Clarke, 2006). 
During the interviews and analysis, it became apparent that certain functions and units 



 
 

3. Methodology 

22 
 

used business analytics to a greater extent than others, like marketing, and 
categorization of themes by function was discussed. However, as many codes were 
cross-functional, categorizing according to functions involved many overlapping codes, 
which made the portrayal of how value was created unnecessarily intricate, and thereby 
this idea was rejected. Further, it was concluded that the first-order themes should 
represent the expressed effect that each code/action had on the organization, which the 
codes were categorized into accordingly and constituted the first-order themes. Several 
first-order themes were multidimensional and could have been placed into several first-
order themes. These codes were placed into several first-order themes if they had 
significant implications for multiple themes, and if the code was primarily affecting one 
theme, it was placed under a singular theme. Later, the first-order themes were 
categorized into second-order themes, based on how the first-order effects implicated 
business value. These common second-order themes formed into operational 
efficiency, increased sales, and the competitiveness of the business.  
 
The thematic analysis was done separately by the authors, meaning that each 
transcribed interview was read and analyzed for finding areas of interest by both 
authors. The found areas of interest were then discussed among the authors, and when 
they were interpreted differently, discussed if to be included. The entire analysis 
involved subjective interpretation, and the involvement of both parties allowed for 
discussion and, thus limiting subjective interpretations influencing the result. Figure 3.1 
shows the coding, which forms the basis for sections 4.1 to 4.3. 
 

Code First-order theme Organizing theme 

Attribution modeling 

Digital marketing efficiency 

 
 
 
 
 
 
 
 

 
 
 
 

Operational 
Efficiency 

 
 
 
 
 
 
 
 
 
 
 
 

Automized marketing channel investments 

Customer value segmentation guiding 
marketing 

Modeling of marketing channel 
investments 

Report data to Google on customer 
choices 

Customer requests guiding customers 
service staffing 

Administration and customer 
service efficiency 

Identification of bottlenecks 

Less required personnel due to 
automatization 

Machine learning categorization of 
products 

Sales data predicting future sales 

Follow up on employment satisfaction 

Sales predictions guiding warehouse 
staffing Warehousing and logistics 

efficiency Follow up on efficiency in warehousing 
activities 



 
 

3. Methodology 
 

23 
 

Add-on service advising product fit based 
on data 

Decrease returns 

 
 

 
 

 
 
 

Operational 
efficiency 

Identification of characteristics of 
products being returned 

Identification of customers with excessive 
returns 

Identification of reasons for returns on 
certain products 

Automated purchases based on decided 
levels 

Optimization of purchases and 
stock 

Identification of bad selling inventory 

Marketing and logistics capacity 
coordination 

Marketing and stock availability 
coordination 

A/B test of shipment price effect on sales 

Average order value 

Increase sales 

Add-on service for upselling 

Product recommendations 

Product listings based on customer data 

A/B testing 

Conversion rate 

Analysis of customer journey 

Analysis of newsletter effectiveness 

Analysis of website speed 

Price scraping 

(Marketplace spec.) Product 
categorization improvement 

 
Customer retention 

 
 
 
 
 

Customer retention 

Customer requests guiding customers 
service staffing 

Customer service performance 

Customer-value segmentation guiding 
service level 

Drop shipment analysis of supplier’s 
delivery times 

Follow up on customer happiness 

Identification of bottlenecks 

Newsletter sent based on time intervals 
and purchase Increase traffic to the site 

Visualization of marketing channel data 

Add-on service to suppliers based on data Inhouse marketing sales 

Assessment of customer lifetime value 
Value segmentation of customers 

 
 

 
Competitiveness of 

business 
 
 
 

Identification of key signs of high-value 
customers 

Brand recognition tracking 
Branding, positioning, and 

assortment Identification of market and product trends 
with external data 



 
 

3. Methodology 

24 
 

Figure 3.1. Thematic coding of interviews 
 

3.4 Research Quality 

One way described by Bell et al. (2018) of assessing the quality of research is proposed 
by Lincoln and Guba (1985) and Guba and Lincoln (1994), who introduce 
trustworthiness and authenticity as criteria assessing the research quality. The 
trustworthiness criteria consist of credibility, transferability, dependability, and 
confirmability (Bell et al., 2018). 
 
Credibility has to do with the match between the observations of the researchers and 
the ideas and theory created (Bell et al., 2018). The researcher used current literature 
on qualitative research methodology to design the study in a preferable way matching 
the aim of the study. Further, transcripts and quotes were sent to the interviewees to 
check that the author’s interpretations reflected what the interviewee meant, and both 
actions were taken to increase the study’s credibility.  
 
Transferability involves the generalization of the findings, meaning how the results can 
be applicable to a broader population than the sample (Bell et al., 2018). In order to 
judge the generalizability of findings within the group sample, the researchers of this 
study have made a thorough description of the sample and each interviewee, including 
company name and type, person’s name and role, and date of interview. Secondly, 
interviews were held until theoretical saturation was reached to include as many factors 
as possible.  
 
Peers should understand the study process and be able to audit the process to achieve 
high trustworthiness. Trustworthiness is proposed by Lincoln and Guba (1994) to be 
achieved partly through dependability. It is adhered to by having a thorough description 
of the process, such as the sampling, interview guide, transcripts of interviews, and 
analysis (Lincoln & Guba, 1994). 
 
The last part of trustworthiness proposed by Lincoln and Guba (1994) includes 
confirmability, involving the researcher’s efforts to limit personal values affecting the 
results, achieving as high objectivity as possible in each part of the study. While Bell 
et al. (2018) describes that it is impossible to achieve complete objectivity, the 
researchers have made efforts to create as high objectivity in the study as possible. 
Firstly, all transcripts from interviews are available on request. Secondly, the 

Matching assortment to demand  
 
 

Competitiveness of 
business 

Social media analysis 

Price adjustments based on historical 
customer demand Increase average profitability of 

sales Price adjustments based on competitor’s 
incapacity 



 
 

3. Methodology 
 

25 
 

researchers analyzed each interview transcript separately to minimize individual 
personal values.   
 

3.5 Research Ethics 

Bell et al. (2018) state that there are four main ethical concerns to be considered when 
conducting a study: whether there is - harm to participants, lack of informed consent, 
invasion of privacy, or deception involved.  
 
To address these considerations, information regarding the objective of the study was 
informed to all interviewees. Further, every interview started with discussing whether 
the interviewee agreed on the conversation being recorded, whether they wanted 
themselves or the company to be anonymous, and if they wanted to approve eventual 
quotes used before publishing. By being upright about the intent and purpose of the 
study and giving the interviewees a second review of quotes and the opportunity to be 
anonymous, ethical concerns have been addressed. 



 

26 
 

4 

Results 
 
 
This chapter presents the value that the online retailers described was gained from the 
use of BI&A. Further, insights regarding how BI&A supports their current 
organizations’ abilities to exploit and explore are presented. The chapter is firstly 
divided into three sections (operational efficiency, increase sales, and competitiveness 
of business) which emerged from the coding as the main areas BI&A are used to create 
value. Associated subheadings are those found from the thematic analysis seen in figure 
3.1. Certain first-order themes are multidimensional and are represented under multiple 
second-order themes. These themes are accounted for under the first-order theme it 
affects most significantly. A complete representation can be seen in figure 3.1. Lastly, 
a section follows, presenting the findings on how BI&A is used to support online 
retailers’ exploitive and explorative processes. 
 

4.1 Operational efficiency 

Several interviewees describe B2C online retailing as a low-margin industry with large 
direct costs associated with each sale. An online retailer must efficiently run the 
operations to lower the direct and indirect costs to achieve profitability. The 
interviewees explained that the large direct costs associated with online sales include 
online marketing, logistics, third-party payer system fees, warehousing, procurement, 
and in some sales, costs associated with returns. The interviewees described that 
lowering the direct costs create value by a substantial effect on the overall profitability. 
A few interviewees have also described product obsolescence as an essential metric that 
influences profitability and should, consequently, be low. Product obsolescence refers 
to the loss of value for products that become harder to sell for various reasons. The 
following sections, 4.1.1 to 4.1.5, describe how the interviewees use data and BI&A to 
create value by lowering costs and increasing the inventory turnover rate. 
 

4.1.1 Digital marketing efficiency 

Customer acquisition is, by many interviewees, described to be an essential part of 
driving sales and should, consequently, be managed in a way that increases sales 
efficiently. Person C (Company C) describes:  



 
 

4. Results 
 

27 
 

“The online team looks at what the cost of marketing is in relation to the 
order intake. It is maybe the most important metric because one thing that 
often distinguishes successful from unsuccessful online retailers is the 
amount they spend on getting customers to buy. Marketing spending can 
vary from 5-20% of total the turnover, which is why we monitor this metric 
daily.” 

 
To ensure that the marketing costs do not become too high, several interviewees have 
emphasized the importance of clearly gathering and visualizing the marketing data and 
its expenditures in different channels. The interviewees further describe several 
different methods to use data and BI&A to decrease marketing costs. An issue multiple 
interviewees described is not knowing exactly how effective their used marketing 
channels are when it comes to cost in relation to the number of customers acquired. 
While all marketing channels have the same purpose of acquiring new customers, the 
different channels create problems attributing the value from the investment. 
Attribution modeling is one method to understand better how each marketing channel 
contributes. Adeline Sterner (Kids Brand Store) highlights this: 

“Many times, you work with last-click, meaning that you see the acquired 
customer’s last clicked advertisement. However, a person choosing to click 
a certain ad might do so because he or she has seen the product and our 
brand in ten different other places, which is influencing the decision to click 
on the advertisement. This makes it seem like the last clicked ad is 
exaggeratedly cheap when in reality, other communication also influenced 
the decision. This is why we work with attribution, to actually see how 
effective each channel is”- Adeline Sterner (Kids Brand Store) 

 
Person B (Company B) further exemplifies this: 

“The advertising on Facebook is never on its own getting enough credit 
without us examining the ad influence. Hence, we ask the customer how 
they came to our site, giving us data which we then analyze to understand 
the actual number of acquired customers from Facebook.”- Person B 
(Company B)  

 
Attribution models try to assign the newly acquired customers to the correct marketing 
channel by capturing data on each interaction the customer has with the different 
marketing campaign and when the customer converted into a buying customer. These 
models are then adjusted based on these interactions to make the model as accurate as 
possible. Further, some interviewees stated that they asked the customers how they had 
found their site.  
 



 
 

4. Results 

28 
 

Some of the interviewees take this attribution modeling of seeing how profitable they 
are in each marketing channel one step further by automating the processes of the 
investments in the different marketing channels. Göran Sällvin (Pierce) is one of the 
interviewees who illustrate this: 

“Marketing is automated based on different data points from which we 
drive either to maximize profitability or revenue. We have built a dataset to 
see how profitable we are in each marketing channel. [...] To achieve the 
set profitability goals, we, therefore, built a model based on 2020 data 
where we can simulate 2020 again at different investment levels, creating 
a normal distribution curve showing the point of investment to achieve 
maximum growth in 2020. This model is then applied to 2021 to see how 
large the marketing investment should be to maximize profitability and/or 
growth” – Göran Sällvin (Pierce) 

 
Another way of decreasing cost in marketing described by interviewees is using 
segmentation of customers based on value. Online retailers want to advertise to 
customers with a high likelihood of buying upon seeing the advertisement since this 
lowers the amount needed to be spent to achieve a certain number of sales. Peter 
Jørgensen (Boozt) describes this: 

“In early days, we decided to make customer segments based on their 
behavior, divided into three tiers in terms of value. [...] The customers are 
divided into 15 different segments based on what they in the past have 
bought, to what price, and the number of products they have returned. If 
you, for example, are a customer who has bought much without discount, 
not returning any products, then you are regarded as a very high-value 
customer. By doing this, we can focus marketing on the high-value 
customers but also make sure to exclude the ones we know buys very small 
amounts at a discount, or are frequently returning products.” - Peter 
Jørgensen (Boozt) 

 
All interviewees emphasized using Google Analytics and Facebook in marketing for 
advertisement and as a means to understand the customer journey better. While the data 
presented by Google was widely seen as very good and valuable for all the online 
retailers, Person B (Company B) highlighted one additional way of decreasing 
unnecessary spending in marketing with Google Analytics by reporting data back to 
Google and Facebook. One example, described by Person B (Company B), is when a 
customer decides to withdraw a product from their shopping cart. When this happens, 
it might not be desired to advertise the same product to the customer on Facebook or 
Google since they have shown not to want the product. Person B (Company B) 
describes this as a way to lower the cost and protect the brand by not annoying the 



 
 

4. Results 
 

29 
 

customer with advertisements on things they have clearly shown that they do not want 
to buy at the moment.  
 

 4.1.2 Administration and customer service efficiency 

In some cases, administration costs in online retailing are possible to decrease with no 
or small effect on operations with BI&A. To understand how much administration is 
required, online retailers need to forecast and set budgets. The previously introduced 
forecasting model based on value segments also have the possibility to do this, Peter 
Jørgensen (Boozt) explains: 

“We have seen that customers in different value segments come back with 
a certain likelihood. Based on the number of customers in each value 
segment, we can fairly precisely predict how much they will spend with us 
in 2021. This, together with past years’ data on the number of newly 
acquired customers, becomes the basis for our forecast. With an added 
predicted number of returns you can get a projected revenue over the next 
months and years” - Peter Jørgensen (Boozt) 

 
With these predictions, operations can be planned accordingly. Thus, keeping costs of 
administration such as the number of employees at the correct level makes the company 
deliver on sales effectively while keeping costs down. Another way of keeping costs of 
administration is to automate certain processes. Adeline Sterner (Kids Brand Store) 
describes, “Since we started being more data-driven, we have managed to cut costs by 
removing certain roles which instead can be automated.”  Person C (Company C) 
exemplifies this with a machine learning use case lowering administrative costs, 
“Categorization of articles on our websites is now classified with machine learning, 
allowing our articles to be classified, for example, rustic and Scandinavian with a 
certain color.”  
 
Person C (Company C) further explains that this allows them to decrease cost and 
improve categorization, as the machine learning algorithm does more precise work than 
employees manually categorizing.  
 
Another cost of running an online retailing site is caused by providing customer service. 
Magdalena Lindh (Adlibris) describes how their development in BI&A has created a 
better understanding of bottlenecks in their operations. Rickard Johansson (CDON) 
describes a use case of analyzing past data to decrease customer service cost: 
“Customer service decides on the number of working hours to be put in, based on the 
number of customer requests. This is based on prior data, which makes them keep 
control of employee costs”. This analysis of past data on the number of requests and 
time it takes to answer the requests enables CDON to limit the number of workers in 
customer service being too unoccupied at work.  



 
 

4. Results 

30 
 

 

4.1.3 Warehousing and logistics efficiency 

An online retailer needs to store its products. Storage can be done internally by 
operating a warehouse or by letting a third-party store the products. When operating a 
warehouse, the retailer needs to run warehousing operations as efficiently as possible 
to keep the associated costs down. In warehousing, as with customer service, the 
number of employees needs to be adapted to demand and sales, as Person B (Company 
B) describes: 

“The staffed hours in the warehouse are based on the forecast of sales 
ahead. If the forecast turns out to be far from the actual sales, then either 
the products will be delivered late to the customers due to lack of personnel 
or the other way around, our personnel doesn’t have enough to do, bearing 
unnecessary cost for our company.” – Person B (Company B) 

 
As mentioned by Rickard Johansson (CDON), this type of forecast also applies to 
logistics: “The logistics function has a good understanding of which and when products 
will sell. This is then supported by adding additional trucks, so we are able to deliver 
to the customers”. While forecasts can support the planning, it does not help the 
efficiency of the actual work. Several interviewees have described the use of different 
metrics measuring the efficiency of the warehouse operations requiring data and 
analysis.   

“One of the highest costs for us is those associated with the logistics of the 
product. To continuously monitor those costs, to see the change of, for 
example, how effective we are at picking and packing the product and then 
to take action to improve the processes is essential” - Esko Österbacka 
(Bubbleroom) 

“It is very valuable in the warehouse to optimize different processes such 
as where the product should be placed. Preferably having the ones being 
sold the most, closest to the packing stations, to decrease distance for pick-
up. And also, to optimize how large the packaging to different orders should 
be, so that not too much air is paid for” - Adrien Mathieu (Nelly) 

 
Several companies interviewed work continuously trying to streamline warehouse 
processes to cut costs and improve delivery speed. E.g., Boozt has automated processes 
in the warehouse, using robots and a warehouse management system. Peter Jørgensen 
(Boozt) describes their warehousing operations: 



 
 

4. Results 
 

31 
 

“Basically, the warehouse is close to being automated, having around 600 
robots picking items. The use of robots takes some cost out of the system. 
The robots work very fast and all around the clock, cutting costs but also 
decreases the time from picking to delivering to the customer. In this, the 
data is essential.” - Peter Jørgensen (Boozt) 

 

4.1.4 Decrease returns 

Returns of products bear a substantial cost for online retailers, as they often pay for 
transportation and the costs associated with inspecting, repackaging, and reselling the 
product. How high the cost is and, consequently, the priority to lower returns depends 
largely on what type of product the retailer sells. For example, Adlibris selling books 
describes having a small number of returns on sold products, hence a small effect on 
their bottom-line result. In contrast, several of the interviewees representing online 
retailers selling clothes describe the opposite, where returns are associated with high 
costs. While the importance of decreasing returns varies, almost all interviewees apply 
analysis to decrease returns. The different analyses the retailers use to decrease returns 
are by identifying and managing; characteristics of products being returned; issues 
leading to returns; and certain customers excessively returning products. Esko 
Österbacka (Bubbleroom) describes one example of how they analyze product returns 
in terms of product characteristics. 

“We have certain products that we analyze when it comes to returns. For 
example, certain dresses need to fit perfectly, not to be returned, while other 
dresses are more forgiving in terms of perceived fit. This gives us 
information on returns and with that the actual cost of a product, when 
adjusted for the return cost of the past. This guides our pricing of that 
specific product and whether the product is profitable or not. This 
information then guides our purchasing department on what to buy and 
what not to buy. From that analysis, we can see that, for example, a certain 
dress is returned too often, making it non-profitable, which makes us 
exclude it from the overall assortment. If it isn’t important to have it in the 
assortment for another reason” – Esko Österbacka (Bubbleroom) 

 
Adeline Sterner (Kids Brand Store) describes how they try to identify reasons for 
product returns and how these insights drive actions to decrease returns. 



 
 

4. Results 

32 
 

“We look at what categories are being returned to most. Of the products 
that are being returned the most, we try to find the underlying reason/s. 
Often, it is simple things such as the product being wrongly stated as small 
in size or the picture showing a certain color of the garment not 
representing how it looks in real life. When we have identified the reason, 
we try to correct it.” – Adeline Sterner (Kids Brand Store) 

 
Some returns can be explained with product defects or inaccurate demonstration, but 
not all returns. Several interviewees explained that through customer data analysis, they 
have identified that certain customers buy excessive amounts of products knowing that 
they will likely return them, as the return is free of charge. In this case, the analysis is 
made on the individuals instead of the product. When the customers misuse the free 
returns system, actors have taken different actions, for example, by banning these 
customers from making purchases, making sure not to advertise to these customers, or 
by giving out warnings. 
 

4.1.5 Optimization of purchases and stock 

Managing purchasing and stock levels are essential for an online retailer to deliver the 
right products at the right time efficiently. Further, it is essential for retailers with their 
own warehouses to sell the products as quickly as possible, to decrease the risk of long-
term storing products that lose value, and to make room for other products in storage. 
If certain products do not get sold, direct losses in the form of lost sales and indirect 
storage costs arise. Products that do not sell will lose value if they become outdated 
technically, style-wise, or expire and likely must be sold at a lower price. To decrease 
the number of products not getting sold, several interviewees mention applying analysis 
on product sales data to early identify products that should be decreased or cut from 
purchases. By doing so, the interviewees describe that it lowers the costs of product 
obsolescence and more effective use of invested capital, as less capital is tied to stored 
products. A large part of the interviewees further describes having predetermined levels 
of automated purchases, triggering new signals of purchase when stock levels of certain 
products reach below a certain number.  Göran Sällvin (Pierce) describes one example 
of this “In terms of the supply chain, we have a system in place which predicts what 
purchases we should do, based on how the sales of the products develop. It helps us 
minimize the risk of having empty inventory”. While one of the main objectives of the 
purchasing departments of the online retailers is to optimize the purchases to meet the 
demand, a few interviewees mention the use of BI&A for improving coordination of 
cross-functional activities. Since the marketing department wants to market products 
that the customer is most likely to buy, the company must have the right products in 
stock to deliver. To make sure that this is the case, Person B (Company B) describes 
how they use data to increase this coordination. 



 
 

4. Results 
 

33 
 

“We have data on stock balances of all products together with their 
coverage of sizes. Before deciding on pushing certain product to potential 
customers through marketing, we make an analysis and check that we have 
a good stock balance and coverage of sizes.” – Person B (Company B) 

 
Capacity coordination can also involve other activities such as measuring and 
understanding capacity in logistics. A few other interviewees have additionally stated 
that the coordination of the capacity of the company and the marketing activities should 
align, not only on stock balance but also on logistics, such as the ability to pack and 
transport products. 
 

4.2 Increase sales  

Increasing sales in an efficient manner were described to be, at least, equally