Organizational aspects to consider in order to implement machine learning
and create business value
- A qualitative study on a technology industry leader
Spring 2020:2019KANI47 Bachelor’s thesis in informatics Jacob Cressy Pontus Friis-Liby
II
Swedish Title: Organisatoriska faktorer att överväga för att kunna implementera maskininlärning & skapa affärsvärde. En kvalitativ studie på ledande företag inom teknologiindustrin.
English Title: Organizational aspects to consider in order to implement machine learning &
create business value. A qualitative study on a technology industry leader.
Year: 2020
Researchers: Jacob Cressy & Pontus Friis-Liby
Supervisor: Anders Hjalmarsson Jordanius
Abstract
This thesis investigates the relationship between technology, specifically machine learning and business value. By incorporating fundamentals of today such as a socio technical
perspective and the fourth industrial revolution, commonly referred to as Industry 4.0 and this thesis will explain if machine learning can create business value in the organization as well as what organizational aspects a company needs to obtain in order to create business value from its data.
This was achieved through a qualitative study by conducting semi-structured interviews with a technology industry leader, two experts on the subject and complemented with closely evaluated sources gathered from articles, journals, published literature, websites and case studies.
The analysis of this thesis is based on an analytical framework that concludes five
organizational aspects. It shows a clear relationship between the five organizational aspects the analysis is based on in regards to data and internal structure that needs close consideration in order to create business value from data as well as the possibilities of implementing and utilizing machine learning. By having a socio technical perspective in mind the analysis shows and confirms the theories this thesis is based on as well as it gives a nuanced perspective on an industry that primarily focuses on technical factors.
The researchers conclude that apart from the five organizational aspects the analysis is based on, a sixth aspects needs to be incorporated as well. By conducting a study on a large
technology industry leader as well as two additional experts on the subject, the researchers conclude that machine learning can benefit large parts of the organization as well as the workforce employed.
Keywords: Machine learning, socio technical perspective, industry 4.0, business value &
organizational support systems.
III Sammanfattning
Uppsatsen undersöker relationen mellan teknologi, specifikt maskininlärning och affärsvärde.
Genom att inkorporera begrepp som är fundamentala idag såsom ett socio tekniskt perspektiv och den fjärde industriella revolutionen ska den här uppsatsen förklara om maskininlärning kan skapa affärsvärde i organisationen men även vilka organisatoriska aspekter ett företag behöver erhålla för att skapa affärsvärde från sin data.
Det uppnåddes genom en kvalitativ studie där semi-strukturerade intervjuer genomfördes med ett ledande företag inom teknologi industrin, två experter inom området och kompletterades med noga utvärderade källor som samlades in via artiklar, tidskrifter, publicerad litteratur, webbsidor och fallstudier.
Analysen i uppsatsen är baserad på ett analytiskt ramverk som innehåller fem organisatoriska aspekter. Analysen visar en tydlig relation mellan de fem organisatoriska aspekterna som analysen är baserad på i relation till data och intern struktur som kräver noga övervägande för att generera affärsvärde från data men även möjligheterna av implementering och användning av maskininlärning. Med ett socio tekniskt perspektiv i åtanke visar och bekräftar analysen de teorier som presenterats och ger även ett nyanserat perspektiv på en industri som primärt fokuserar på tekniska faktorer.
Forskarna drar slutsatsen att förutom de fem organisatoriska aspekterna som analysen är baserad på, krävs även att en sjätte aspekt introduceras. Genom att genomföra en studie på ett ledande företag inom teknologi industrin samt två experter inom ämnet, drar forskarna
slutsatsen att maskininlärning kan gynna stora delar av organisationen samt även de anställda.
Nyckelord: Maskininlärning, socio tekniskt perspektiv, industri 4.0, affärsvärde &
organisatoriska supportsystem.
IV
Acknowledgements
We would first and foremost like to offer our sincere gratitude to the company examined, the respondent, the experts and the person within the researchers network for taking their time to be a part of our study and provide an insight into their organization. We would also like to thank our supervisor Anders Hjalmarsson Jordanius for his help, advice and guidance.
V
Table of content
1.0 INTRODUCTION ... 1
1.1BACKGROUND ... 1
1.2PURPOSE ... 2
1.3PROBLEM DISCUSSION ... 2
1.3.1 Research questions ... 3
1.4TARGET GROUP ... 3
1.5DELIMITATIONS ... 3
1.6DISPOSITION ... 3
2.0 METHODOLOGY ... 5
2.1RESEARCH DESIGN ... 5
2.1.1 Qualitative method ... 5
2.2SAMPLE TECHNIQUE ... 5
2.3DATA COLLECTION ... 6
2.3.1 Primary data collection ... 6
2.3.2 Secondary data collection ... 6
2.4INTERVIEWS... 7
2.4.1 Conducting the interviews ... 8
2.4.2 Interview guide ... 9
2.5DATA ANALYSIS ... 10
2.6ETHICAL CONSIDERATIONS ... 10
2.6.1 Reliability & validity ... 11
2.7LIMITATIONS ... 12
2.7.1 Primary data ... 12
2.7.2 Secondary data ... 13
3.0 THEORETICAL FRAMEWORK ... 14
3.1SOCIO TECHNICAL PERSPECTIVE ... 15
3.2INDUSTRY 4.0 ... 16
3.3MACHINE LEARNING ... 18
3.4BUSINESS VALUE ... 19
3.5ORGANIZATIONAL SUPPORT SYSTEMS... 21
3.5.1 Decision support systems ... 21
3.6SUMMARY ... 22
4.0 RESULT ... 24
4.1USE OF MACHINE LEARNING AND ARTIFICIAL INTELLIGENCE TODAY ... 24
4.1.1 Text analysis case ... 25
4.2CREATING BUSINESS VALUE WITHIN THE ORGANIZATION ... 25
4.3OBSTACLES ... 26
4.3.1 Data quality ... 26
4.3.2 Unstructured data ... 26
4.3.3 Security ... 27
4.3.4 Employees ... 27
4.3.5 Risks ... 28
4.4FUTURE ... 29
5.0 ANALYSIS ... 32
5.1DATA QUALITY ... 32
5.2DATA INTEGRATION... 33
5.3DATA SECURITY ... 33
VI
5.4ANALYTICS ... 34
5.5HUMAN TALENT ... 34
5.6SUMMARY ... 35
6.0 DISCUSSION... 37
6.1OMEGA ... 37
6.2DATA QUALITY ... 37
6.3DATA INTEGRATION... 37
6.4DATA SECURITY ... 38
6.5HUMAN TALENT ... 39
6.6DISCUSSION OF RELIABILITY, VALIDITY & TRANSFERABILITY ... 40
7.0 CONCLUSION ... 42
7.1FURTHER RESEARCH ... 43
8.0 REFERENCES ... 44
9.0 APPENDIX ... 47
9.1APPENDIX 1 ... 47
9.2APPENDIX 2 ... 48
9.3APPENDIX 3 ... 48
1
1.0 Introduction
Technology has presented opportunities to people and organizations around the globe that no one, or at least very few people saw coming. The effects of the technology and internet progression have resulted in a globalization that has freed companies from operating solely in a specific geographical environment and/or in a small market. Obviously with globalization comes competitors to a much larger extent, which means that time is of the essence in production and delivery.
The global Machine Learning (ML) market was valued at around 2.5 billion USD in 2017 and is expected to reach around 12.3 billion USD by 2026. The global ML market includes a wide range of services, solutions and techniques closely connected to Artificial Intelligence (AI).
The market is influenced by a number of factors such as a growing demand for improved application areas, development associated with AI and lack of trained professionals. The forecast of the market is that it will grow exponentially during the period mentioned above due to increasing automation and advanced technology (The Express Wire 2019).
The subject of ML is a hot topic at the moment due to the fact that with globalization and urging demand from customers, requires organizations to produce and deliver with a rapid pace and continuity. By researching the matter of the subject and conducting a study on an organization in the technology industry a clear relationship between ML implementation and business value creation has been obtained. This was done through the perspectives of socio technical factors and the fundamentals of Industry 4.0.
This thesis will try to explain the fundamentals of ML in relationship to business value and how it can be created in an organization. In order for the researchers to see the relationship between business value and ML multiple organizational aspects has been taken into close consideration.
1.1 Background
In today’s highly competitive society technology leads the way. The importance lies in seeking new opportunities and solutions for organizations considering that business possibilities has grown due to globalization. The highly competitive technology industry makes no difference in this field. The importance of speed and accuracy is vital to any organization, not the least in the manufacturing industry.
The company this thesis is based on is a technology industry leader in pulp, paper, business and automation technologies. The organization covers four main business lines: providing new machinery and technology to paper import businesses, serving pulp and energy
corporations, providing automation and technology to the processing industry of automation businesses and providing service for the pulp & energy industry as well as the paper import industry. The key processes of the organization are Research & Development (R&D), marketing, sales, engineering, procurement, production, delivery and service. The annual turnover is over three billion euros and the company employs around 13 000 employees.
The company is active in five business areas covering the majority of the globe: Europe, Middle East and Africa (EMEA), North America, South America, China and Asia-Pacific.
Although the company is in every way possible a global competitor, the core of the business is in the Nordics with technology and general know-how.
2
In addition to the company examined two external parties were interviewed in order to process and validate the data gathered. By incorporating two additional parties in the same industry but with varied experience and knowledge in the field, a broadened perspective and understanding for the subject was reached. The first additional party that was interviewed was SAS Institute which is a large, global organization working with software analytics
development. The second party was a small company based in Sweden focused on system development.
Technology and systems are crucial to any organization and in this day and age multiple different aspects have to be taken into consideration when developing a new system. The socio technical perspective and Industry 4.0 includes aspects that previously weren’t taken into consideration in the same way. Bansler (1989) argues that the end users should be participating in the development of systems due to the fact that they are the ones that will execute the day-to-day operations in the system, they are familiar with the organizational structure but most importantly the actual work process. One of the areas where Industry 4.0 will differ from previous industrial revolutions is how the human factors and technological factors will work together rather than be separated.
ML is a subset of what is commonly known as AI which is a technology that in recent years has received a lot of attention. As the evolution of internet progresses an increasing amount of data is generated which in this day and age is considered a valuable asset to organizations around the world. Directing yourself through the maze of data and technology is a hard task to anyone and any organization often because of the sheer amount of data a large company possesses. ML provides technologies to solve complex problems as well as finding patterns in large amounts of data which cuts time and makes processes more effective (Marton, Hounsell
& Entwistle 2005).
What is the definition of value? Well, the word has different meanings depending on the context it’s put in. This thesis will explain how business value can be created in relation to data, technology and implementation of organizational support systems. Different aspects have to be considered when creating value from data in a business such as data quality, data integration, data security, analytics and human talent (Grover, Chiang, Liang, & Zhang 2018).
These are what the researchers refer to as organizational aspects. Creating business value from data is no easy task and takes a lot of effort from the company at all levels of the hierarchy.
1.2 Purpose
The purpose of this thesis is to determine what organizational aspects, according to the researchers’ analytical framework a company must consider when implementing machine learning in order to create business value through organizational support systems. With a socio technical perspective as well as the fundamentals of Industry 4.0 these aspects were taken into consideration and laid the ground for the analytical framework.
1.3 Problem discussion
In order for industrial organizations to cope with the urging demands technology is and will continue to be a major contributing factor. The fact that we are moving towards Industry 4.0 which emphasizes on technology and its surroundings confirms the increasing power
technology together with social factors will have in the future of industries. One of the urging
3
problems the organization faces today is unstructured data which makes implementation of new technologies, in particular ML challenging. However, that is not the only obstacle an organization has to overcome in order to utilize new technologies efficiently. Other aspects such as data security and human factors needs to be considered as well in order to stay competitive in the future. In order for the researchers to try to address the problem a socio- technical approach has been adapted as well as the cornerstones of machine learning.
1.3.1 Research questions
In the light of the industry sector moving into Industry 4.0 it becomes necessary to explore if ML can create value for different organizational systems within a company. This could be accessed from either a technical perspective, social perspective or a combination of both.
Given this the following research questions were formulated:
Can machine learning help create business value for a company in the technology industry?
What organizational aspects must a company obtain in order to implement machine learning to create business value?
1.4 Target group
The main target group for this thesis are organizations in the technology industry who are looking to implement ML. This thesis will address the problems facing organizations today and the organizational aspects they need to overcome in order to utilize the full potential of ML and if it can help to create business value. Although this study was conducted with an organization in the technology industry this thesis will address problems that in addition to the technology industry will provide valuable information for organizations in other fields of business.
1.5 Delimitations
ML can be implemented into multiple departments of an organization. This thesis will focus on if ML can be of aid and create business value in a large company within the technology industry based in the Nordics. Even though implementation of ML will be discussed this thesis will not provide concrete examples of how to implement ML but rather what is required before an implementation can take place in order to create business value.
1.6 Disposition
Chapter 1 – Introduction
The first chapter is an introduction to the subject of ML and how it can help to create business value. It also serves as an introduction to the company examined so that the reader can
understand what type of company that the research is based around.
Chapter 2 – Methodology
The second chapter explains what type of methodology that has been used for this thesis as well as how the collection and analyzing of both primary and secondary data for the thesis have been made. It will also explain how the interviews were conducted in more detail as well as the researchers reasoning from the first to the fourth and final interview.
4 Chapter 3 – Theoretical framework
The third chapter gives the reader relevant information about different concepts used in the thesis. Such as socio technical perspective, machine learning, industry 4.0, business value and organizational support systems and how they all intertwine.
Chapter 4 – Result
The fourth chapter provides the results from the interviews held with a representative from the company examined as well as the additional interviews from the external parties.
Chapter 5 – Analysis
The fifth chapter analyses the results found in chapter four through a thematic analysis method that laid the ground for the analytical framework.
Chapter 6 – Discussion
The sixth chapter discusses the findings from the analysis in relation to the research questions.
It will also discuss the reliability, validity and transferability of the thesis.
Chapter 7 – Conclusion
The seventh chapter concludes the thesis by answering the research questions. It also provides examples for further research.
Chapter 8 – References
The eight chapter presents the sources used in the thesis.
Chapter 9 – Appendix
The ninth chapter presents the questions asked during the interviews in chronological order through appendix one, two and three.
5
2.0 Methodology
This chapter will explain how the researchers have conducted the study for this thesis and how the data was collected and analyzed. This chapter also contains the researchers’
reflections about the chosen method as well as sample technique, the data collected, the interviews, the interview guide, ethics and the limitation of the study.
2.1 Research design
Creswell (2009) emphasizes two main methods that could be used for data collection, quantitative and qualitative method. The data collected in a quantitative method is organized because the result of the study is measurable and statistical. The data in a quantitative method can be categorized and placed into diagrams for analyzing. In a quantitative method, data from a large number of participants can be gathered and further analyzed as it has the benefit of being able to cope with a larger scale then a qualitative method. In a qualitative method the data collection focuses on words instead of a large number of respondents. According to Bryman (2012) the focus is on conversations for instance interviews and not numbers.
However, a drawback with choosing a qualitative method is that the number of respondents are quite small compared to a quantitative method. It’s not simply possible to conduct the amount of in-depth interviews and to analyze them compared to a quantitative method where it's easier to analyze larger numbers of respondents (Creswell 2009).
2.1.1 Qualitative method
To get a deeper understanding of how a company in the technology industry operates and the challenges they face when using and implementing ML, it was important for the researchers to study the company in its natural environment. The research strategy therefore selected was a qualitative strategy. The main reason why the researchers chose this strategy was because the thesis aims to solve a problem from a company´s standpoint. To be able to solve this problem the researchers needed to speak to a company to be able to get an insight into their organization. Furthermore, to validate the data presented by the company examined expert knowledge was needed. This meant that a qualitative interview also fulfilled the requirement for the interviews with the experts and provided an external point of view on the subject.
The data collection, together with the company and the experts was done by conducting two semi-structured interviews with a company employee from the company examined, one semi- structured interview with a representative from SAS Institute and one semi-structured
interview with a software developer. The interviews were conducted in the form of semi- structured interviews with the help of interview guides that followed specific topics. A semi- structured interview enabled the researchers to ask more questions following a pre specified question to get more information from the respondent if that was necessary (Bell, Bryman &
Harley 2018). This allowed the conversation during the interview to flow more freely but it also allowed the researchers to collect more data on topics that firstly weren’t intended.
According to Creswell (2009) a semi-structured interview allows the respondent to answer the questions in their own words while the technique provides structure.
2.2 Sample technique
The researchers opted to get in touch with employees and experts within organizations with very good knowledge, expertise and experience on his/her department as well as a good understanding of the subject and a strive for new technology implementation. Through
6
individuals with this kind of expertise the researchers would be able to answer the research questions of this thesis. It would also fulfill the purpose of the study as the respondents would be highly credible and therefore providing a reliable result for the thesis.
According to Bryman (2012) there are three different categories when choosing companies, comfort, strategic and network selection. The researchers opted to choose a network selection as it is based around prior connections to a company. The researchers thought this was the best option for the study as it increased the possibility of a company’s involvement.
In advance the study was limited to only ask organizations and experts for involvement, as the thesis aims to solve a problem from a company’s standpoint. Therefore, the researchers saw no reason to contact private individuals or to conduct a quantitative study as focus lied on getting an insight into how a company operates rather than how an individual think.
At the beginning of the study a company from the researchers’ network was asked to take part in the study via email. The first email sent to the company included what the purpose of the thesis was, what would be required from the company and how they would benefit from taking part. After the interviews with the company were conducted, the researchers contacted experts in the field to get their point of view on the collected data. The experts were also contacted via email and were given the same introduction to the subject as the company and also information about the prior interviews.
2.3 Data collection
Data collection often represent a crucial part of a research project (Bryman 2012). The researchers therefore placed great importance in the data collection and divided the data collection into primary and secondary data. The primary data collection was gathered through semi-structured interviews with an employee from a company within the technology industry as well as with two experts in the field, one software analytics developer and one software developer. The secondary data was gathered through articles, journals, published literature, websites and case studies.
2.3.1 Primary data collection
The primary data collected is based on a qualitative method and was gathered through semi- structured interviews with a company employee and experts (Bryman 2012). The primary data collected provided a foundation to answer the research questions, objectives and by that provide perceptions and understandings of the company, their processes & structure. By conducting semi-structured interviews with an employee from the organization provided the researchers with information mainly from a managerial perception as well as an internal perspective. The semi-structured interviews held with the experts provided the researchers with a broader perspective on the subject as well as an analysis of the data collected from the company.
2.3.2 Secondary data collection
Secondary data was used and collected simultaneously during the writing of this thesis. The secondary data was gathered from multiple sources such as published literature, journals, articles, websites and case studies that provided an understanding on the subject both from a technological point of view as well as an understanding if implementation of ML is possible.
According to Bryman (2012) a good basis for theory is published literature. To find published literature the researchers have primarily used the University of Borås’ search engine Primo,
7
Google scholar as well as visited libraries. To search the web for information can however have its drawbacks. Bryman (2012) warns researchers that by using the web to search for information can be dangerous as search engines provides all the sites it finds and do not assess the information. Therefore, the researchers had to be critical towards the information found when searching the web and evaluate it. The researchers had to reflect about the author of the information, the reason why the author published it and also when the site was last updated to ensure the trustworthiness of the site (Bryman 2012).
Keywords that the researchers have used when searching literature on the web have primarily been: Machine Learning, socio technical perspective, industry 4.0, business value and support systems.
2.4 Interviews
As stated in the sample technique the researchers opted for a person with good knowledge, expertise and experience in order for the result to be credible. Based on the requirements from the researchers, the organization selected a person that fulfilled the criterias. The respondent is employed as a manager at Global Financial Operations (GFO) and works with project management and implementations in regards to finance and technological development. The respondent manages a team with nine employees who are fully dedicated to technology development within the organization. The researchers argue that the interviewee was the correct fit to fulfill the purpose of the thesis due to the fact that the respondents only focus is on technology and development which is what the researchers opted for. The researchers view the respondent as an essential part of the organizations continuous strive for development in general and in particular the technological developments at GFO. Because of the fact that the respondent solely works with development and technology the researchers were confident that the respondent could provide a solid picture of the organization and their strategies towards future use of technologies.
The researchers opted to first conduct an explorative interview with the respondent from the company. Due to the fact that the subject of the matter did fit both parties well the researchers were granted the opportunity to conduct a second, in-depth interview with the respondent.
The fact that the researchers were able to conduct two interviews with an individual with that sort of knowledge, expertise and experience is perceived as a major benefit for the thesis in the eyes of the researchers.
The requirements for the experts were equal to the respondent from the company in order for them to be credible. Based on these requirements the researchers got in contact, through their network, with two different people that are perceived as experts in the field. Firstly, there is the respondent from SAS Institute, Christer Bodell. Christer has over 35 years of experience in the field of software for analytics and is employed by SAS Institute as a business advisor for the manufacturing industry in the Nordic region.
The second expert that was interviewed for this thesis is employed as a software developer for a small sized company in Sweden. Both parties expressed their wishes to be anonymous. The software developer is a younger person and the researchers argue that this person provided the thesis with a nuanced picture of the subject. The subject of ML is something that the software developer works with on a daily basis and was therefore the correct fit for the thesis as the person is an expert at the technological side of ML.
8
To establish contact with the respondents’ email conversations took place. During the email conversations it was stated that two of the respondents and the company they worked for wished to be anonymous while one declined the option to be anonymous. The respondents were notified about the themes the interview questions revolved around in order to prepare without knowing the questions.
Before the researchers started to ask questions according to the appendices, the same general information was stated to all respondents. Firstly, the researchers asked if they had permission to record the interviews in order to transcribe the material so that they could draw as much valuable data as possible from it. Following this, the researchers declared that, if the respondent and the company had chosen to be anonymous the researchers assured that information that could be directly linked to either one of them would be removed during the transcription. The respondents were also given the option to decline to answer any question due to any reason in order to make sure that the respondent felt comfortable. Lastly the respondents were informed that the entire thesis along with the key findings would be sent to them after approval.
The interviews with the company were conducted on video conference over Skype due to geographical reasons and the interviews with both experts were held in person. The two interviews with the company were held on the 13th and the 19th of December 2019. The interviews with the experts were held on the 6th and the 9th of March 2020. All the interviews were recorded on an either an iPhone 6s or an iPhone 11. The researchers decided to divide the interview between them in order to give the interview structure. One of the researchers asked all the questions while the other researcher listened, took notes and asked additional questions at the end of the interview.
2.4.1 Conducting the interviews
The interviews conducted for this thesis were semi-structured interviews. Before the first interview an interview guide was created by the researchers. The interview guide for the first interview was constructed with respondent’s area of expertise in mind in order to draw as much valuable information as possible from the interview. To further ensure that the data collected would be valuable, the questions constructed had to be open-ended questions in order for the respondent to answer the questions in their own words (Creswell 2009). This worked well as the respondent gave detailed answers to the questions asked.
A semi-structured interview also makes it possible for the researchers to ask follow-up questions if needed (Bell, Bryman & Harley 2018). Follow-up questions were something the researchers utilized to collect more information on certain topics which weren’t firstly intended. Examples of follow up questions that emerged during the interviews were “Could you explain a bit more about how you think that jobs will stay in Scandinavia rather than leave with utilizing Machine learning?” And “So rather than laying of people this will be a natural transition due to the technology improvement?”.
The exploratory interview started by asking the respondent about their duties at the company to make sure that the respondent felt comfortable. Following the information about the
respondent, more specific questions were asked about how the company uses ML and/or AI at present day to provide a background. The third theme revolved around business value to see if that was something they were actively working with and their approach in regards to the subject. The final theme revolved around obstacles and how they see the future in regards to ML. The idea behind this theme was to get an introduction to the research topic from the
9
respondents point of view and so that the researchers could prepare for a second, in-depth interview. Lastly the researchers asked if there were some additional things to bring up in regards to the subject.
After the first interview the researchers transcribed the data and started to analyze it. The researchers focused on obstacles with ML and why the company hasn’t implemented it to a larger extent, which became the basis for the second, in-depth interview.
The second interview served as a more detailed interview with the purpose of collecting information to be able to answer the research questions. The focus, as mentioned was on obstacles and to understand the company’s point of view on ML. Since the first interview covered the obstacles in general, the researchers constructed more specific questions to try to dive deeper into what is causing issues.
After the second interview the researchers analyzed the data gathered from both interviews in order to prepare questions for the experts in order to later analyze the result. An interview guide was then constructed and used for both interviews with the experts in order to get their opinion on the same questions. The focus on these interviews still revolved around the same four themes as previous interviews but the questions were formed in a way so that the experts could provide their opinion on how to solve the obstacles presented.
The results from the interviews are presented in chapter 4 and are divided according to each theme. Both interviews with the respondent from the company examined revolved around the same themes and are not separated due to the fact that the first interview, which was of explorative character, revolved around the themes in general and the second interview, which was an in-depth interview revolved around the same themes. In order to give the reader an easy understanding of the results the researchers chose to present data from both interviews under the same theme. The interviews with the experts are divided in order to give the chapter a better structure and therefore making it easier to read and understand.
2.4.2 Interview guide
The interview guides for all interviews can be found in chapter 9 under appendix 1, appendix 2 and appendix 3. The first interview (appendix 1) which was an explorative interview revolved around four major themes of questions, the respondents’ role at the company, use of ML and/or AI today, how the company creates business value today & obstacles and the future of ML. The themes were sent to the respondent before the interview took place in order for the respondent to prepare for the interview without knowing the questions. The
researchers opted to sort the questions into four different themes in order to give the interview structure.
The second interview (appendix 2) was an in-depth interview with the same respondent at the same company. The second interview focused more in-depth into certain areas where the researchers wanted to gain more knowledge from the previous interview in order to answer the research question. The areas that the researchers chose to focus on revolved mainly around obstacles that the company are facing and why they haven't implemented ML to a larger extent.
The third and fourth interview (appendix 3) had the same interview guide in order for the researchers to analyze and compare the answers provided by the experts. The interview guide was built upon the data gathered from the first and second interview in order to get the experts
10
point of view on the data. Furthermore, the interviews revolved around the same four themes and was structured in the same way as interview one and two. Interview three and four also provided the researchers with external opinions from two experts, broadening the research.
2.5 Data analysis
The analytical method for this thesis is based on a thematic analysis and was adapted to get a deeper understanding on the data collected on the subject of business value and ML without having its base or adaptation in existing theories (Bryman 2012).
Bryman (2012) recommends that after conducting interviews it’s important to transcribe the interview as close to the actual interview as possible so that the researchers can recall the subjects discussed. A transcript of an interview also helps to guarantee that the data found is qualitative as well as helps the researchers to find new themes that can be discussed in further interviews. In regards to the transcription of the entire interviews one thing that is worth mentioning is anonymity. Two of the respondents and the company they work for chose to be anonymous, which is why the researchers intentionally removed material that could directly be linked to either one of them during the transcription of the interviews.
Since the transcript of an interview is the main source of information from that interview it’s important to thoroughly transcribe the interview as well as give it structure (Bryman 2012).
Transcribing an interview is however time consuming so the researchers opted to split the interviews between them in order to be as efficient as possible. The researchers helped each other to understand anything that was unclear or difficult to hear from the recording in order to avoid misunderstandings.
According to Braun and Clarke (2006) coding begins when the researchers have good knowledge about the data collected. Bryman (2012) explains that coding is a method to evaluate the transcript and mark different factors that the researchers think is of value for the research. With this method the researchers were able to find different themes within the transcript so that the data collected could be analyzed and compared against the theoretical framework. The researchers chose to first analyze and find themes individually to avoid influencing each other and therefore prevent discoveries in the data. During the coding, data that was found irrelevant for the purpose of this thesis was deliberately removed. After the individual analysis the researchers conducted a second analysis together to go through the themes and highlight important data.
2.6 Ethical considerations
Research ethics are about harm to the respondent, lack of information, invasion of privacy and lastly deception (Bryman 2012). The researchers place great importance in these principles to ensure the confidentiality in the research. Not harming the respondent played a big role in the outlining of the semi-structured interviews which is why all respondent and the company they work for were given the option to be anonymous. The reason why anonymity was an option was to protect respondents, the companies and their involvement in this thesis by not naming either one of them. The respondents was also given the option to decline to answer any question in order to not disclose any information that could potentially harm the respondent, the company or any external parties linked to either one of them.
11
The second principle, lack of information means that the respondent has to be provided with the amount of information needed to decide if they want to participate in this thesis or not.
The second principle, lack of information means that the respondent has to be provided with the amount of information needed to decide if they want to participate. The respondent should also be informed, prior to the interview if certain observation techniques will be used or if the interview will be recorded (Bryman 2012). The researchers provided all respondents with the same general information about the subject, the purpose and the research questions. The experts also provided further information in regards to analyzing previously collected material. This was done in order to make sure that the respondents knew what they were participating in and how they could benefit from taking part. The respondent, the company examined and the experts expressed a positive attitude towards participating in this thesis.
When it first was confirmed that the company wanted to participate, questions and themes for the explorative interview were formed by the researchers and the themes were later sent to the respondent.
Bryman (2012) explains that the third principle, invasion of privacy has a close link to the second principle. As an example, the interviewee might not want to answer a question which often occur if the question is about their private character. To minimize the risk that the respondents didn’t want to answer a specific question the researchers chose to not include any type of question that were not work related or that could potentially harm either the
respondents or the companies.
Lastly Bryman (2012) speaks about deception. This means that the researchers show the research as something that it is not. To lie to a respondent is not professional and can harm researchers in the future. To avoid this, the researchers of this thesis have been truthful and transparent towards both the company the thesis is based around, the respondent from the company and the external experts to ensure that they are well informed about the thesis and what was intended with the information that they provided.
2.6.1 Reliability & validity
Saunders, Lewis and Thornhill (2012) describes the theory of research evolution to be about reliability and validity from an internal and external point of view. The theory illustrates if the analytical method and data collection provides reliable results if it were to be used for future purposes in the research or repeated by another researcher within other areas. Threats towards the reliability of a research that has to be taken into account when collecting data is,
participant bias and participant error. To minimize the risk of participant bias and therefore increase reliability the researchers opted to offer full anonymity to the company and the interviewee. Furthermore, the researchers also opted to include external experts in order to get an external view on the subject and a broadened perspective. The choice to include one expert from a large company that works with software for analysis and one expert that works for a small company that makes custom software solutions was done deliberately. The reasoning behind this choice was to include respondents from both parts of the spectrum in order to get as broad perspective as possible, increasing the reliability. Participant error in the semi- structured interviews was prevented by enabling all interviewees to choose the time, date and location for the interview in order to minimize the chance of them not taking part. To include three respondents with different types of expertise, background and experience also
heightened the reliability since the researchers got three, independent of each other, perspectives on the subject in question.
12
Internal validity is about a context where specific findings show an occasional relationship among two factors. As an example, if the research proves a relationship among elements from the theoretical framework and the qualitative data collected, internal validity is achieved (LeCompte & Goetz 1982). The researchers believe that internal validity was achieved in the thesis since the respondent and the experts had high credibility. Material from the respondent, the experts and the theoretical framework can all be linked to on and other. This makes not only the respondent and the experts credible but also the theories presented relevant, which increases the internal validity.
External validity is attained when a study can be generalized in a broader population (Bryman
& Bell 2011). This means that the researchers have reached the goal of external validity if other actors within the same industry can apply the results of the study. The researchers of this thesis believe that external validity was achieved. One company was examined
thoroughly and two independent experts validated, evaluated and analyzed the data collected from the company. This means that two external parties with no connection to the company shared their thoughts and opinion on the subject making for a broadened perspective.
The researchers of this thesis believe that external validity was difficult to reach as only one company within the industry were examined. However, the researchers believe that to only examine one company was a better strategy than to conduct research on multiple actors. To only examine one company gave the researchers the opportunity to not only have a highly credible respondent but also made it possible to conduct two interviews. One explorative interview which provided the researchers with a broad perspective on the subject, as well as an in-depth interview that provided the researchers a deeper understanding of the company’s views and ideas towards technology development.
2.7 Limitations
2.7.1 Primary data
As mentioned, a semi-structured interview method was chosen in order to let the respondents answer questions in a slightly more open way. However, one can argue that a limitation with this type of collection technique is that the outcome and effectiveness relies on verbal level as well as skills of the respondent. In order to diminish this limitation and by doing so increasing the quality of the interview, the researchers developed and performed the questions as clear as possible. Even if the interviews were outlined on a structured base of questions, the
researchers encouraged the participant to answer freely and rephrase questions if they could not understand a specific question. One could argue as well that the questions include bias and subjectivity as the questions were entirely developed by the researchers.
Due to the limitations of time and resources, both from the researchers, the company
examined and the experts the researchers were confident that four separate interviews would provide enough data to cover the subject of the matter. This could be perceived as a
limitation, as one could argue that it isn’t enough interview data to cover the subject. The researchers however argue that the knowledge, expertise, experience and credibility of the respondents is satisfying the needs of the thesis which is why the researchers argue that the amount of data collected is fulfilling the purpose of the thesis. Due to the fact that there is a clear relationship between the theoretical framework, the data collected and the analytical framework confirms the researchers approach.
13
Anonymity can be perceived as a limitation since only the researchers know the identity of the respondent and the company examined. The researchers however, argue that preserving the confidentiality of the respondents makes the thesis more credible than exposing the respondents which potentially could harm not only the respondents, but the organizations in the long run. In the writing of this thesis the researchers made it clear from the very start that confidentiality and anonymity was an option for the respondents and the company.
2.7.2 Secondary data
The secondary data collected in this thesis ranges from quite large timespans. Some of the sources used dates back quite a long time and one could argue that this could be seen as a limitation as it could be slightly or heavily outdated in terms of findings and statistics.
Therefore, were all sources carefully evaluated before used in the thesis to ensure that they were relevant.
14
3.0 Theoretical framework
The third chapter of this thesis is the theoretical framework. In this chapter previous research and theory that is relevant for the research question is explained.
Figure 1 displays the theories in order from a broad perspective to an increasingly narrower perspective. In order for the researchers to give an understanding of how ML can create business value the broader theories (Socio technical perspective & Industry 4.0) have to be incorporated in order to narrow the scope down to ML, business value and organizational support systems.
In order for the researchers to give a nuanced and broad description of ML and how it can be used in order to create business value a number of theories are incorporated and related to one and other. The broader theories such as the socio technical perspective and industry 4.0 are incorporated in the theory in order to provide a perspective of the technological side of ML as well as an organizational perspective. The research that the researchers have conducted indicates that these theories are becoming increasingly important for the organizations and the relationship between the organizational and technological sides of it. Furthermore, ML is introduced as a narrower perspective in order to provide the reader with the researchers’
definition of what ML is. ML is also directly connected to the purpose of the thesis. Lastly business value and organizational support systems are introduced as the final theories
presented in order to narrow down the broader scope even more. Business value is explained as it is directly connected to the purpose of the thesis. Organizational support systems is presented as the subject of the matter requires the reader to understand its importance and also the framework that the analysis is based on.
15
3.1 Socio technical perspective
Socio technical is the interrelatedness of social and technical aspects of an organization. A socio technical system is an open system that manages the fine lines between the system and the environment internally as well as externally in and outside the organization (Botla &
Kondur 2018).
“Socio-Technical Systems Design (STSD) methods are an approach to design that consider human, social and organizational factors, as well as technical factors in the design of organizational systems.” (Baxter & Sommerville 2010, p. 4). When developed and
implemented correctly should eventually result in better understanding of how human, social and organizational factors affects the work itself, how it’s done and the usage of the technical system (Baxter & Sommerville 2010).
Socio technical perspective is the perspective of organization and technology merging together creating one unit. When developing a system, the developers should have the
organization at large in mind as well as the individuals within the organization. Increased job satisfaction and higher productivity from employees could be achieved by having employees in mind when developing the system. The end users must participate in the design and
implementation of systems. Bansler (1989) argues that the end users are the ones that possess knowledge from a wider perspective meaning that they are the ones that know the
organizational structure but most importantly the actual work processes (Bansler 1989).
Baxter and Sommerville (2010) lists five key categories of socio technical systems:
• “Systems should have interdependent parts.
• Systems should adapt to and pursue goals in external environments.
• Systems have an internal environment compromising, separate but interdependent technical and social subsystems.
• Systems have equifinality. In other words, systems goals can be achieved by more than one means. This implies that there are design choices to be made during system development.
• System performance relies on the joint of optimization on the technical and social subsystems. Focusing on one of these systems to the exclusion of the other is likely to lead to degraded system performance and utility” (Baxter & Sommerville 2010, p. 5).
To achieve legitimacy in ML systems requires embracing a socio technical view, according to Selbst, Boyd, Friedler, Venkatasubramanian and Vertesi (2018). Technical actors must shift from seeking a solution to grappling with different frameworks that provide guidance in identifying, articulating, responding to tensions, uncertainties and conflicts inherent in socio technical systems. It might be the case that the system can’t provide definitive solutions although it will help in enabling process and order.
According to Selbst et al. (2018) it is equally important to understand the behaviors of hiring managers and job candidates when using an automated resume screen as it is to understand the role of the software, the most important insight is to understand that the social aspects have to be considered alongside the technical aspect in any design enterprise. Different social environments, actors and social groups shape the different kinds of technology that eventually become successful.
16
According to Buhr (2015) social and technological factors will impact businesses, both factors will arise simultaneously and create new ways of working and developing businesses.
The real world and the virtual world will, to some extent, merge together. By linking people, objects, systems and production facilities together will create a more dynamic and self- organizing business.
Law (2011) means that we must be able to think about three things simultaneously. We need to find a way of thinking about the social, technological and the natural. Although these three aspects are usually seen as three individual units they should be looked at as one unit. He argues that in the event of a catastrophic collapse these three individual units are all
intertwined and linked to each other, which is the reason why they should be seen as one unit.
Law (2011) suggests that looking at a disaster, whatever it may be should be looked at as a failure of a heterogeneous system. Whatever has gone wrong represents a tangled and
complex network of technical relations, social relations and natural relations. He argues that it doesn’t help to separate them because the technical relations are at the same time as being technical relations social relations as well as natural relations. “They are social relations because they unavoidably reflect and embed organizational decision making. And they are natural relations because they reflect and embed the natural properties of materials and their interactions.” (Law 2011, p. 5).
Selbst et al. (2018) confirms what Law said and states that heterogeneity refers to the
requirement that we think simultaneously about what different technical parts of the apparatus will do, and what the humans that operate, live alongside and otherwise contribute to them will do as well. It isn’t possible to engineer human decisions, what is possible however is to engineer so that boundaries of abstraction to include people and social systems both from a micro and macro perspective, meaning both at a local level as well as a global level. Working with many different elements, that is the socio technical puzzle allows developers to design and model fair outcomes as properties of systems that are well thought out from a socio technical perspective (Selbst et al. 2018).
3.2 Industry 4.0
The part of the economy that produces material goods is called an industry. Industries have been a part of the economy for a very long time and the differences between the different times are how material goods are produced. The prior industrial revolutions were all triggered by technical innovations. The first industrial revolution came when manufacturing was developed with water and steam-powered production at the end of the 18th century, the second industrial revolution was based on the division of labor at the beginning of the 20th century and the third revolution came in the 1970’s when programmable logic controllers were introduced to automation purposes (Brettel, Friedrichsen, Keller & Rosenberg 2014).
The combination of internet technologies and future-oriented technologies of machines and products seem to result in a fourth industrial revolution, according to Lasi, Fettke, Kemper, Feld and Hoffman (2014).
Lasi et al. (2014) describes the term Industry 4.0 from two development directions which are application-pull and the technology-push. Revolutions that the world have gone through in the past are due to social, economic and political changes. Lasi et al. (2014) describes some of the changes specifically intertwined with application pull such as shorter development periods, flexibility, decentralization and resource efficiency.
17
The technology-push has already been developed to a large extent in the private sectors. Most of the people in the world are dependent on applications, smartphones, laptops etc. However, the push isn’t widespread in the context of industry therefore a few approaches to a
technology-push in what Lasi et al. (2014) describes as job-related areas, specifically in industries of production. A few examples that are mentioned are increasing mechanization and automation, digitalization and networking, new systems in distribution and procurement, adaptation to human needs and Corporate Social Responsibility (CSR) (Lasi et al. 2014).
According to Lasi et al. (2014), the ideas and approaches to industry 4.0 are developed in the context of electrical engineering, business administration and computer science to name a few. Business and Information Systems Engineering (BISE) have always been a central part of industries. The focus of BISE is certainly applied on other branches than just industry although the integration of information systems and the modelling and design play a huge role in industries.
Brettel et al. (2014) refers to the term Cyber-Physical-Systems (CPS) which is a system that allows communications between humans and machines throughout large networks. A factory of the future will be referred to as a smart factory where CPS will enable communications between humans, machines and products. The factories of the future will be developed with a socio technical perspective in mind in order to bring those different parts into one unit.
Factories have to cope with rapid product development and flexible production which confirms what Lasi et al. (2014) describes above.
The Socio Technical Systems (STS) theory considers both technical and social factors in the search of change in an organization, whether it’s change in the form of new technologies being introduced or organizational change. In the early stages of STS theory it was used to discover and revel dysfunctions between the social factors (the people working in the systems) and the technological factors (what the purpose of the system was). Obviously this has gained future research and development of new technological systems. As mentioned previously the end users need to be incorporated in some form or fashion in the development stages in order to be as efficient as possible (Imran & Kantola 2018).
The interaction between organization and technology is vital to understand. A holistic approach demands that technology isn't only seen as an artefact, it's rather seen as an
embedded feature in organizational and human behavior. As well as the other way around, an organization doesn't solely consist of the people working it's shaped and customized through the use of technology (Bygstad, Nielsen & Munkvold 2005).
Organizations are large complex systems with multiple moving parts that need consideration in the event of modification or introduction of new technologies. To make changes or
modifications in an organization without thinking of the parts that aren’t directly affected could result in an ineffective change. Organizations need to adopt a holistic approach, covering all fields in the event of change in order for all the parts to work together. The approach need to be emphasized specifically on the social and technical factors (Imran &
Kantola 2018).
According to Imran and Kantola (2018) all industrial revolutions introduced a new technology which forced change in organizations and by employees. The fourth industrial revolution is no difference, although it seems to emphasize more on the employees in organizations. The
18
fourth industrial revolution is introducing new ways of working such as decoupling of work and time which will force flexibility and new competences into organizations.
3.3 Machine Learning
Machine learning means to create algorithms in a formal computer language that the computer can understand and execute mechanically. The algorithms are created to solve problems in different ways and in that way detect patterns in large amounts of data. These algorithms aren’t developed to copy human behavior although they can learn to some extent during the ongoing process according to Marton, Hounsell and Entwistle (2005).
First of all, we should establish what learning is. According to Carbonell, Michalski and Mitchell (1983) there are two basic forms of learning in a human sense: knowledge acquisition and skill refinement. The definition of knowledge acquisition is “learning new symbolic information coupled with the ability to apply that information in an effective manner” (Carbonell, Michalski & Mitchell 1983, p.6). Skill refinement is described as “the gradual improvement of motor and cognitive skills through practice” (Carbonell, Michalski &
Mitchell 1983, p.7).
According to Chui, Kamalnath and McCarthy (2019) ML is used to provide predictions and prescriptions on large amounts of data. There are three types of analytics that are used in companies globally ranging from less complex to increasing complexity. The three types of analytics are; descriptive, predictive and prescriptive.
• Descriptive analytics describes what occurred and is employed heavily throughout all businesses globally.
• Predictive analytics predicts what will happen and is employed in data-driven organizations as a major contributing source of insight.
• Prescriptive analytics provide recommendations on what to do to achieve certain specific goals, these analytics are employed heavily by Internet companies.
A definition of learning in a technological sense is “learning is any change in a system that allows it to perform better the second time on reputation of the same task or on another task drawn from the same population.” (Simon 1983, p. 28).
According to Nilsson (1998) ML refers to the changes in a system that performs task associated with AI. Nilsson names recognition, diagnosis, planning and prediction as a few examples of tasks that can be performed by using ML.
Flexibility seems to be a keyword when discussing Industry 4.0 and machine learning. Digital technologies and cognitive computing are changing the manufacturing industries. One of the main factors that’s changing manufacturing industries is the usage of self-learning solutions in which machine learning is a key factor. Cognitive computing aims at reproducing human skills through artificial models and computable algorithms and in the long run transferring human decision-making processes to intelligent machines (Ansari, Erol & Sihn 2018).
Ansari et al. (2018) describes the factories of the future with mutual human-machine learning, which is a socio technical perspective on Industry 4.0. Ansari et al. (2018) has identified two groups of learners, human and machine. Mutual learning describes the interaction between the two groups in context of Industry 4.0. Quality and performance when executing a task are the
19
two key indicators identified which distinguishes the differences in capability by human and machine on the performance of the specific task.
Mutual learning is described as “a bidirectional process involving reciprocal exchange, dependence, action or influence within human and machine collaboration, which results in creating new meaning or concept, enriching the existing ones or improving skills and abilities in association with each group of learners” (Ansari, Erol, Sihn 2018, p. 119). What Ansari et al. (2018) describes is the interaction between two key factors in an organization which is what Industry 4.0 describes. The interaction between humans and machines will be a key factor when entering the fourth industrial revolution as well as having a socio technical perspective in mind.
According to Ansari et al. (2018) three task pools are a central part of redefining job roles, human-specific, machine specific and shared tasks. Considering the fact that flexibility is a keyword the three task pools identified are a central role. Being able to divide and/or intertwine work between the groups of learners is a key factor when moving into the fourth industrial revolution.
3.4 Business value
AI and ML have today a big impact on business automation as it aids to create value for a business. The technology is used in many different areas, from managing transport loads to selecting loan applicants without human interaction (Canhoto & Clear 2019). In the authors report they write about a chatbot as an example of how ML can help to create business value and not have employees interact with customers directly. However, to get the desired effect of a chatbot the algorithm needs to have knowledge about different areas to meet customer needs. The chatbot needs to be able to process free-form text, understand natural language, identify customer wishes, understand if a customer is upset or not and decide how to fulfill customer needs. There are many different areas where a chatbot needs knowledge and is therefore also difficult to implement successfully. However, if done successfully it will help the company to create value even though it’s a challenging process to get the program to work correctly (Canhoto & Clear 2019).
To create business value from an organization’s data a few steps are required to ensure that the data presented is trustworthy. According to Dean (2014), first the data that have been collected needs to be manage and prepared so that it can be analyzed. Normally organizations have several people or teams across different departments within the organization that share the task of data mining (finding patterns in data) in order to analyze the data in an efficient way.
With the help of different methodology, algorithms and approaches organizations can use data mining to extract value from the data and by doing so create business value for the
organization. Furthermore, creating business value from existing data has the potential to save money for the organization but also an increase of income for the organization (Dean 2014).
In the report by Sharam, Krishnan and Grewal (2001) the authors discuss how resource allocation to a specific customer segment can help a company to create value. They explain that with the help of technology delivery, product delivery and customer delivery processes value can be created and therefore improve the business. They also speak about, Value-based pricing and Value-based communications, as different methods that can be used by managers
20
to reach their selected target group. The findings in this report is that a combination of all these different processes gives the company the value that they need but if not priced and communicated effectively they don’t do so much for the company. Therefore, for a company’s survival it’s important to learn how to use technology processes to improve customer value and therefore have a sustainable business model.
Furthermore, other research is also showing that a combination of different aspects within data needs to be considered to create value. Grover, Chiang, Liang and Zhang (2018) writes in their report Creating Strategic Business Value from Big Data Analytics: A Researchers
Framework that to create value from data, a business needs to consider different aspects to make the implementation successful. Their framework for value creation is based around five different aspects based on Big Data Analytics (BDA) as a mean to help companies create value from their data. Within these five different areas there are challenges and obstacles that a company have to take into account or overcome before a successful implementation can be completed. Below are the five aspects listed:
• Data Quality – There are numerous data quality challenges that has to be considered for an organization such as, incorrect or lost data, as data is growing it’s difficult to ensure that its trustworthy, unlabeled data and imbalance in data.
• Data Integration – The diversity in data, both structured and unstructured regulates the complexity of data integration particularly when data comes from external sources.
It is therefore important to create a data management design that is capable of
supporting variety in data so that it can be delivered on-demand, real time data and be accessed quickly when needed.
• Data Security – Data must be secured towards different cyber threats to avoid data breaches. A company must comply with data security regulations but the problem is when data infrastructure gets bigger so must also the security mechanisms that traditionally is formed for small-scale data centers.
• Analytics – Organization that focuses on making existing processes more efficient or autonomous to cut cost often have the lowest analytical competence as they lack vital parts such as humans, activities and instruments to gather or understand their data.
• Human talent – The main challenge in developing BDA. Finding the correct person and to get the organization to support it has proven to be much more difficult than different technological aspects (Grover et. al 2018).
In concussion, the authors found that it is not enough to only invest in data infrastructure and analytic technologies but it also requires skilled people who understand what technology to use. Investing and building BDA to create business value for an organization is a real
challenge as it includes developing data and analytical skills. However, if done successfully it can bring much value to an organization (Grover et. al 2018).
To conclude what business value is or how its created is difficult as no real definition exist.
According to Marinova, Larimo and Nummela (2017) even value is difficult to give an exact definition to because it means different things to different people. The authors write in their book Value Creation in International Business “Somehow, it seems we know what “value”
means, but if we try to use it in different processes and context in relation to diverse actors, we might be surprised by the various interpretations given to it” (Marinova, Larimo &
Nummela 2017, p. 1). In form of creating business value from data is only one aspect of how an organization can create business value and as it is hard to give an exact definition to what value is, it’s difficult to summarize how to create business value from data. However, one
