Today’s most precious raw material is not gold, but Big Data: Each one of us generates a huge amount of information every single day, rendering thus both ourselves and our choices transparent. But in addition to that, Big Data helps companies to improve their decision-making.
Since managers have to address highly complex issues in an ever more complicated world, they cannot do without Big Data and Artificial Intelligence, as Carolin Nothof explains. By taking into account various external factors, their algorithms predict right entrepreneurial choices.
These choices can be made in areas such as retail, Human Resources, the Internet of Things, and marketing. Nothof’s publication is not only rich in theoretical explanations, but also gives examples of the practical use of Big Data in various industries. Machines are a man’s best co-workers.
In this book:
- Big Data;
- decision-making;
- AI;
- Behavorial Economics;
- Machine Learning;
- algorithms
Table of Contents
Abstract
List of Figures
1 Introduction
1.1 Problem and relevance of the subject
1.2 Objectives
1.3 Structure
2 Theoretical principles
2.1 Classification of the term decision-making
2.2 Classification of the term Big Data
2.3 Big Data’s role in decision-making
3 Analysis: Big Data in practice
3.1 Structure of the analysis
3.2 Big Data-decision use cases in different industries
3.3 Artificial data-based decision makers
3.4 Parallels between human and artificial decision makers
3.5 Might managers be replaced by Big Data?
4 Closing remarks
4.1 Summary
4.2 Result: How changes Big Data managerial decision making?
4.3 Challenges
4.4 Trends and developments
Bibliography
Abstract
Big Data is one of the buzzwords of the last years when talking about digitization. Data is being generated anywhere, anytime and at an increasing scale. This amount of data can be valuable for companies in many ways if they are analyzed properly.
People do not just generate data, but they also have access to many data sources. Besides, more and more people have the chance to enjoy good, long education and collect valuable experiences during the course of their lives. Nonetheless, they face an apparent insuperable obstacle when it comes to decision making. Due to cognitive limitations, humans are tricked by their own mind and behave irrationally, which makes them lose quality in their decision processes, even in professional contexts.
The purpose of this thesis is to find out whether Big Data can improve managerial decision making. Therefore has been analyzed how Big Data is used in companies, how artificial intelligent decision makers function and if they rationalize similar to human decision makers.
List of Figures
Figure 1: Development of Google-inquiries for Big Data since 2011
Figure 2: Structure of the Bachelor Thesis
Figure 3: Simon's decision phases
Figure 4: Two Thinking Modes
Figure 5: The RAPID decision model
Figure 6: Moore's Law
Figure 7: Data Warehouse
Figure 8: Machine Learning
Figure 9: Data Mining
Figure 10: Proposal for an Organization of Big Data related topics
Figure 11: Development of data-driven decision making
Figure 12: Variation of the term analytics
Figure 13: Decision model including Big Data, Business Intelligence and Decision Support Systems
Figure 14: Model of Deep Blue's Functioning
Figure 15: Model of Watson's Functioning
Figure 16: Functioning of DeepMind
Figure 17: Distribution of strategic decision rights
Figure 18: Key strategic decisions
Figure 19: Limitations of strategic decision-making
Figure 20: Data-driven decision-making
Figure 21: Kind of analytics used
Figure 22: What strategic decisions rely on
Figure 23: Data-based decision-making process
1 Introduction
1.1 Problem and relevance of the subject
Every day an incredible amount of data is created, without people making any effort: a switched on mobile phone or any other device connected to the internet is sufficient to transmit data streams. In this context, Erik Brynjolfsson, economist and director of the Massachusetts Institute of Technology’s (MIT) Center for Digital Business, said:
“We’re rapidly entering a world where everything can be monitored and measured. But the big problem is going to be the ability of humans to use, analyze and make sense of the data.”1
The current term describing this vast amount of unstructured data that needs to be analyzed in order to be valuable for organizations is called Big Data. The prominence of this term has increased considerably during the last years. By investigating on Google, one of the most famous examples of a business making successful use of data, it becomes evident that the term Big Data started gaining attention in the public sphere about five years ago and the term did not vanish from the public scene until today.
Abbildung in dieser Leseprobe nicht enthalten
Figure 1: Development of Google-inquiries for Big Data since 2011
Source: Google Trends (2016), online source.
Today, the topic is very present in public discussions: While its critics are mostly afraid of privacy and data protection issues, its supporters praise Big Data as a wonder weapon. Since data is the raw material for information and knowledge it has effectively the potential to represent a high value for any organization - on condition that it can be analyzed properly. Getting to know the customer better, optimizing organizational processes and products or making better decisions - the list of where Big Data can be valuable or useful is long. Several companies have already started implementing Big Data strategies and others will surely follow, since it appears to be a competitive advantage. At the same time, research does not stop generating possibilities and developing new instruments: Currently Big Data is being connected with Artificial Intelligence (AI) and with self-learning algorithms in order to create smart supporters in professional and private life. The common reputation of Big Data and AI represented in films is not necessarily a very positive one: In the recent Tatort episode HAL2 appears a program able to make predictions on consumer behavior thanks to Big Data analytics.3 The software develops omnipresence by monitoring everything what is going on everywhere and it takes away any form of privacy. Since it is programmed with the objective to survive, it does everything to protect itself, and even manipulates data to achieve its goals. Another recent film dealing with an intelligent, artificial being getting out of its inventors control is Morgan. Interestingly it is IBM’s computer Watson which edited the trailer for this film after having analyzed 100 other trailers of horror films and their characteristics.4
On the one hand, there is this negative science-fiction marked reputation and on the one hand the fear that computers gaining human capabilities will take away jobs. Both aspects neither support the development nor the application of AI. However, reality is drawing another picture so far, as will be shown in the further course of this work.
Decision making, the other big topic in this work, has a more traditional and larger research history since it is one of the most important activities of any manager. But even though a lot has been said about it, people continue to make bad choices, even in professional contexts. Studies reveal that an effective introduction of data into decision processes makes companies more profitable5. Operational decisions need to be made day by day and strategic decisions such as production, cost or restructuring decisions need to be made regularly in order to steer a company. But what does a manager need in order to face this challenge? It is common knowledge that the most traditional ways of making decisions based on a manager’s experience, and the famous gut-feeling, often are not the most rational or the best ones. That is why some sort of decision-support systems have already been introduced into management departments in the 1950. However, it must be pointed that decision making today and 70 years ago does not refer to the same problem as decision processes today are more complex due to the large amounts of structured and unstructured data from various sources that a company needs to deal with every day. That is where Big Data comes into play.
1.2 Objectives
Bearing this in mind, there is a problem to solve: bad or inefficient decision making and a potential solution: Big Data, or how to leverage it. The objective of this thesis is to compare both topics in order to understand if and how the phenomenon of Big Data can improve organizational decision making. This means that the thesis aims to answer what Big Data is currently used for, what areas it is connected with and what further developments and applications can be expected in the future. As the title suggests, another objective is to find out how Big Data influences decision making and what changes this means for the managers of today, and the ones of tomorrow.
The intention of this paper is to incorporate scientific literature and non-scientific sources considering Big Data, to extract the sometimes conflicting points of view and to check the potential for decision making improvement. It is important for companies to deal with this subject in order to stay competitive and so is the scientific significance of this topic. Insights from various sources will be collected and compared with current empirical results.
1.3 Structure
This thesis is divided into four major parts. First of all, the topic and its relevance, as well as the objectives and the structure of the work are presented as an introduction.
The second part is about the theoretical principles underlying the work. In order to find answers to the problems presented earlier, it must be first understood how human beings actually make decisions today, why they sometimes make irrational choices and what strategies are used today to overcome this problem. The term Big Data is also presented and it is put into context with related topics in order to understand the whole framework. A short historical review on data analytics systems follows afterwards. This part is then closed by converging both topics, Decision Making and Big Data, by presenting a solution on how to integrate both.
The third chapter gives a series of use cases that illustrate how Big Data is already successfully being implemented in different areas or industries. As noted before, the connection between Artificial Intelligence and Big Data will also be examined through three artificial data based decision makers. It is analyzed how these computer programs function and if there are parallels that can be made in regards to how humans make decisions. If so, it would be important to know if these parallels are sufficient to replace managers in the future: This is what the last part of this chapter is about. Two surveys dealing with the same questions are presented and the importance of Big Data is weighed up against human intuition in a complex world. This chapter concludes with a differentiation between decisions Big Data is useful for and others where it might be more difficult to introduce. Several behaviours managers need to adopt for a successful usage of Big Data will be derived from those insights.
The fourth, and final, chapter will summarize the findings of the work and present the results. The challenges that companies need to be aware of will be pointed out as well as a hypothetic outlook at the development of the entire Big Data subject in decision making. Figure 1 gives a graphic overview of the thesis.
Abbildung in dieser Leseprobe nicht enthalten
Figure 2: Structure of the Bachelor Thesis
Source: Own representation.
2 Theoretical principles
2.1 Classification of the term decision-making
2.1.1 Definition
Unlike Big Data, decision making has been a topic human beings have always dealt with, since it is directly associated with a person’s independence, life planning, and even survival. From an economical point of view, the crucial actions a decision should involve influence the performance of every organization. Decision making signifies
“the thought process of selecting a logicalchoice from the available options. When trying to make a good decision, a person must weigh the positives and negatives of each option, and consider all the alternatives. For effective decision making, a person must be able to forecast the outcome of each option as well, and based on all these items, determine which option is the best for that particular situation.”6
Filtering out the key characteristics of this definition, the following attributes remain. They will be explained or discussed in the further course of this work.
- Process à See Simon's decision process and Sjöberg’s (p.7).
- Logical à See Sjöberg’s decision rules (p.8) and Decision Traps (p.10).
- Ends with a choice à See Simon’s decision process (p.7).
- About forecasting à See anchoring (p.10) and forecasting trap (p.11) and Big Data’s role in decision making (p.24).
- Individual according to the respective situation à See decision types identified by Kepner and Tregoe (p.5).
2.1.2 Explanation
Decision types (Kepner/Tregoe)
In order to determine how to make a good decision, first it must be considered that there are fundamentally differing types. Decisions vary greatly from one to another and the degrees of difficulty depend on the particular situation. The two researchers on rational decision making and problem solving, Kepner and Tregoe, identified four types of decisions:
- Complex decisions demand the screening of a lot of information and the assessment of this information. It includes problem solving and strategic planning.
-“Yes/No” decisions only allow two alternatives.
- Examinations check if the proposed alternative is good enough or if the team should develop a new one.
- Routine decisions such as employing new staff, ordering of technical equipment, development of human resource policies and other daily decisions such as resource allocation.7
Two decision types (Simon)
Herbert Simon, who won a Nobel Prize in economics “for his pioneering research into the decision-making process within economic organizations”8, also distinguished between different decision types. The two main segments for him are “programmed” decisions, respectively routine and repetitive decisions on the one hand versus “non-programmed” decisions referring to policy and strategic decisions on the other hand.
“Modern” vs. traditional techniques (Simon)
Simon also distinguished how decision were made in the past and at his time (in 1960) and already discerned that there were fundamentally differing techniques for both routine and complex decisions. The traditional way for making routine decisions was habit or clerical routine whereas the modern ways, at this time, were mathematical analysis, computer simulation via Operations Research (a topic which will be dealt with later on) and electronic data processing.
The non-programmed decisions, however, were traditionally made by intuition and creativity, rules of thumb and training of executives whereas the “modern” approach were heuristic problem-solving techniques human decision makers were trained with and the construction of heuristic computer programs9 - this is a technique that offers a fast solution to the problem, with an imperfect but mostly sufficient outcome. In the development of decision support methods a row of further techniques and systems appeared which will be depicted later on.
As demonstrated, all decisions are not equal, but still most of these managerial activities involve several equal components which will be defined in the following as the decision-making process. One of the most widely accepted decision making processes was first introduced by Simon, who also grasped the importance of the subject. According to him, decision making is just a synonym for management, and therefore has the same powerful meaning.
Decision process (Simon)
Simon’s model of the decision-making process distinguishes between three phases. In the first place, a situation occurs which calls for a decision. This involves that the problem must be defined which leads to a decision statement. Afterwards, a set of alternatives will be developed by considering a large set of information. Eventually, the alternatives are evaluated and one is chosen. The three phases are not necessarily seperated but may blur into one another.10 This model is still valid today and can be implement for Big Data strategies, as will be seen.
Abbildung in dieser Leseprobe nicht enthalten
Figure 3: Simon's decision phases
Source: cf. Simon, H. (1960), p.1-2.
How do humans learn to make decisions? (Simon)
By having the information of how to develop a solid decision-making process, unfortunately, it is not done. It requires steady learning from mistakes or bad decisions in order to integrate proper decision processes. According to Simon the skill of being a good decision maker can be compared with the skill of being an athlete. The latter one is born with biological potential for athletic endowment which he develops through training and experience into a mature skill. The former one is as well born with the natural gift of some intelligence and the ability to interact with their fellows. It is likewise through learning, practice and experience that this skill will be developed.
However, only by gathering experience a person does not gain automatically knowledge as pointed out by Russo and Schoemaker in their book Decision Traps: People make experiences, through which they collect data, or rather, the necessary raw material to build knowledge. Nonetheless, even intelligent or highly motivated people do not learn automatically from experiences - in fact, it requires profound skills and training to turn gathered data into knowledge.11
Decision process (Sjöberg)
In his book on human decision making, the Swedish professor of psychology Lennart Sjöberg detects steps in the decision process similar to the ones of Simon: Everything starts with a person being in an unsatisfied situation that he does not want to continue (=Intelligence). In a second step, the decision maker would generate different alternatives to his problem and would afterwards do an evaluation of those alternatives as to analyze the best one (=Design). This selected best solution will be considered for an implementation. After the alternative being tested in the decision makers environment, the consequences will become clear over time (=Choice).12 In the first place, the setting of a goal is crucial to the execution of a proper decision process. The operational tools this process is driven by are cognitive simulators, some of which are automatic without the decision maker’s perception and others which are consciously made, demanding an active consideration.13
Decision rules (Sjöberg)
During the step of evaluating the different alternatives, certain decision rules are considered by a rational decision maker. In most cases the rule which applies first is the rule of dominance: the chosen alternative should be better than the other alternatives on at least one attribute and be equal on all the other attributes. However, even though there is a row of rules, not all of them are applicable for every decision and the order in which decision rules are considered for application may vary according to the decision’s and the alternative’s characteristics.14 At this point an endless list of decision rules which should be implemented whenever making a well thought-out decision could follow, but it will not. The reason is that in reality even professional decision makers in reality struggle with the following rules. There is a phenomenon limiting the quality of decisions on an alarming scale and which must be considered first whenever constructing a decision rule: irrational behavior.
2.1.3 Cognitive limitations to rational decision making
System One and System Two
Making decisions is the most important, toughest and riskiest job of an executive: the decision outcome may influence not only his own career but also the company’s well-being or even survival. So where do bad decisions come from? It is possible that the decision process might lack professionalism: all the alternatives have not been correctly identified, irrelevant data has been analyzed or costs and benefits have not been weighed up against each other accurately. Nonetheless, even if the process and external conditions are optimal, there is still a high chance that humans do not make the best possible choice because of their mind. Even though they are convinced of having everything under control, there is a great part of their mind not controllable. Human minds use unconscious routines in order to respond fast to complex problems: Heuristics. These routines are usually good servants since they allow the reduction of complexity and deliver quick and mostly helpful outcomes. However, they have a characteristic which makes them dangerous: their invisibility. Heuristics are firmly established in our thinking process and therefore are very hard to recognize.15
The idea of cognitive biases and their impact on decision making was first published by Daniel Kahnemann and Amos Tversky in 1974. Their theory explains why people are incapable of recognizing their own biases: Cognitive scientists identified two modes of thinking, namely intuitive and reflective thinking, also known as System One and System Two. The former allows people to easily perform in a world full of different impressions, making them walk, avoiding obstacles, talking - all at the same time without any special difficulty. The latter becomes active when people consciously are thinking about how to solve a calculation task, to complete a tax form or how to make an important decision.
Abbildung in dieser Leseprobe nicht enthalten
Figure 4: Two Thinking Modes
cf. Kahnemann, D., Lovallo, D., Sibony, O. (2011), p. 52.
System One is able to combine a variety of inputs such as external information, experiences, associations and goals to a (mostly accurate) context. If this picture, however, is not that accurate and people make cognitive mistakes, they have no clue about it. Then again, the mistakes people make while using System Two, can easily be identified: the result of a calculation may be difficult to find, but the examination is quick. The fact that people usually cannot see their mistakes from System One makes them accept the intuitive thinking.16 What are the results of this phenomenon of the divided mind and the invisible acting System One?
Linked to the heuristics which the intuitive thinking automatically performs, a row of traps appear, harming the quality of decision processes and outcomes. In fact, according to a McKinsey study from 2011 examining 1.048 major investment decisions, companies can actually achieve a return on investment seven percentage points higher by reducing bias in their decision-making processes.17
In order to find a sustainable way to identify and avoid human bias, it is necessary to first investigate what kind of decision traps do exist.
Decision traps
A series of these traps, particularly the ones most likely to undermine business decisions, are outlined below.
The Anchoring Trap influences people’s decisions by something they have heard, read or seen before. In business for example, a common anchor is a past event or trend: when estimating sales volumes for the coming year, marketers use to have a look at the same key figure from the year before and afterwards adjust it according to the corresponding factors. By having seen them before, the ancient numbers become anchors and may influence the forecast too much compared to other factors and consequently lead to poor choices.
The Status-Quo Trap also reduces people’s rational decisions because they find the status quo comfortable and do not pull themselves up to change the situation in order to favor the outcome situation. In business, this trap can occur particularly strongly because bad outcomes of wrong actions tend to be punished more severely than bad outcomes resulting from inactivity or omission. Mergers, for example, often fail because the responsible managers of the acquiring company do not take actions to implement their management values and structure in the acquired company. They want to “wait until the situation becomes more stable”, but in fact, it becomes much more complicated to change entrenched structures afterwards.
The Sunk-Cost Trap makes people justify their past choices even though they are not valid anymore. Again, especially in business environment, this trap appears to be even more frequent, because people are unwilling, consciously or not, to reveal their apparent poor judgment in public. They close their eyes to reality and continue throwing good money after bad in the vain hope that they can transform their decision outcomes into successes in the end.
The Confirming-Evidence Trap leads people to focus their search for information only on data supporting their already existing opinion and avoiding data which contradicts their point of view. This bias will not only influence where the decision maker seeks for evidence, but also how they interpret it and how they weigh supporting and conflicting information up against each other. The two psychological forces having an effect here are the people’s tendency to subconsciously decide before even knowing why they do so and that humans, by nature, prefer engaging with things they like rather than with those they do not like.
The Framing Trap can influence a decision outcome considerably just by asking the question or by presenting the problem in a certain manner. This means, that initial frames must not be accepted automatically but that problems always must be reframed in various ways in order to identify possible distortions.
Estimating and Forecasting Traps concern humans’ difficulty to judge probabilities appropriately because of the missing feedback about their estimations accuracy. On the one hand it can occur that managers are overconfident about their accuracy and waste a lot of money because they did not consider the risk of market failure, for example. On the other hand it can also happen, that managers, contrary to the earlier trap, are overcautious and their extremely prudent choices hinder the company to fulfill its potential.18
Inconsistency of decision making
As a result of all these decision traps, people make inconsistent decisions: the same problem is solved differently one day than it would have been solved another day by the same person and the decision outcome depends considerably on who is the decision maker. Even though professionals claim to follow decision rules, there are several studies demonstrating that they contradict their prior judgments when confronted with the same data again at another occasion.19
Overcoming the barriers to learning
It becomes evident, that humans are not the “Homo oeconomicus” often described in economy books. They behave irrationally and cannot do anything about it. Experience is widely accepted as one of the aspects that make a manager a good decision maker. To make this true, managers need not only to learn from their experience as said Sjöberg, but they need also to implement decision models that help detecting possible decision traps. In the first place it will be described how managers can learn from experience, or more exactly, how they can overcome their barriers to learning and afterwards it will be presented how to implement decision models that help avoiding the traps. According to Russo and Schoemaker, there are principally three ways to overcome barriers:
- Successful decision outcomes must not simply be considered as the result of the one’s skills since it (partly) also results from other sources - it must be evaluated exactly which of one’s actions contributed to the success, ideally considering feedback provided by the colleagues.
- The same rule applies for failures: they should not be rationalized as the result of bad luck, but the true reasons for the bad decision outcome must be analyzed.
- The two professors also point out the hindsight bias. Decision makers must record their expectations after having made the decision. Comparing these expectations to the decision’s real outcome may reveal lessons to learn, since this method prevents that expectations are adjusted unconsciously afterwards.20
It must also be considered that a decision outcome does not always reflect the quality of the decision process. When passing a judgment on a decision, people must set the focus on the process and not on the outcome as most human beings do. Psychologists call this the outcome bias. Obviously, when an executive is criticized for a bad decision outcome, even if they did a good job during the decision process, there is no chance for them to learn. Also, poor decision makers rewarded with a good outcome will not learn if their job is not evaluated on the process.21
A crucial point for learning here is feedback by fellows and superiors. Especially regarding Kahnemann and Tversky’s findings that the human mind which is not able to identify own bias, this becomes evident. Even after having read this and knowing about our own bias, this still does not make us able to eliminate them.22 Nonetheless, it is useful to know about all the traps in order to identify them when others commit them.
2.1.4 Decision model to distribute responsibilities
So, even though executives are not very powerful in identifying their own biases, they have the possibility to recognize and neutralize those of their team members. With the implementation of effective decision processes, managers contribute to a better quality of decision-making in their organization.23
For a rapid and effective decision making process it is important for any company to have an implemented system on how strategic decisions are made. In fact, a crucial factor is that the different roles and responsibilities the decision process involves are clearly assigned to one person or a group of persons. The RAPID model proposes that one person should recommend a proposal in the beginning, which should be supported with data and analytics. This proposal must be agreed with by a different person in the next step. If both parties cannot agree, an alternative must be generated or the final, accountable decision maker must be involved. The person or group responsible for the agreement could be either from legal department or from the head of the unit which will be affected by the decision. Now, Input will be provided by those people who afterwards will implement the decision to discuss how practical it is. Finally, one executive will decide and take all the accountability. They need to understand the trade-offs that have taken place before. Further characteristics of them should be keenness to act and an awareness of the concerned business unit. The performers do eventually implement the decision promptly and effectively.24
Abbildung in dieser Leseprobe nicht enthalten
Figure 5: The RAPID decision model
Source: cf. Blenko, M., Rogers, P. (2006), p.4.
This model allows everyone to know what to do (and what not to), distribute responsibilities clearly, and allow the final decision maker to have a view apart on the decision making process and possible traps in order to eliminate them.
2.2 Classification of the term Big Data
2.2.1 The term Big Data
When Thomas H. Davenport, professor for information technology and -management and scientist at the MIT Center for Digital Business first came across the term Big Data and the associated hype in 2010, he thought that the phenomenon was just a new label for analytics and even considered just replacing the term Big Data with analytics in his books. Doing his research on Big Data, however, he noticed that there was no need to be skeptical and that Big Data was not just a new term for an already existing approach.25
Even though people are quite familiar to the term Big Data today, this confusion of what it actually stands for still has not disappeared. This becomes clear when reading through articles about Big Data: unfortunately, people - especially vendors - started using the term Big Data for analytics and sometimes even for reporting or Business Intelligence. This demands a careful selection and rejection when investigating about the term.26 Discussing about the term Big Data requires nonetheless a precise understanding of what it really describes and why it implies an immense potential for value extraction. Gartner defines Big Data as follows:
“Big data is high- volume, high- velocity and/or high- variety information assets that demand cost-effective, innovative forms of information processing that enable enhanced insight, decision making, and process automation.”27
Reading through this and other definitions of Big Data, there is a row of attributes which can be extracted: volume, velocity and variety are considered as the three V’s distinguishing Big Data from the topic ” analytics”.
The first noun “volume” refers to Moore’s law which says that capacity of IT-systems is increasing exponentially: since 1965, when the law was written down, capacity has been doubling every 12 to 24 months. To be more precise, Moore, cofounder of Intel, predicted that the number of transistors in integrated circuits would double approximately every two years.28 This fact is illustrated in the following figure, portraying the number of transistors in Intel processors. A transistor is “made up of semi-conductors and is a component used to control the amount of current or voltage or used for amplification/modulation or switching of an electronic signal.”29
Abbildung in dieser Leseprobe nicht enthalten
Figure 6: Moore's Law
Source: cf. Intel (n.d.), online source.
Parallel to this increasing capacity of IT-systems, there is also an exponentially increasing volume of data - one of the principles of Big Data. In fact, the data being created every second has skyrocketed to 30.000 gigabytes30, or in other words, about 2.5 quintillion bytes per day. This number is still accelerating which becomes evident by with the fact that in the last two years alone 90% of the data in the world has been created.31
“Velocity” is understood in various ways: it is the rapid pace in which new data appears, in which it is produced and changed but it also refers to the higher velocity of IT-systems Big Data requires for its processing.32 In fact, this characteristic of Big Data involves an important competitive advantage for companies: gaining real-time or nearly real-time information can make a company act much more agile than the competition does if it takes the advantage.33
The attribute “Variety” describes Big Data’s ability to extract information from a high diversity of different unstructured source types such as texts and audio-data in natural human language, blog posts, tweets, images and interactions on social networks. GPS signals provided from mobile devices or information from sensors belong to the different data sets companies are confronted with, too.34
Beside these three main characteristics, some definitions also include the following one or two V’s:
Big Data has the potential to increase the companies “Value”. Investments are made in human resources and technical infrastructure in order to generate an added value.35
“Veracity” is another attribute that only appears in some definitions. It refers to the completeness, accuracy and reliability of data. Big Data applications typically also include data which possibly does not fulfill those characteristics. Social media data, which is deeply subjective, is an example for such unreliable data. Therefore, one central task of Big Data applications is the consideration of these factors when analytics are made.36 Special algorithms must be used to judge the quality of the result because of the data’s vagueness. High volumes of data do not automatically generate a better evaluation outcome; it depends on the quality of the introduced data.37
This immense growth of data amounts has profoundly affected business: traditional database systems, such as relational databases and data management techniques have reached their limit and broke under the pressure of Big Data. This is why new technologies have emerged, many of them under the term NoSQL (“Not only SQL”). Logically, big companies working with large amounts of data such as Google or Amazon were the pioneers in creating Big Data systems. Hadoop from the Apache Software Foundation, for example, is one among the most common ones. It allows the loading, storage and query of massive data sets on a large grid of servers and the simultaneous execution of advanced analytics.38
2.2.2 Understanding the Framework
Since within Big Data and its environment quite a lot of technical terms do exist, the most important ones which appear in this work will be explained here and will be embedded in a contextual framework.
Data Warehouse (DWH): This kind of data stores was first described by Bill Inmon as “a subject-oriented, integrated, nonvolatile, and time-variant collection of data in support of management’s decisions” 39. Later he refined this concept to an enterprise data warehouse, which was proposed as the single repository of the company’s entire historic data.40
According to Dorschel, Data-Warehousing and Business Intelligence are concepts which change various data via ETL-processes into a structured form in order to analyze them afterwards whereas Big Data aims at making the computer understand this unstructured data and to interpret it without having to change it before into structured forms.41 Davenport does not totally agree since he points out that unstructured data needs to be converted into structured data before it can be analyzed - this is what Big Data tools are doing for example with Sentiment Analytics on Social Media: Contributions are classified as positive, negative or neutral in quantitative descriptions (-1,0,1) in order to be analyzed afterwards.42 The illustration below represents the main characteristics of a data warehouse (DWH):
[...]
1 Brynjolfsson, E. (n.d.), qtd. in Lohr, S. (2009), online source.
2 HAL’s original creator was Arthur C. Clarke with his Space Odyssey series, 1968.
3 cf. Das Erste (2016), online source.
4 cf. Gaede, L. (2016), online source.
5 cf. Brynjolfsson, E., Hitt, L., Kim, H. (2011), p.1.
6 Business Dictionary (n.d.b), online source.
7 cf. Kepner, C., Tregoe, B. (1985), p.109.
8 nobelprize.org (1978), online source.
9 cf. Simon, H. (1960), p.8.
10 cf. Simon, H. (1960), p. 1-2.
11 cf. Russo, J.E., Schoemaker, P. (1989), 174.
12 cf. Sjöberg, L. (1983), p. 131, 132.
13 cf. Toda, M. (1976), p. 79, 80.
14 cf. Sjöberg, L. (1983), p. 134, 135.
15 cf. Hammond, J.S., Keeney, R.L., Raiffa, H. (1998), p. 47.
16 cf. Kahnemann, D., Lovallo, D., Sibony, O. (2011), p. 52.
17 cf. Lovallo, D., Sibony, O. (2010), online source.
18 cf. Hammond, J.S., Keeney, R.L., Raiffa, H. (1998), p. 48-58.
19 cf. Kahnemann, D., [et al.] (2016), online source.
20 cf. Russo, J.E., Schoemaker, P. (1989), 187, 188.
21 cf. Gino, F. (2016), online source.
22 cf. Kahnemann, D., Lovallo, D., Sibony, O. (2011), p. 52.
23 cf. Kahnemann, D., Lovallo, D., Sibony, O. (2011), p. 52.
24 cf. Blenko, M., Rogers, P. (2006), p. 1-4.
25 cf. Davenport, T.H. (2014a), p. 3.
26 cf. Davenport, T.H. (2014a), p. 7.
27 Gartner Inc. (n.d.a), online source.
28 cf. Moore, G.E. (1965), pp. 114–117.
29 Computer Hope (n.d.), online source.
30 cf. Marz, N., Warren, J. (2015), p.1.
31 cf. IBM (n.d.a), online source.
32 cf. Dorschel, J. (2015), pp. 6-8.
33 cf. Brynjolfsson, E., McAfee, A. (2012b), p. 63.
34 cf. Marz, N., Warren, J. (2015), p.1
35 cf. Meier, A., Kaufmann, M. (2016), p. 13.
36 cf. Dorschel, J. (2015), p. 8.
37 cf. Meier, A., Kaufmann, M. (2016), p. 13.
38 cf. Davenport, T.H., Barth, P, Bean, R. (2012), p. 24.
39 cf. Inmon W.H. (2005), p.29.
40 cf. Stackowiak [et al.] (2015), p. 3.
41 cf. Dorschel, J. (2015), p. 8.
42 cf. Davenport, T.H. (2013), p. 17.
-
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X.