Business Intelligence

Jerzy Surma

• Publisher: Business Expert Press
• Year: 2011
• ISBN:
  9781606491867
• Topic: Business
• Subtopic: Information Management
• Index: Business

This book is about using business intelligence as a management information system for supporting managerial decision making. It concentrates primarily on practical business issues and demonstrates how to apply data warehousing and data analytics to support business decision making. This book progresses through a logical sequence, starting with data model infrastructure, then data preparation, followed by data analysis, integration, knowledge discovery, and finally the actual use of discovered knowledge. All examples are based on the most recent achievements in business intelligence. Finally this book outlines an overview of a methodology that takes into account the complexity of developing applications in an integrated business intelligence environment. This book is written for managers, business consultants, and undergraduate and postgraduates students in business administration.

Chapter 1

An Introduction to Business Intelligence

1.1. The Origins of Business Intelligence

During the 1970s Herbert Simon, a Nobel Prize winner in economics, was developing his world-famous concept of bounded rationality. He was certainly inspired by an interest in examining the cognitive boundaries of a man following the disappointing results of his own trials concerning the computer stimulation of human decision-making processes.¹ However, the intensive development of computer technologies gave him great hope for building systems to support human activities related to thinking and rational behavior. Those hopes were at least partly fulfilled thanks to the development of business intelligence (BI)—that is, a system that supports managerial decision making in enterprise management in the broadest sense of this word. A distinctive feature of BI is its powerful pragmatism: Out of the broad spectrum of technologies, only those that can be applied to business are selected. The four fundamental sources of information for BI and its tools are as follows:

1. Statistics and econometrics, including inter alia statistical theories of pattern recognition, econometric methods, statistical reasoning, and forecasting techniques
2. Operations research, including inter alia linear programming, decision theory, and game theory
3. Artificial intelligence, including inter alia heuristic search methods, machine learning, expert systems, genetic algorithms, artificial neural networks, and case-based reasoning systems
4. Database technologies, including inter alia data modeling, query languages, query optimization, and indexing methods

The application of computers in statistics and operations research resulted in the creation of so-called decision support systems. These systems required the application of formal mathematical models and were mainly based on quantitative data. Simultaneously, they naturally reduced the areas of use and application of computers in modeling real decision-making problems.

Artificial intelligence (AI) faced a much more challenging task. Work on AI was initiated in the 1950s, but to date it has not been crowned with any spectacular success. It was then a lesson in humility for the academic environment and proved how complex and refined human intelligence is. Nevertheless, thanks to those attempts numerous algorithms for supporting real decision-making processes were worked out that go far beyond the capabilities of decision support systems.

Although both AI and decision support systems had a large influence on BI, it was the development of database technologies initiated in the 1960s that was the most important. The development of databases based on the relational data model that allows for the relatively simple interpretation of business data and structured query language (SQL), which was easy to use for those times, proved to be particularly important. Progress in database technologies led to a boom in business applications of enterprise resource planning (ERP), which allows the standard processes of an enterprise to be automatic and well arranged. These transaction data (e.g., expense entering, invoice registering, and recording of bank account transactions or phone call records in a billing system) were soon discovered to be a source of interesting insights into the activities of a business. First, transaction data was aggregated into various reports, which were generated by decision support systems usually by means of SQL. Then techniques derived from decision support systems and artificial intelligence, which could conduct more sophisticated analyses, started to be applied. Those activities, performed at the turn of the 1970s and 1980s, were occasional and originally developed in two areas—namely, in shopping malls² and telecommunications companies.

1.2. BI as an Autonomous Discipline

In the 1980s, business applications became so advanced that a separate discipline of designing and creating databases for business decision support emerged. So-called data warehouses³ and specialist toolsets appeared. Although the term “business intelligence” was first used in 1958 in a paper for IBM Journal,⁴ a new sense was imparted to it by Howard Dresner from the Gartner company in 1988. Having analyzed the information technology (IT) market, he referred to business intelligence as to a kind of “umbrella” that covers numerous methods, technologies, and applications oriented to real business decision support in an enterprise:⁵

Business Intelligence is a user-oriented process of gathering, exploring, interpreting and analyzing of data, which leads to the streamlining and rationalization of the decision-making process. Those systems support managers in business decision-making in order to create economy value growth of an enterprise.

Such a definition explicitly points out that BI is an IT management system and, strictly speaking, a third-generation IT management system.⁶ In light of such an understanding of decision support systems, they encompass a broad spectrum of technologies, including the following:

• Online analytical processing (OLAP) tools. Software for multidimensional analysis of business data by integration, aggregation, and adequate mode of presentation and visualization of different data
• Data-mining tools. Algorithms for automatic analysis of great volumes of data using statistical and econometric methods, as well as machine learning methods that can analyze not only quantitative but also qualitative data
• Knowledge management tools. Tools that allow for storage, indexing, and analysis of textual documents and their further linkage with other data

This class of technological systems is based on the data collected by data warehouses—that is, database systems that gather data from various sources and make it readily available to businesses.

In the 1990s, BI became a widely known term among specialists, and on the level of tools, it was a standard offered not only by specialist companies but also by the greatest software manufacturers to enterprises, such as IBM, Microsoft, Oracle, or SAP.

1.3. BI and Company Management

At the beginning of the 21st century, IT technologies were developing extremely rapidly due to sudden Internet developments. Despite the almost total computerization of fundamental business processes, managers still have a fragmentary knowledge of their own businesses and often make decisions intuitively. Simultaneously, confusion caused by the excess of available data and a lack of its organization can be observed. Moreover, errors in data, lack of cohesion, and having a few versions of “the truth” in an enterprise have also led to a lack of trust in the gathered data. All of these factors aided the increased use of BI by enterprises. There are some sectors that already cannot do without such solutions—for instance, telecommunications and banking. The belief that analytical technologies are key tools to gaining a competitive advantage is also clearly visible.⁷ Generally speaking, the successful development of BI contributed to the fulfillment of Herbert Simon’s vision. According to his theories, the tasks managers deal with fall into three categories:⁸

1. Supervising the standard activities connected with the management of business processes and subordinates
2. Solving well-structured tasks (problems), that is, programmable decisions that are routine and repeatable and for which strict procedures have been worked out: For these tasks (e.g., establishing a selling price when logistics costs and purchase price are familiar), every single case does not have to be considered individually
3. Solving ill-structured tasks (problems)—that is, nonprogrammable decisions that don’t have a cut-and-dried answer—related to new cases in which no pattern of behavior is established, results are unknown, and there is also no ready-made solution (e.g., a strategic decision about starting up manufacturing abroad)

It should be stressed that at the moment BI is only applied to the first and the second category. The third category is definitely the most interesting one. At present, trials are being conducted in the application of BI methods as solutions for this type of problem. In order to understand the complexity of this issue, let us look at Table 1.1, which describes well- and ill-structured problems.

In formal terms, managerial tasks are decision processes, by which a decision means the selection of one possibility from a set of possible solutions. A BI system can generally support managerial decisions in the following ways (see Figure 1.1):

1. Providing a decision maker with some information. This mode encompasses the preparation of adequate information: business reports and outcome from complex analysis.
2. Proposing managerial decisions. This approach also includes the possibility of a system making decisions itself.⁹

Making decisions in supervision activities and solving well-structured problems can be supported by providing information and by proposing managerial solutions, while solving ill-structured problems might be supported by BI by their rationalization, that is, by providing the management board of an enterprise with suitable information (see chapter 3). However, proposing managerial solutions for ill-structured problems is not the subject of BI systems’ activities.

Table 1.1. Well- and Ill-Structured Problems

Source: Based on Turner (1988).

Table 1.1. Well- and Ill-Structured Problems
	Well-structured problems	Ill-structured problems
Data	Quantitative, specific	Qualitative, unspecific
Knowledge	Mathematic model, al...