Chapter 1

Introduction

1.1 INTRODUCTION TO DECISION SUPPORT SYSTEM

1.2 CHARACTERISTICS OF DECISION SUPPORT SYSTEM

1.3 Types of Decisions

1.4 Applications of DSS

1.5 Types of DSS

Chapter 1

Introduction

As a new graduate, equipped with the modeling and algorithmic skills taught in a standard Operations management curriculum, Kiran is ready to solve real-world problems. With a knowledge and understanding of theory and applications of mathematical programming, simulation techniques, inventory management, supply-chain management, and other industrial engineering (IE) or operations research (OR) and business topics, she is ready to help her company solve distribution problems by linear programming, inventory problems by applying the economic order quantity (EOQ) model, and manpower planning problems by integer programming. However, as she works on projects with her experienced colleagues and presents results to management, she realizes that she needs more than her models and equations. The management needs Kiran to help solve decision problems, but they only want to know the final results of the analysis; they have no time or interest in understanding the mathematical model for these problems. They want this decision analysis tool to be available as a software system in which they can modify parameters and see different results for various scenarios. However, Kiran is clueless about how to develop such a system. She knows the right model but she does not know how to package her model and how to present it with friendly graphical user interface. She feels that her education did not impart to her the skills she needs to meet her job requirements.

Kiran is not the only one facing problems in her job as an operations research or business decision analyst.This is a widely prevalent problem which is not addressed in the current IE/

OR or business curriculums. As OR practitioners and business analysts, students are support staff members and are required to build systems for non-OR users. They must know how to package OR/business models so that they can be comfortably used by top managers and other co-workers. Real-life decision making often requires building interactive systems, which students must know how to design and implement. To summarize, students must learn sufficient information technology skills so that they can build intelligent information systems, alternatively, called decision support systems, which can run sophisticated models at the back-end, but are friendly enough at the front end to be used comfortably by any user.

A decision support system (DSS) gives its users access to a variety of data sources, modeling techniques, and stored domain knowledge via an easy to use graphical user interface (GUI).

For example, a DSS can use the data residing in spreadsheets or databases, prepare a mathematical model using this data, solve or analyze this model using problem-specific methodologies, and can assist the user in the decision-making process through a graphical user interface. Students are frequently being employed in positions that require developing DSS which are gaining widespread popularity. As more and more companies install enterprise resource planning (ERP) packages and invest in building data warehouses, those who are able to create decision technology driven applications that interface with these systems and analyze the data they provide will become increasingly valuable. Indeed, imparting DSS development skills, which combine OR/business skills with information technology (IT) skills, will make students highly sought after in the modern workplace.

Developing courses that teach our students how to build decision support systems has been a challenging task so far since it requires the availability of platforms which allow the integration of various technologies (data, models, codes, etc.). However, in the past few years, several platforms have become available which allow such integration. One such platform is Microsoft Excel. Excel, which is the most widely used spreadsheet package among managers and engineers, allows data storage and model building. Excel also has many built-in programs as well as many add-on programs available that allow optimization and simulation of various models built in Excel. Excel also has a macro programming language, Visual Basic for Applications (VBA), which allows building GUIs and manipulating Excel objects. Thus, Excel provides a platform in which fairly sophisticated DSS applications can be built. This book imparts the skills needed to build such systems.

1.1 INTRODUCTION TO DECISION SUPPORT

SYSTEM

Virtually everyone makes hundreds of decisions each day. These decisions range from the inconsequential, such as what to eat for breakfast, to the significant, such as how best to get the economy out of a recession. All other things being equal, good outcomes from those decisions are better than bad outcomes. For example, all of us would like to have a tasty, nutritional breakfast (especially if it is fast and easy), and the country would like to have a stable, well-functioning economy again. Some individuals are "lucky" in their decision processes. They can muddle through the decision not really looking at all of the options or at useful data and still experience good consequences. We have all met people who instinctively put together foods to make good meals and have seen companies that seem to do things wrong but still make a good profit. For most of us, however, good outcomes in decision making are a result of making good decisions.

"Good decision making" means we are informed and have relevant and appropriate information on which to base our choices among alternatives. In some cases, we support decisions using existing, historical data, while other times we collect the information, especially for a particular choice process. The information comes in the form of facts, numbers, impressions, graphics, pictures, and sounds. It needs to be collected from various sources, joined together, and organized. The process of organizing and examining the information about the various options is the process of modeling. Models are created to help decision makers understand the ramifications of selecting an option. The models can range from quite informal representations to complex mathematical relationships.

For example, when deciding on what to eat for a meal, we might rely upon historical data, such as those available from tasting and eating the various meal options over time and our degree of enjoyment of those options. We might also use specially collected data, such as cost or availability of the options. Our model in this case might be simple: Select the first available option that appeals to us. Or, we might approach it with a more complex approach: Use linear programming to solve the "diet problem"1 to find the cheapest combination of foods that will satisfy all the daily nutritional requirements of a person.

1 !The diet problem was one of the first large-scale optimization problems solved using modern modeling

techniques. The Army wished to find the cheapest way to provide the necessary nutrition to the field soldiers. The National Bureau of Standards solved the problem with the simplex method (which was new then) with 9 equations and 77 variables. To solve the problem, it took nine clerks using hand-operated calculators 120 days to find the

In today's business world, we might use models to help refine our understanding of what and how our customers purchase from us to improve our customer relationship management. In that case we might collect information from point-of-sale systems for all of our customers for multiple years and use data-mining tools to determine profiles of our customers. Those profiles could in tum profile information about trends with which managers could change marketing campaigns and even target some marketing campaigns.

The quality of the decision depends on the adequacy of the available information, the quality of the information, the number of options, and the appropriateness of the modelling effort available at the time of the decision. While it is not true that more information (or even more analysis) is better, it is true that more of the appropriate type of information (and analysis) is better. In fact, one might say that to improve the choice process, we need to improve the information collection and analysis processes. Increasingly corporations are attempting to make more informed decisions to improve their bottom lines. Some refer to these efforts to use better information and better models to improve decision making as business intelligence. Others refer to it as analytics. In either case, the goal is to bring together the right information and the right models to understand what is going on in the business and to consider problems from multiple perspectives so as to provide the best guidance for the decision maker.

One way to accomplish the goal of bringing together the appropriate information and models for informed decision making is to use decision support systems (DSS). Decision support systems are computer-based systems that bring together information from a variety of sources, assist in the organization and analysis of information, and facilitate the evaluation of assumptions underlying the use of specific models. In other words, these systems allow decision makers to access relevant data across the organization as they need it to make choices among alternatives. The DSS allow decision makers to analyze data generated from transaction processing systems and other internal information sources easily. In addition, DSS allow access to information external from the organization. Finally, DSS allow the decision makers the ability to analyze the information in a manner that will be helpful to that particular decision and will provide that support interactively.

So, the availability of DSS provides the opportunity to improve the data collection and analyses processes associated with decision making. Taking the logic one step further, the availability of DSS provides the opportunity to improve the quality and responsiveness of decision making and hence the opportunity to improve the management of corporations. Said differently, the DSS provides decision makers the ability to explore business intelligence in an effective and timely fashion.

To see how DSS can change the way in which decisions are made, consider the following example of a Manhattan court. Consider the problem. New York spends in excess of $3 billion each year on criminal justice and the number of jail beds has increased by over 110% in 20 years. In Manhattan, in particular, developers have spent billions of dollars refurbishing neighborhoods and providing good-quality living, business, and entertainment areas. Yet people continue not to feel safe in them, and minor crimes depreciate the quality of life for residents. Furthermore, the likelihood of repeat offenses is high; over 40% of the defendants seen in a year already have three or more convictions.

optimal solution. For more information on the diet problem, including a demonstration of the software, check the NEOS page at http://wwwneos. mcs.anl.gov/CaseStudies/dietpy/WebForms/index.html.

While clearly there is a problem, those facts (that crime exists, that enormous amounts of money are spent, and that people do not feel safe) are examples of bad outcomes, not necessarily bad decisions. However, three facts do suggest the quality of the decision could be improved:

• Criminal justice workers know very little about the hundreds of thousands of people who go through the New York court systems.

• There has been little creative thinking about the sanctions judges can use over time.

• Most defendants get the same punishment in the same fashion.

Specifically, they suggest with more information, more modeling capabilities, and better alternative generation tools that better decisions, which could result in superior outcomes, might be achieved.

In this case, citizens, court officials, and criminal justice researchers noted the problem of information availability and have developed a process to address it for "quality-of-life" crimes, such as shoplifting and street hustling. Specifically, the city, landlords, and federal funding jointly created a new court and located the judge in the same building as city health workers, drug counselors, teachers, and nontraditional community service outlets to increase the likelihood of the court working with these providers to address the crime problem innovatively. The centerpiece of this effort is a DSS that provides judges with more and better information as well as a better way for processing that information so as to make an impact on the crime in Manhattan.

1.2 CHARACTERISTICS OF DECISION SUPPORT

SYSTEM

A DSS is not meant to replace a decision maker, but to extend his/her decision making capabilities. It uses data, provides a clear user interface, and can incorporate the decision maker’s own insights. Some of the major DSS capabilities are the following:

• A DSS brings together human judgment and computerized information for semistructured decision situations. Such problems cannot be conveniently solved by standard quantitative techniques or computerized systems.

• A DSS is designed to be easy to use. User friendliness, graphical capabilities, and an interactive human-machine interface greatly increase the effectiveness of a DSS.

• A DSS usually uses models for analyzing decision-making situations and may also include a knowledge component.

• A DSS attempts to improve the effectiveness of decision making rather than its efficiency.

• A DSS provides support for various managerial levels from line mangers to top executives. It provides support to individuals as well as groups. It can be PC-based or web-based

A DSS must access data from a variety of sources. In our court example, the system accesses the arresting officer's report, including the complaint against the offender and the court date. In addition, the DSS provides access to the defendant's criminal record through connections with the New York Division of Criminal Justice. These police records are supplemented with information gained by an independent interviewer either at the police precinct or at the courthouse. These interviewers query the defendant regarding their lifestyle, such as access to housing, employment status, health conditions, and drug dependencies. Finally, an intermediary between the court and the services available, called a court resource coordinator, scans the person's history, makes suggestions for treatment, and enters the information into the system.

A second characteristic of a DSS is that it facilitates the development and evaluation of a

model of the choice process. That is, the DSS must allow users to transform the enormous amount of "data" into "information" which helps them make a good decision. The models may be simple summarization or may be sophisticated mathematical models. In this case, the modeling takes on a variety of forms. The simple ability to summarize arrest records allows judges to estimate recidivism if no intervention occurs. Further, the summarization of lifestyle information encourages the development of a treatment model. In addition, with the DSS, the judge can track community service programs and sites to determine which is likely to be most effective for what kinds of offenses. Hence, the judge can model the expected impact of the sanctions on a defendant with particular characteristics. In other words, it can facilitate the evaluation of programs to determine if there is a way to have greater impact on particular defendants or on a greater number of defendants.

The design team is in the process of adding additional modeling capabilities. Soon, they hope to integrate mapping technology that will plot a defendant's prior arrest record.The judge can evaluate this map to determine (a) if there is a pattern in offenses that can be addressed or (b) where to assign community service sentence to optimize the payback to society.

The third characteristic that is demonstrated by this DSS is that they must provide a good user

interface through which users can easily navigate and interact.

A decision support system is a model-based or knowledge-based system intended to support managerial decision making in semi-structured or unstructured situations (Turban and Aronson, 2001). A DSS is not meant to replace a decision maker, but to extend his/her decision making capabilities. It uses data, provides a clear user interface, and can incorporate the decision maker’s own insights. Some of the major DSS capabilities are the following:

• A DSS brings together human judgment and computerized information for semistructured decision situations. Such problems cannot be conveniently solved by standard quantitative techniques or computerized systems.

• A DSS is designed to be easy to use. User friendliness, graphical capabilities, and an interactive human-machine interface greatly increase the effectiveness of a DSS.

• A DSS usually uses models for analyzing decision-making situations and may also include a knowledge component.

• A DSS attempts to improve the effectiveness of decision making rather than its efficiency.

• A DSS provides support for various managerial levels from line mangers to top executives. It provides support to individuals as well as groups. It can be PC-based or web-based

Figure 1.1 A schematic view of a decision support system.

A DSS application contains five components: database, model base, knowledge base, GUI, and user (see Figure 1.1). The database stores the data, model and knowledge bases store the collections of models and knowledge, respectively, and the GUI allows the user to interact with the database, model base, and knowledge base. The database and knowledge base can be found in a basic information system. The knowledge base may contain simple search results for analyzing the data in the database. For example, the knowledge base may contain how many employees in a company database have worked at the company for over ten years. A decision support system is an intelligent information system because of the addition of the model base. The model base has the models used to perform optimization, simulation, or other algorithms for advanced calculations and analysis. These models allow the decision support system to not only supply information to the user but aid the user in making a decision. We now present a more detailed look at each of these components.

Database: The database provides the data with which decisions are made. The data may reside in spreadsheets or a data warehouse, a repository for relevant corporate decision-making data. The database allows a user to access, manipulate, and query data. Some examples of databases would include a spreadsheet containing personal banking account information or a data warehouse containing shipment records of various products.

Model Base: A model base contains statistical, financial, optimization, or simulation models that provide the analysis capabilities in a DSS. Some popular optimization models include linear programming, integer programming, and nonlinear programming. The DSS allows the ability to invoke, run, and change any model or combine multiple models. An example of a model base would be an integer programming model used to solve a capital budgeting problem.

Knowledge Base: Many managerial decision making problems are so complex that they require special expertise for their solution. The knowledge base part of a DSS allows this expertise to be stored and accessed to enhance the operation of other DSS components. For example, credit card companies use a DSS to identify credit card thefts. They store in their knowledge base the spending patterns that usually follow credit card thefts; any abnormal activity in an account would trigger checking for the presence of those patterns and a possible suspension of the account.

Decision Support System

Database

GUI

Model base Knowledge

User

GUI: The graphical user interface covers all aspects of communication between a user and a DSS application. The user interface interacts with the database, model base, and knowledge base. It allows the user to enter data or update data, run the chosen model, view the results of the model, and possibly rerun the application with different data and/or model combinations. The user interface is perhaps one of the most important components of a DSS because much of the flexibility and ease of use of a DSS are derived from this component.

User: The person who uses the DSS to support the decision making process is called the user, or decision maker. A DSS has two broad classes of users: managers and staff specialists, or engineers. When designing a DSS, it is important to know for which class of users the DSS is being designed. In general, managers expect a DSS to be more user-friendly than do staff specialists.

A DSS should be distinguished from more common management information systems (MIS). An MIS can be viewed as an information system that can generate standard and exception reports and summaries for managers, provide answers to queries, and help in monitoring the performance of a system using simple data processing. A DSS can be viewed as a more sophisticated MIS where we allow the use of models and knowledge bases to process the data and perform analysis.

1.3 Types of Decisions

The characteristics of decisions faced by managers at different levels are quite different. Decisions can be classified as

• structured

• semi - structured

• unstructured

Unstructured decisions are those in which the decision maker must provide judgement, evaluation, and insights into the problem definition. Each of these decisions is novel, important, and non-routine, and there is no well understood or agreed - on procedure for making them.

Structured decisions by contrast, are repetitive and routine, and decision makers can follow a definite procedure for handling them to be efficient. In general, structured decisions are made more prevalently at lower organizational levels, whereas unstructured decision making is more common at higher levels of the firm.

Many decisions having elements of both and are considered semi- structured decisions, in which only part of the problem has a clear cut answer provided by an accepted procedure. Thus in reality, however, there are many decisions that fall somewhere in between the two extremes and these decisions are the semi - structured ones.

Senior executives tend to be exposed to many unstructured decision situations that are open ended and evaluative and that require insight based on many sources of information and personal experience. For example, a CEO in today’s music industry might ask, “Whom should we choose as a distribution partner for our online music catalogue Apple, Microsoft, or Sony?” Answering this question would require access to news, government reports, and industry views as well as high level summaries of firm performance. However, the answer would also require senior managers to use their own best judgement and poll other managers for their opinions.

Middle management and operational management tend to face more structured decision scenarios, but their decisions may include unstructured components. A typical middle level management decision might be “Why is the order fulfilment report showing a decline over the last six months at a distribution centre in Ashanti?” This middle manager could obtain a report from the firm’s enterprise system or distribution management system on order activity and operational efficiency at the Ashanti distribution centre. This is the structured part of the decision. But before arriving at an answer, this middle manager will have to interview employees and gather more unstructured information from external sources about local economic conditions or sales trends.

Rank and file employees tend to make more structured decisions. For example, a sales account representative often has to make decisions about extending credit to customers by consulting the firm’s customer database that contains credit information. In this case the decision is highly structured, it is a routine decision made thousands of times each day in most firms, and the answer has been pre-programmed into a corporate risk management or credit reporting system.

The types of decisions faced by project teams cannot be classified neatly by organizational level. Teams are small groups of middle and operational managers and perhaps employees assigned specific tasks that may last a few months to a few years. Their tasks may involve unstructured or semi-structured decisions such as designing new products, devising new ways to enter the marketplace, or reorganizing sales territories and compensation systems.

1.4 Applications of DSS

Given the above definition of a decision support system, we have developed several spreadsheet-based DSS applications in Excel. Using the spreadsheet functionality in Excel and the VBA programming capabilities, a complete decision support system can be developed. As an IE/OR or business graduate, students will discover opportunities in their careers to develop and use DSS applications. Many students who have taken DSS courses have reported that their skills of combining modeling with an information technology packaging have truly helped them be outstanding at their jobs. Let us consider two examples of DSS applications which may be found in industry.

Car Production: Consider a factory which produces cars. The manager of the factory, probably a business or IE/OR graduate, may need to make some large-scale decisions about ordering parts, hiring/firing employees, finding new suppliers, or making changes to the production process. Let us focus on a production process DSS application. The manager may be deciding where to place a new piece of equipment or how to add a new product part to the production sequence. With the use of some basic simulation and analysis tools, one could develop a DSS which allows the manager to enter the parameters to describe a possible scenario and see how it would affect production. The manager may not want to know the details of the models used, but rather what would be the affect on cost and production time and quantity if a specific change was made. This is just one example of how a DSS could be used to aid in the car production industry.

Railroad Car Management: Consider a railroad company which owns several trains on which they place several thousand railroad cars which ship to several cities in the country. A distribution manager, again with an IE/OR or business background, may need to decide which

cars should go on which trains to which cities. He would benefit from using an optimization model which allows him to modify certain constraints or focus on various objectives and compare the resulting distribution plans. He may want to display the car and train plans visually, may be projected on a country map, to have a better understanding of the effects of one solution compared to another. A DSS would aid him in accomplishing this analysis and making a decision which considers all scenarios and possible outcomes.

The applications we develop are basic illustrations of decisions which are made in IE/OR and business industries. Two examples, selected from the DSS applications developed in Part III of this book, are described below.

Portfolio Management and Optimization: In this application, we allow users to create and/or edit their portfolio as well as optimize their investments. To create a portfolio, users can choose from a list of stocks in a database in Excel and add or remove them to/from their portfolio. The stocks are listed by name and category. These stocks can also be compared using their historical annual returns and changes in market price. Once users have created their portfolio, they may also edit it at any time. Users may then optimize their investment by specifying a desired return on their portfolio and the amount available for investing. The application then solves an optimization problem which minimizes users’ risk on their selected portfolio. The suggested investment strategy is then displayed to the users. This application allows users to interact with the database of stock information and the knowledge base of comparative statistics. The model base for this DSS uses the Excel Solver to perform an optimization which minimizes the user’s risk on the portfolio investments.

Facility Layout: In this application, we study a facility location problem which consists of placing n facilities at n locations to minimize the total handling cost. This problem is also known as the Quadratic Assignment Problem (QAP). The QAP arises in many other applications, such as the allocation of plants to candidate locations, the backboard wiring problem, design of control panels and typewriter keyboards, turbine balancing, etc. The user begins the application by providing the size of their facility. From the dimensions provided, a layout is displayed and a distance matrix is created. Random flow matrix values are generated which the user may overwrite if desired. From these two matrices, the cost matrix is derived; the total of the costs from this matrix is minimized by performing a pair-wise local search on the user’s facility. The user may also opt to fix some facilities so that they cannot be moved when the local search is performed. The user may run this local search algorithm automatically or participate in the decision taken at each iteration. The final layout is then displayed to the user. The model base for this application uses an algorithm developed in the VBA code. This DSS application aids a facility designer in creating a facility layout which minimizes total handling cost. It can be used to solve typical facility layout problems which may be encountered by plant managers, school administrators, or in other applications. This DSS can also be used as a pedagogical tool in facility planning courses to illustrate the pairwise local search technique.

1.5 Types of DSS

We can differentiate among types of DSS by looking at their major purpose. Holsapple and Whinston (1996) identified six types of DSSnamely:

• text-oriented DSS

• database-oriented DSS

• spreadsheet-oriented DSS

• solver-oriented DSS

• rule-oriented DSS

• compound DSS.

Examples of text-oriented systems include catalog books, periodicals, reports, memos, and other written documents so that their contents can be made available to decision makers. Each document, or a portion of that document, provides some information or even knowledge that could be important to a decision maker when making choices. The system allows you to categorize, consolidate, and merge documents as well as to write comments about the contents and the value thereof. By allowing users to focus on portions of documents, the system helps decision makers save time when they need to refer to the document. In addition, intelligent systems can perform content analyses of the texts and recommend sections (and thus information) the decision maker might not otherwise consider.

A variation on the text-oriented DSS is the hypertext-oriented DSS. The hypertext-oriented DSS provides the same basic functions that text-oriented systems do, but the documents are logically related and linked. This allows the decision makers to follow specific subjects among documents when making choices. No longer do they need to go through documents in a linear fashion to find the important information. They can instead transverse the information in all of the various sources, thereby supplementing his or her abilities to associate relevant portions of the text. Of course, since we now are accustomed to such links because of Web surfing, we generally take such abilities for granted in our online documents.

Database-oriented DSS are similar to the text systems in that they provide descriptive information that is of relevance to a choice under consideration. Instead of providing text, though, these systems focus on discrete data that are stored in a database. The system controlling these databases allow for manipulating and joining the data and presenting those data in ways that will benefit decision makers. Generally such systems use Structured Query Language (SQL) through which to identify and manipulate the data. Some minimal summaries of the data can be provided through the use of these SQL commands

Spreadsheet-oriented DSS, as the name suggests, use the tools available in a spreadsheet to summarize and analyze the data. Instead of just providing access to data, these DSS allow the decision maker to create some basic models and to evaluate those models in a quick and efficient manner.

Similarly, solver-oriented DSS provide some kind of modeling package as the basis of the DSS. These systems allow decision makers to identify more varied and sophisticated relationships among the data. The modeling package may be integrated into the DSS or simply used by the DSS depending on the architecture of the system.

A rule-oriented DSS or intelligent DSS provides advisory support to decision makers. Early examples were rule based of the form

IF <some premise is true>

THEN <some condition is true>

ELSE <some other condition is true>

By linking the rules together, these systems could provide some cognitive functions and prove something to be true (or sometimes false) or reason as far as the data allowed toward a conclusion. Improvements in artificial intelligence technologies have allowed these systems to demonstrate more sophisticated reasoning and even some learning.

The compound DSS are hybrid combinations of the individual types of DSS. Such systems have mixed capabilities, such as a solver-database combination or a spreadsheet database intelligence combination. The different components exist equally within the system and allow complete flexibility in their use. As you might expect, such hybrid designs are the most common form of DSS today.