Problem analysis paradigms

Problem analysis paradigms

15.1 Purpose

This chapter introduces several paradigms for locating, pinpointing, and identifying a problem or an opportunity, including decomposition, factoring, synthesis, and generate and test. These paradigms can also be applied to problem solving.

15.2 Strengths, weaknesses, and limitations

The strengths and weaknesses of each paradigm will be noted in context.

15.3 Inputs and related ideas

These paradigms serve as the philosophical basis for numerous tools, techniques, and methodologies. Significant links will be noted in context.

15.4 Concepts

This chapter introduces several paradigms for locating, pinpointing, and identifying a problem or an opportunity. These paradigms serve as the philosophical basis for numerous problem-solving tools, techniques, and methodologies.

15.4.1 Decomposition

Decomposition is a top-down, goal-oriented approach that is used when the problem is too complex or too abstract to study directly. The idea is to divide (or decompose) the problem into logically consistent, more manageable sub-problems, and then to attack the sub-problems. Much as a book can be broken into chapters, sections, and then paragraphs, the decomposition approach divides a large, abstract problem into several small, concrete sub-problems, each with clear goals or specific tasks to perform.

For example, imagine that a firm requires seven days to process an order and deliver the merchandise to a customer, but a competitor needs only three days to perform the same service. Rather than trying to solve the excessive turnaround time problem directly, it might be more effective to decompose the problem into order taking, order entry, order authorization, order filling, and shipping components, and then independently study those sub-problems.

The primary weakness of decomposition is that it can be difficult to track the interrelationships between the sub-problems. Additionally, independently solving a number of sub-problems can be time consuming. Determining acceptable criteria for decomposing the main problem can also be a difficult task.

Decomposition is used throughout the information engineering (Chapter 2) and structured analysis and design (Chapter 3) methodologies and plays an important role in such tools and techniques as data flow diagrams (Chapter 24), data normalization (Chapter 28), functional decomposition (Chapter 62), and HIPO (Chapter 64). Although this paradigm might be applied to selected sub-problems, decomposition is not as effective for bottom-up, data-oriented, or output-oriented tools and techniques. The decomposition paradigm is widely used in database design and is sometimes called normalization.

15.4.2 Factoring

The essential idea of factoring is to merge several small, isolated, overlapping, or related problems to form a meta-problem. Generally, a problem can be reformulated by identifying those sub-problems that share similar characteristics, and then grouping the related sub-problems.

For example, a system analyst investigating low profits might identify several possible causes, including excess warehouse personnel, sales floor understaffing, high stock expenses, poor quality control, poor sales effort, inadequate advertising, product shortages, excessive rework, and so on. With so many possible causes to consider, it is difficult to distinguish the trivial from the significant. Consequently, the analyst might begin studying the low profit problem by factoring the sub-problems to form the following meta-problems:

•  High production costs resulting from poor quality control and excessive rework.

•  Low sales resulting from poor sales effort, inadequate advertising, and sales floor understaffing.

•  High shipping and handling costs resulting from high stock expenses, product shortages, and excess warehouse personnel.

Focusing on the meta-problems is likely to be more efficient than attempting to independently analyze the sub-problems.

The factoring process calls for judgment. Often, a given problem can be factored in several different ways, and individual systems analysts or information system consultants might reasonably view the same problem differently. Consequently, it is essential that agreement on the sub-problems, the factoring criteria, and the meta-problems be reached early in the process.

Factoring is a bottom-up approach that lends itself to data-oriented methodologies and tools, such as the structured requirements specification methodology (Chapter 4) and Warnier-Orr diagrams (Chapter 33). Additionally, this method is widely used by expert systems (Chapter 7) to perform reasoning.

15.4.3 Synthesis

Synthesis is an evolutionary paradigm. It starts with a major or influential user’s viewpoint and expands (perhaps with revisions and/or modifications to the original problem description) by incorporating other users’ perspectives until all relevant viewpoints are included. It is useful when the core problem is well-defined and well-structured and the sub-problems are simple add-on functions that use the core as a base.

Inventory is a good example of a core problem. The starting viewpoint might be that of the functional group in charge of the warehouse. Once an effective inventory control system is implemented and a stable inventory database is established, other viewpoints can be considered. For example, the system might be enhanced to incorporate time-to-ship commitments for the sales department, on-demand inventory status reports and queries for the purchasing department, a just-in-time inventory system for production, such applications as inventory aging, continuous physical inventory, and inventory shrinkage analyses for accounting, and so on. Note that the sub-problems cannot be solved until the core problem (inventory control) is solved.

The major concern with synthesis is correctly identifying the core problem. Also, comprehensive testing is difficult because of the evolutionary nature of the paradigm. However, once the core problem is identified and solved, it becomes relatively easy to identify and solve the sub-problems. Prototyping (Chapter 31) is particularly effective for problems that fit the synthesis paradigm.

15.4.4 Generate and test

Generate and test is a hierarchical, test-oriented paradigm that is used in expert systems to define a solution that meets certain criteria or constraints. The technique starts at the top of a hierarchy with the main problem and continues down the hierarchy through the sub-problems, conducting tests of the appropriate criteria and constraints at each level until the bottom is reached and no more testing is necessary.

For example, imagine that an analyst has identified three problems, all of which contribute to lower than expected profits (the main problem):

•  Production costs are too high.

•  Sales revenues are too low.

•  Inventory costs are too high.

Management is concerned about the lower than expected profits and expects to see results within one month (a time constraint).

An initial study suggests that high production costs are probably the result of poor quality control, excessive rework, and frequent shortages of essential raw materials. Solving the first two sub-problems (quality control and rework) will require the purchase of new inspection equipment. Delivery time on that equipment is two months, which clearly exceeds management’s time constraint. Raw material shortages result from poor production planning and inadequate coordination between the production and the warehouse. Consequently, the raw material shortage problem must be solved in concert with certain warehousing problems.

The likely causes of low sales revenue appear to include poor sales effort, inadequate advertising, and understaffing in the sales department. The solutions to these three sub-problems might include better management and the reallocation of resources, and those solutions can be implemented within management’s one-month target.

High warehousing and distribution costs result from poor materials handling procedures and poor inventory management. Preliminary analysis suggests that solving the materials handling sub-problem will require a lengthy study of the existing materials handling procedures followed by the purchase of new materials handling equipment and several weeks of employee retraining. Total elapsed time to complete these tasks is expected to be three to four months. The inventory management sub-problem appears to be related to production’s raw material shortage sub-problem, so a one-month study of the relationship between production and inventory will be needed before the true scope of the problem can even be determined.

The initial low profit problem can now be viewed as five sub-problems:

  1. Poor quality control and excessive rework.

  2. Raw material shortages.

  3. Poor sales effort, inadequate advertising, and understaffing in the sales department.

  4. Poor materials handling procedures.

  5. Poor inventory management.

Solving sub-problem 1 or sub-problem 4 will exceed management’s time constraint. Sub-problems 2 and 5 are interrelated, and the need to study inventory management for a month before the true scope of the problem can be estimated means that it, too, will exceed management’s constraint. Consequently, in the short run the analyst should start by attacking sub-problem 3 (poor sales effort, inadequate advertising, and understaffing in the sales department) because it is the only sub-problem that has a chance of yielding results within the time constraint imposed by management.

The generate and test paradigm can be used to pinpoint the correct sub-problem to be solved, particularly for complex (large domain) problems with time and budgetary constraints and additional constraints on specific sub-problems. Perhaps the most important strength of the generate-and-test paradigm is its focus on real world constraints. Note, however, that other paradigms may be needed to identify the main problem and the sub-problems. Also, selecting a small set of sub-problems based on artificial constraints can lead to sub-optimization and may increase the time and cost to solve the main problem.

15.5 Key terms

Bottom-up —

An approach to problem solving that starts with the details and works upward.

Data-oriented —

A tool or technique that starts with the data and derives the necessary processes.

Decomposition —

A problem analysis paradigm that calls for breaking a problem into more manageable sub-problems and then attacking the sub-problems.

Factoring —

Merging several small, isolated, overlapping, or related problems to form a meta-problem.

Generate and test —

A hierarchical, test-oriented paradigm that starts at the top of a hierarchy with a main problem and continues down the hierarchy through the sub-problems, conducting tests of the appropriate criteria and constraints at each level until the bottom is reached and no more testing is necessary.

Goal-oriented —

A method or technique which searches through a process until a predefined goal is accomplished.

Meta-problem —

A large problem defined by combining several smaller problems.

Sub-problem —

A problem that is part of a larger problem.

Synthesis —

An evolutionary paradigm that starts with a major or influential user’s viewpoint and incorporates other users’ perspectives until all relevant viewpoints are included.

Top-down —

An approach to problem solving that starts with the high-level control structures and works down to the details.

15.6 Software

Not applicable.

15.7 References

1.  Awad, E. W., Building Expert Systems: Principles, Procedures, and Applications, West, Minneapolis/St. Paul, MN, 1996.

2.  Buchanan, B. G., and Shortliffe, E. H., Rule-Based Expert Systems: The MYCIN Experiments of the Standard Heuristic Programming Project, Addison-Wesley, Reading, MA, 1984.

3.  Connor, D., Information Systems Specification & Design Road Map, Prentice-Hall, Englewood Cliffs, NJ, 1985.

4.  Hayes-Roth, F., Waterman, D. A., and Lenat, D. B., Building Expert Systems, Addison-Wesley, Reading, MA, 1983.

5.  Holsapple, C. W., and Whinston, A. B., Business Expert Systems, Irwin, Homewood, IL, 1987.

6.  Liebowitz, J., and De Salvo, D. A., Structured Expert Systems: Domain, Design, and Development, Prentice-Hall, Englewood Cliffs, NJ, 1989.

7.  Waterman, D. A., A Guide to Expert Systems, Addison-Wesley, Reading, MA, 1986.

8.  Zahedi F., Intelligent Systems for Business: Expert Systems with Neural Networks, Wadsworth, Belmont, CA, 1993.