Systems Analysis: Golubkov

"The Systems Approach in Long-Range Planning" by Golubkov, Raizenberg, and Pekarsky, 1975 (in Russian).

Systems analysis comprises a set of scientific methods and practical techniques for solving complex problems (technical, natural science, economic, socio-political, and others).

Above all, it employs the concept of a system as a unity of interrelated elements acting jointly to achieve a common goal. The properties possessed by a system as a whole differ from the properties of its constituent elements. Any system is characterized by its own specific regularities that do not follow directly from the modes of action of its elements.

The main parts of a system are the input, the process (structure), and the output. The input of a system is a complex concept. On the one hand, it is the substance that enters the system and undergoes certain transformations within it. On the other hand, it is the external (surrounding) environment, that is, the totality of factors and phenomena affecting the system (natural conditions, foreign policy situation, market conditions). Input also refers to the established modes of operation for the system's elements, for example, instructions, regulations, and orders that define the procedures, rules, constraints, and goals of the system's operation.

The second part of a system is its internal structure, that is, the channels through which matter, energy, and information entering the system via its inputs pass, and the processes or operations that transform them (for economic systems, these are the processes of reproduction of material, labor, and financial resources).

The third part of a system is its output, the product or result of its activity.

Based on the degree of interaction with the external environment, systems are divided into open and closed ones. Open systems intensively exchange matter, energy, or information with their environment, while closed, isolated systems function with relatively little exchange (for example, a closed cycle of processing planning and economic information, or a closed technological cycle). It should be noted that a system may be closed or open with respect to matter, energy, or information separately. A system has either natural boundaries or boundaries established based on the research objectives; in the latter case, the system boundaries are defined by the researcher.

The same object can be represented as several different systems. If a manufacturing enterprise is viewed as an aggregate of machines, technological processes, materials, and products processed on those machines, the enterprise appears as a technological system. The enterprise can be viewed from a different perspective: examining the workers' relationship to the means of production, their participation in the labor process and the distribution of its results, the enterprise's place in the broader economic system, and so forth. In this case, the enterprise is an economic system.

In the course of social development and the scientific and technological revolution, a new object of research in the field of management emerged, known as a large-scale system. Generalizing the viewpoints of various authors, the following most important features of large-scale systems can be identified:

• the presence of subsystems with clearly expressed local properties, which together constitute the large-scale system;
• the goal-directedness and controllability of the system, that is, the existence of a common goal and general purpose, set and adjusted within higher-level systems;
• a complex hierarchical (multi-level) organizational structure of the system, providing for a combination of centralized control with the autonomy of parts (subsystems), the presence of vertical links between elements of different levels, and horizontal links between elements of the same level;
• the large size of the system, that is, a large number of elements, inputs, and outputs, and a diversity of functions;
• integrity and complexity of behavior, complex intertwining relationships among variables, and feedback loops causing a change in one variable to entail changes in many others.

Systems analysis provides special techniques by which a large-scale system, difficult to examine as a whole, can be divided into a number of smaller interacting systems, or subsystems.

Among the important concepts of systems analysis is the fundamental cybernetic concept of feedback. It was precisely this concept that helped establish the similarity between the organization of control in qualitatively different systems. Feedback means the existence of a communication channel between the output and the input of a system, either direct or through other elements of the system (for example, through a control unit). By means of feedback, data on the functioning of the controlled system are transmitted from its output to the control system. There, these data are compared with the data specifying the content and scope of work (for example, with the plan). In case of a discrepancy between the actual and the specified state of the system, measures are developed to eliminate this discrepancy.

An important concept used in systems analysis is the concept of uncertainty. The factor of uncertainty is present in the solution of many complex problems in the area of national economic planning and management.

Uncertainty encountered in the course of systems analysis may result from insufficient knowledge of the phenomenon being studied, or may be due to the fact that the consequences of decisions taken manifest themselves over a sufficiently long period and cannot be predicted with sufficient accuracy.

Other causes of uncertainty include the impossibility of quantitative assessment of many phenomena, in particular the difficulty of quantitative interpretation of the concepts "better–worse," "more–less" in decision making, imprecise problem formulation, and individual characteristics of decision makers.

Systems analysis is characterized primarily not by a specific scientific formalism, but by an ordered, logically grounded approach to the study of problems. With the aid of systems analysis methods, complex systems and situations are investigated, primarily those associated with the need to determine goals, objectives, and courses of action.

Systems analysis represents an aggregate of methods applied in the investigation of possible approaches to solving complex problems and simultaneously serves as a means of ordering and more effectively utilizing the knowledge, judgments, and intuition of specialists in the process of decision-making on these problems. In other words, systems analysis is a scientific tool for researchers and decision-makers.

The task of systems analysis consists of clarifying the problems facing decision makers in such a way that the executive becomes aware of all the principal consequences of decisions and can take them into account in his actions. Quantitative analysis helps the person responsible for a decision to approach the evaluation of possible courses of action more rigorously and to select the best one, taking into account additional, non-formalizable factors and considerations that may be unknown to the specialists preparing the decision (the systems analysts). Systems analysis serves to determine the most realistic ways of solving the problems that have arisen, ensuring maximum satisfaction of the requirements posed.

The experience of applying systems analysis in military affairs had a great influence on its development as a methodology, which left a certain imprint on the terminology of this and closely related scientific disciplines (operations research, systems engineering).

Between individual directions of systems analysis bearing different names, there often exist no fundamental differences or clearly delineated boundaries; at the same time, each of these directions is characterized by certain distinctive features and nuances. Thus, operations research is usually understood as the science concerned with developing quantitative recommendations necessary for the planning and organization of operations. An operation here is understood as any goal-directed action of a person, a group of people, or a human–machine system. Systems engineering creates methods for the synthesis of large-scale systems based on the study of the functioning of their individual elements. Although many methods and results of systems engineering have been obtained in application to technical systems, they can, to a significant extent, be transferred to objects of a different nature, including economic ones.

The emergence of systems analysis in the methodology of substantiating management decisions marks a transition from solving well-structured, formalizable problems (where goals, paths, and criteria are clearly defined, achievable through operations research methods) to solving ill-structured problems that arise under conditions of uncertainty and contain non-formalizable elements not translatable into the language of mathematics.

The operations research specialist employs methods of mathematical or logical analysis under conditions where there is a clear understanding of what constitutes "more effective" performance. He rarely delves into defining the goal of the work or the methods for evaluating its success; this is one of the main tasks of the systems analyst.

A planning specialist virtually always acts as a systems analyst, since he sees the problem being solved as a whole. The results of systems analysis depend on the adopted system of criteria. The choice of criteria is an integral part of the analysis. The role of the analysis consists of establishing which criteria are advisable to use for making certain decisions in the area of management.

One variety of operational analysis, known as cost-effectiveness analysis, consists of a comprehensive comparison of the effect of implementing one or another course of action with the resources required for its implementation.

The difference between systems analysis and cost-effectiveness analysis lies in the fact that in the latter case, the analysis is directed at finding rational ways to use resources under conditions of a formulated, most often given, goal. Systems analysis poses the problem more broadly, treating the goal itself as an object of choice made in accordance with certain principles.

The methodology of systems analysis can be defined as an aggregate of its principles, that is, the most general, fundamental regularities governing the conduct of specific analyses of various systems and the processes occurring within them.

The defining principle of systems analysis is goal-directedness. Every system exists and develops in accordance with goals set externally or actively formed by the system itself. From the standpoint of systems analysis, it is the totality of goals that defines and delineates the system, unifying the system and its activities into a single whole.

The next principle of systems analysis follows from the necessity of applying a systems approach and consists of identifying and investigating all essential interrelationships both within the system and between the system and its external environment, as well as in selecting particular solutions, taking into account their influence on the system as a whole.

In systems analysis, one must bear in mind that the same goals can be achieved through the use of several different means and methods. From this follows a principle of systems analysis that consists of seeking several alternative solutions to the problems that have arisen. Examining possible variants of a system's structure, its organization, and the processes occurring within it (technological, planning, management, and others), with the aim of selecting the best among them, is a critical feature of systems analysis.

Often, the main task of systems analysis is finding optimal solutions. Optimal solutions are those that, based on one or another criterion, are preferable to others. Typically, in selecting an optimal solution based on some criterion, the effect obtained from implementing a particular strategy of behavior is compared with the resource expenditure required for its implementation.

The depth and completeness of analysis of processes occurring in various systems are largely determined by the degree to which the dynamic nature of their functioning and development is studied. Systems analysis proceeds from the necessity of understanding not only the static interrelationships in a system but also the dynamic interactions, of studying not only the state of the system at a given moment but also its change over time. Examination of phenomena in their dynamics and development is also an important principle of systems analysis.

The principle of separation of responsibility for the recommendations resulting from systems analysis and the decisions made on their basis presupposes the establishment of a clearly delineated set of rights and responsibilities for the systems analysts and the decision makers who act on the results of the analysis. In ultimately approving a decision, those responsible for it may take into account, alongside the recommendations resulting from systems analysis, a number of additional quantitative and qualitative considerations that were not accounted for by the analysts.

In cases where a complex system is being studied, the analysis of which can only be carried out by generalizing the results of analyzing its constituent parts (subsystems, elements), the structure of the system (problem) under analysis and the possibilities for dividing it into elements are studied first. Then, a sequential decomposition (structuring) of the system into its constituent parts is performed in accordance with its structure, functions, and internal processes, taking into account all the most important interrelationships. The principle of goal-directed structuring of the system under study finds concrete practical embodiment in a number of systems analysis methods, for example, in the goal tree method.

The domain of application of systems analysis, depending on the complexity of the problems being solved and the feasibility of using mathematical methods, can be roughly divided into five levels:

• optimization of the planning and management of individual operations;
• selection of the type of system or the most effective paths to achieving stated goals;
• development of new systems;
• determination of the place and role of a system in higher-level problems;
• planning and management of an economic system at the national economic level.

This ordering corresponds to increasing problem complexity, level of uncertainty, and difficulty in developing recommendations on which concrete action can be taken. At the first level, analysis acquires a mathematical form to the greatest extent and can be based on the use of formal-logical tools and methods. In essence, this is operations research aimed at increasing system effectiveness under conditions where the concept of "more effective" is clearly defined. The main characteristic of problems in this domain is that they have a clearly delineated structure amenable to formalized description. At the remaining levels, the application of systems analysis is necessarily connected with the need to investigate qualitative factors and conditions, most often of a social nature.

In certain cases, systems analysis is inevitably connected with the need to study contradictions. Examples of contradictions and inconsistencies include unbalanced plans, disruptions in the supply of certain types of materials and equipment, discrepancies between the growth in the volume of production and consumption of certain product types, and contradictions associated with the slow adoption of new technology.

Analysis of problems at the second and third levels includes the design of new systems intended for better performance of known operations or operations never previously performed. When considering problems at this and subsequent levels, the consideration of social and political factors becomes increasingly important.

"The Use of Systems Analysis in Sectoral Planning" by E.P. Golubkov, 1977 (in Russian).

The general method for solving specific problems of systems analysis is differentiated and embodied in a great variety of methods, which, depending on the classification principle adopted, can be divided into different groups — for example, into methods of analysis and synthesis, descriptive and experimental methods. Depending on the degree to which formal elements are used in systems analysis, three groups of methods can be distinguished: mathematical (formal), heuristic (informal), and combined mathematical-heuristic methods.

Systems analysis methods can be considered both in a broad and in a narrow, specific sense. In the broad sense, systems analysis methods include any methods from the three groups listed above that are applied to solve the problem at hand on the basis of a systems approach. In this broad understanding, systems analysis methods are identical to problem-solving methods and can be classified in accordance with the individual stages of the decision-making process.

In the narrow sense, systems analysis methods are specific methods designed for determining the totality of goals of a system's activity and the best paths to their achievement, for choosing models and criteria, for the sequential, directed decomposition of a system (problem) into its constituent elements, for determining the interrelationships and interdependencies among these elements, and for determining the relative significance (preference) of individual goals, measures, criteria, and models. Thus, these are primarily methods aimed at solving ill-structured problems in cases where the application of formal, mathematical methods is limited.

The main body of mathematical methods in systems analysis consists of operations research methods (various types of programming, game theory, queuing theory, and others). These methods are widely applied in systems analysis, but rather for studying individual aspects of the problem being solved than for investigating its essence.

Mathematical methods in systems analysis are most often used:

• for determining numerical values of indicators characterizing the results of a system's functioning;
• for finding the best courses of action leading to the achievement of specific results (optimization);
• for processing and analyzing data of a heuristic, creative nature.

Despite the increasing role of mathematical methods in solving economic problems, they cannot be regarded as a universal means of solution. Methods based on accumulated experience and intuition, that is, heuristic (informal) methods, remain essential.

The desire to obtain precise answers to the questions posed using mathematical methods can lead the engineer and the economist to become so engrossed in their work and the methodology for identifying research needs that the ultimate result of their efforts turns out to be a conclusion about the need for additional studies.

Meanwhile, under conditions where certain data about one or another economic process, phenomenon, or result are absent, it is better to obtain approximate answers to the most important questions than to attempt to give precise answers to questions that are not fully clear and understood. The procedures for formulating system goals, alternatives for their implementation, models, and criteria cannot be fully formalized.

The conduct of such expert reviews and conferences is very often regulated by tradition, that is, ultimately by experience, and in many respects constitutes an art. However, gradually, various mathematical methods for processing source material of heuristic origin are beginning to penetrate this domain as well.

A distinctive feature of heuristic methods is that the specialist, in assessing an event, relies to a significant extent on information based on his experience and intuition. This information is enormously dependent on the personal qualities of the expert. When using the intuitive method of decision making, the expert is unable to articulate the sequence of its implementation in words.

Intuition presupposes a direct, leap-like attainment of a solution instead of a chain of careful, well-thought-out and deliberated steps. An intuitively thinking person often cannot report which aspects of the situation were identified in the process of perception, what portion of the information stored in his memory he used, or what "reasoning" led him from the initial data to the decision.

Experience-based heuristic methods of decision making proceed from a past analysis of a problem identical to the one that has arisen at the present moment (a decision based on analogy with the past). Informal methods also include decision-making based on appeal to authority (seeking advice and guidance from management, experts, or reference sources, such as scientific and educational literature).

Application areas of systems analysis methods:

1. Determining the set of goals and the paths to their achievement (for example, defining the goals of a sectoral plan and developing several alternatives for its implementation).
2. Determining the preference (ranking) of individual goals, paths, measures, results, and so forth (for example, determining the priority and timing for implementing individual measures in a sectoral plan for new technology).
3. Decomposing goals, programs, plans, and so forth into their elements (for example, compiling plans for individual associations or enterprises based on the tasks set for the industry as a whole).
4. Selecting the best paths to achieving stated goals (for example, analyzing and comparing various organizational and technical measures proposed for inclusion in the plan).
5. Selecting criteria for comparing goals and paths to their achievement (for example, selecting criteria for evaluating the effectiveness of various sectoral plan alternatives).
6. Constructing models for selecting goals and paths to their achievement (for example, a model for distributing plan assignments among enterprises of an industry by the criterion of minimizing reduced costs).
7. Generalizing the data of the analysis of individual subsystem functioning to draw conclusions about the optimality of the system's functioning as a whole (for example, compiling an optimal scheme for developing sectoral plans based on the analysis of planning procedures adopted in individual associations or enterprises of the industry).

It should be noted that the boundary between heuristic and combined methods is to some extent conventional, since there is no strict demarcation between them. At present, mathematical techniques for processing source data of a heuristic nature are used in the application of almost all heuristic methods.

In systems analysis, depending on the degree of certainty in the formulation of problems and the conditions for their solution, there may be various types of problems that predetermine the use of one or another specific solution method. These problems, using the classification employed in operations research, can be divided into three groups, although certain clarifications are also necessary here.

1. Deterministic problems. These are problems of selecting the best solution alternative in situations where each strategy leads to a unique outcome (for example, a given plan alternative leads to obtaining the required result with 100 percent probability).
2. Probabilistic problems. These are problems of selecting the best solution alternative in situations where each action may yield different outcomes whose probabilities are known or can be estimated (for example, each plan alternative leads to obtaining known results with a certain probability).
3. Problems under uncertainty. These are problems of selecting the best solution alternative in situations where the probabilities of obtaining different outcomes are unknown, or it is entirely unknown what results may be obtained when choosing one or another strategy from among those under consideration.

The introduction of such terminology for systems analysis problems is to some extent conventional, since, for example, the factor of uncertainty is present even in the solution of probabilistic problems.

It should be borne in mind that the class of probabilistic problems and problems under uncertainty also includes problems for which a complete set of possible strategies has not been defined, and only known strategies are considered. Such a situation is characteristic of decision-making practice, where not all possible alternatives for solving the problem that has arisen are considered, but only a limited number of such alternatives.

Deterministic and uncertainty problems can be considered as limiting cases (for example, complete knowledge and complete ignorance of outcomes) of probabilistic problems. For this reason, we shall first consider probabilistic problems, after which we shall proceed to the consideration of deterministic and uncertainty problems.

In solving probabilistic problems, the selection of the best solution alternative is carried out by maximizing the expected value of the degree of preference. This criterion is applied in solving the majority of problems of this type.

In cases where certain calculations, for example techno-economic ones, make it possible to determine the effectiveness of one or another strategy under various conditions of the external environment, concrete values characterizing the effect of implementing one or another solution alternative must be used instead of the values of the degree of preference. Final recommendations on the choice of a solution should be developed, taking into account the probability of obtaining different outcomes for the solution alternatives under consideration.

A deterministic problem can be regarded as a limiting case of a probabilistic problem, assuming that the probability of obtaining each of the possible outcomes equals either one or zero.

Examples of deterministic problems include, for instance, linear programming problems.

The criteria applied in solving such problems may represent a limiting case of maximizing the expected value of the degree of preference, where the probabilities of realizing possible outcomes equal either one or zero. Various indicators having a concrete economic meaning (reduced costs, profitability level, labor productivity, and others) are often used as criteria in solving deterministic problems.

The most characteristic type of systems analysis problems is problems under uncertainty. In real cases, when it initially appears that any probability estimates for achieving various outcomes are absent, the systems analyst usually makes maximum effort to obtain information about these probabilities by conducting a special study and, as a rule, succeeds. However, cases are also possible where probability estimates are entirely unknown.

In solving problems under uncertainty, such well-known criteria as minimax, maximin, generalized maximin (the Hurwicz criterion), the minimax regret criterion (the Savage criterion), the Laplace criterion, and others are applied.

The use of different criteria in solving a single problem, as a rule, leads to obtaining dissimilar results. There are two approaches to selecting criteria for solving problems under uncertainty. The first is the development of new criteria or requirements for choosing a decision criterion. The second approach consists of using any, even the most meager, information about the probabilities of different environmental conditions being realized (different outcomes obtained from implementing one or another strategy) or in conducting experiments to obtain estimates of these probabilities. Thereby, the uncertainty problem becomes a probabilistic one.

Both approaches are labor-intensive and, as a rule, difficult to implement in practice; however, the second approach is preferable. The first approach leads to searches for new criteria for choosing the best among those known, then to searches for criteria for choosing criteria from among those under consideration, and so on. In other words, there is no decision criterion not based on probability estimates that would satisfy certain justified requirements for a "good" criterion.

Attempts to formulate a criterion for evaluating possible decisions under uncertainty reflect the aspiration to make the advantages and disadvantages of each course of action under various conditions more transparent. None of the existing criteria for choosing decisions under uncertainty is universal, capable of satisfying any decision maker.

Systems analysis methods should not be set in opposition to one another. Each has advantages and disadvantages, but none is suitable for solving all problems. Therefore, the best results can be obtained by combining several methods, as determined by the nature of the problem being solved and the level at which it is examined. When solving planning problems at the lowest level of economic management, one typically encounters sufficiently well-structured problems, which allow for the broad application of mathematical methods of analysis. At higher levels, goals and other elements of systems analysis acquire an increasingly qualitative character, and therefore the importance of informal methods grows.

The complexity of modeling social processes in economic systems further complicates the application of mathematical methods. At the same time, the role of the uncertainty factor increases, and ignoring it through formal methods may lead to erroneous conclusions.