Algorithm Models of Production Decision Making and Adaptation Cost Evaluations

Sergei Perminov
Russian Academy of Sciences
Perminov@serv2.cemi.rssi.ru

Abstract

Adaptation factors are not properly involved yet in decision making and not measured regularly in economic terms. In the paper we mean under adaptation cost an opportunity cost which one has to bear using a certain inefficient adaptation system compare to the better system. Adaptation cost as a parameter may be operationally and efficiently used in considerations how to improve a control. Adaptation cost is really hidden and unlike ordinary costs it requires a special methodology and techniques to be revealed.

The study is focused on a production process considered as a perpetual flow of resources being processed stage by stage towards final products and customer satisfaction. Each action that takes place is a result of control decisions pursuing certain targets. The contra posing of targets and control decisions plays the crucial role in the analysis. It is assumed that control decisions are:

• establishing a compromise between many interests formulated in terms of targets,
• thought over in accordance with the concept of `bounded rationality',
• probabilistic,
• taken as a result of adaptation process.

The problem as a whole originated from the fact that an economic environment is uncertain in principle though their probability distributions can be estimated. Therefore a production structure has to adapt to both changing targets and a metamorphosing environment (disruptions of resource inputs, technological changes, etc.). The ability to adapt to these changes characterizes a production structure’s effectiveness. We deal with two kinds of uncertainty. The external one (targets and environment) and the internal uncertainty actually generated within a system as a result of many feedback contours performed over time.

The computer simulation model of a decision process is formulated and analyzed. Its basic parameters are priority rules and their generalizations - random preferences. The permanent comparison of the targets and the actual levels is the essential part of the adaptation process. This provides signals that reflect how efficient activities in the current situation are. These signals define the direction in which the current random preferences must be corrected.

When all signals have the equal intensity the fixed point of the adaptation process (i.e., a certain compromise of interests) is achieved and there is no possibility for further improvements.

The adaptation process is globally stable. So starting from any initial random preference it converges to the same point that is adequate to the given situation. The strict conditions for this process can be also defined.

The adaptation process shows that a decision maker in absence of external disturbances is experienced afterwards to take optimal decisions. But in a real life it can be only a tendency because the economic environment changes over time. Variations of master plans, targets, resource inputs and other parameters modify the notion of optimum. So before reaching the optimum a decision maker starts to adapt to the different situation. It is natural. It other case it would be no need in a permanent adaptation to ``moving" targets in a turbulent environment. The base for adaptation cost measurements is a deviation of actual performance from targets. The practical reason to use adaptation cost is the necessity to have an instrument to evaluate in the traditional economic manner any local improvements in the adaptation mechanism.

For example, we can undertake a management improvement - replace tone production structure by another one that reduces production cost. For instance, we plan to put a new production control system. The operational advantage of our considerations here is the ability to evaluate the profitability of such an improvement.

Since a description of any real production structure is complicated significantly the ``module" principle has to be applied. We assume a real production structure consists of standard modules, i.e., elementary adaptation activities. If we learn to analyze them properly we become capable to probe various production structures.

The major causal factors, known as cost drivers, and the related technique of analysis are considered below.

Feedback mechanism is defined by periodicity and a structure of signals. The `direct' cost of each feedback signal is easily calculated since we know the cost of computer and communication systems used, labor employed, etc. But the `hidden' cost appears often more significant. If we evaluate actual performance levels too often (before the estimations become representative enough) we take risk to correct our preferences in a wrong direction. In other hand, to wait too long and generate feedback signal beyond reason rarely is also harmful because we keep taking control decisions using probably non-adequate random preferences.

Stereotypes (stock activities) and inertia in decision making. The essential part of learning and adaptation is the ability to focus on stereotypes, i.e., ``stock activities" that are currently the most efficient. This essentially reduces a decision’s dispersion but at the same time increases an inertia in case of any changes in the environment or ``movements" of targets. The ability to redistribute resources in order to eliminate bottlenecks plays the crucial role in production control. The proposed approach offers a tool to evaluate improvements of this kind.

Therefore we can evaluate ( for the certain economic environment and targets) the contribution of any controller as a production structure's element. It opens the real possibility of consecutive improvements, i.e., assembling of a production structure from ``modules".

The practical necessity of these evaluations visible in terms of activity- based costing. We can evaluate its positive effect as the adaptation cost reduction and compare with expenses to be born to maintain this controller.