Modeling brings the power of mathematical abstraction to domain reality to solve problems.
The art of modeling involves articulation of the decision problem, understanding the domain reality, and translating the abstracted solution provided by the decision model to concrete reality. At the heart of decision modeling is the domain expertise that lies largely with the decision-makers. Tools and technology by themselves can seldom identify the problem, key drivers, choose the right modeling approach, and discover qualitative frameworks/relationships (preferences, experience etc) from the given data.
The science of modeling involves converting domain reality- quantitative, and qualitative (like ethics, preferences, experience) to mathematical abstraction, using quantitative tools, and providing solutions as abstracted reality. The ultimate objective is to give quantitative expression to the decision maker's expertise.
Meaningful and relevant modeling is a product of close and continuous interaction between the decision maker and the modeling experts.
Modeling aids decision making through
Logical structuring of decision problem:
"It isn't that they can't see the solution.
It is that they can't see the problem."
The first step in this process is to ask "What is the Question"? Once the problem is clearly posed, the next step is to identify the key variables or parameters necessary for answering the question. With these inputs, the data that is relevant is specified, and at the same time a framework for a model emerges. One is then ready to use state-of-the-art mathematical and computational tools.
Identification of pertinent data:
"A theory has only the alternative of being
right or wrong. A model has a third possibility: it may be right, but
Domain expertise and sound knowledge of modeling tools are keys to selecting the relevant data for a problem. This is illustrated with an example. Consider modeling the evolution of inflation rates. A multivariate model is used to predict inflation rates based on relevant independent variables. Decisions involved are how many independent variables are to be chosen, and what should be the period over which the data needs to be examined. The trade-off is model complexity (too much data) versus model relevance (too little data).
Intelligent assessment of alternatives:
"Technical skill is mastery of complexity
while creativity is mastery of simplicity."
A model would necessarily make assumptions. These have to be tested. Alternatives emerge by changing the assumptions either drastically or in a gradual manner. In either case, different scenarios get generated with different probabilities. Intelligent assessment on the part of the decision maker would involve risk vs. return policy of the organization.
A model enables
1. Testing of hypotheses - reconciliation of intuitive understanding
with mathematical formulations
"The purpose of models is not to fit the
data but to sharpen the questions."
The first steps:
Building advanced models relies a great deal on computing power. Availability of low cost computing is a relatively recent phenomenon. In the absence of availability of low cost computing, modeling in organizations did not receive enough attention. And without any plans for formal modeling, the organization did not focus much on data collection and archival strategies. In most organizations this is the first challenge and the first benefit of formal modeling exercise. It forces organizations to identify strategies to collect and archive pertinent and clean data.
It is often useful to begin initial modeling exercises with the existing data. While this phase may not lead to major value creation, it forces one to define and identify the pertinent data, fine-tunes the existing data collection mechanisms to ensure availability of clean data. Besides, to derive the power of modeling, one cannot escape the first step.
Decision models are vital tools for robust decision-making.
2000-10 DecisionCraft Inc.