Value Models

A value model is a scoring function. The input to a value model can be any number of scalars, vectors, discrete variables (like color = red, blue or green), yes / no answers, or any kind of quantity. The output is a single real number, so that results from the value model can be ranked from best to worst. Optimization literature descibes an objective function, the output of which is maximized or minimized by the optimization process. A value model is an objective function, and is directly used for optimization of the design of systems. The term "value model" was used for a complex, multi-attribute objective functions by Ralph Keeney in his dissertation under Howard Raiffa (Keeney, Ralph L., and Raiffa, Howard. Decisions with Multiple Objectives: Preferences and Value Tradeoffs. Wiley, New York, NY, 1976). The concept was expanded in his 1992 book, Value-Focused Thinking. Although early value models were developed as regression fits to preferences elicited from decision-makers, contemporary value models have elegant internal structures that utilize the laws of economics in a similar way to the use of physics in performance models.

To find the best design, your engineering team must have a way to evaluate designs, a way to measure their goodness. Of course, everyone knows that less weight, less fuel consumption, less cost and more reliability are good, but what about a design that is more reliable, consumes more fuel, weighs less and costs more? Trade factors are sometimes used to balance two properties, but to compare four or ten properties at once requires a full objective function, a mathematical expression that combines all the properties into a single measure of goodness.

Furthermore, the use of formal design optimization packages such as ISIGHT requires a formal definition of value, that is, an objective function.

To generate an objective function specific to your product, DFM Consulting can build a value model, an engineering economic model that estimates the price and profitability of your product in terms of design parameters, such as weight, efficiency, cost and reliability. The model incorporates a rigorous representation of the market and your customers' business, including the impact of your design on their revenues and costs. Probabilistic analysis is used to address demand uncertainties.

Value models are especially powerful for sensitivity studies. What does a one pound increase in weight do to the profitability of your product? How about a one percent improvement in efficiency? DFM Consulting value models make it easy to answer these questions.

Sensitivity results can then be used to price out research and development programs. Your model can be used to answer questions such as, "What is it worth to my business to increase turbine inlet temperature capability by 25ºF?"

A DFM value model can also serve as a custom objective function for your product. The objective function will reflect what your customers desire in terms that your engineering team can design to.