The Checkmark Function for Input Variable Correction
Howarth, D.
Lockheed Martin Aeronautics Company

This paper describes a method for correcting mass inputs to parametric cost equations, thus improving cost estimation accuracy.

Parametric cost models do not always produce estimates that match eventual outcomes. Quite often parametric cost approximations run lower than actual expenditures. However, history shows that this may not be the fault of the estimating models used to generate projections. From initial description to final delivery, product configurations may change significantly. Perhaps the most important of these changes relate to project sizes, as often measured by empty product masses for a broad range of military and commercial manufactured goods. Realized product masses frequently do not match initial predictions. Mass properties groups, using their own parametric guidelines, are responsible for developing mass estimates. Often these organizations reduce their estimates of project sizes concurrent with their delivery to parametric cost estimating groups. Once programs begin, product masses usually creep upward until the products undergo their final weigh-ins, which reveals their true weights, which are typically higher than initially forecasted.

Parametric cost estimators aware of discrepancies between initial size projections and final realized masses broadly have two options: They can either use the initial mass properties provided to them in their cost models, given that they have formed and used cost estimating relationships based on like initial masses from previous endeavors, or they can attempt to project final masses from initial masses, and then use these revised inputs in their equations based on final masses from other programs. In either event, they need to acquire relevant program mass histories over time.

This paper examines the mass histories of 13 different programs running in duration from roughly two to six years, with ultimate empty weights ranging from under 1,000 to over 300,000 pounds. Normalizing each program to a constant span measured in percent provides a common basis for weight growth analysis.

Data reveals that initial masses seldom closely match ultimate masses. Further, the same data indicates that smaller programs demonstrate proportionally more growth from their initial to final mass estimate than do larger programs. Importantly, however, there is a very strong correlation (> 99.8%) between the initially predicted and ultimately realized program sizes using initial mass as an independent variable. Thus, we can use initial mass to more accurately predict final mass by deriving a statistically significant equation that reveals and corrects the organizational downward bias of weight estimates. This removes an important input error and brings parametric cost estimates closer to realized expenditures.