Avoiding Pitfalls When Applying Learning to your Estimate

Methods and Models I Track



Most cost estimators understand the basics of learning curves, but we sometimes fail to recognize the implications of the way that we apply learning.
This presentation points out some major pitfalls and mistakes that can be made when dealing with learning. We address both deriving learning curves from historical data and applying learning to a cost estimate. Additionally, we recommend best practices to use when dealing with these issues.
Some of the issues discussed are:
• The end of learning. What happens when learning is applied to a large number of units, perhaps in the tens of thousands or more? Does the cost continue to decrease? If so, how should we as cost estimators approach estimating a large quantity of units with a learning curve? Also, how does this threshold vary with the commodity being estimated?

• The effect of fixed costs: Often we can see the cost of an item only at the top level. Even though we know part of the unit has a fixed cost, we can’t separate that portion, so we erroneously apply learning to the entire end item. We examine the effects of applying learning to an item that includes fixed costs, and recommend an approach to improve the accuracy of the derived learning curve.

• Sums of learning curves: What is the result when learning is applied to several elements within a work breakdown structure or cost element structure? We derive a relationship between the learning curve slopes of children elements and the associated parent.

• Shared learning: How does one apply learning to a group of items that are similar but not identical, such as a family of vehicles? We explore this scenario and recommend a methodology for applying “shared learning” to a group of similar items.


Andrew Busick
Technomics, Inc.
Andrew Busick is a Cost Analysts with Technomics, Inc. He has over five years of cost estimating and analysis experience with myriad skills in modeling and simulation, descriptive/inferential statistics, linear programming, and constrained optimization. Andrew currently supports both the Army Tank Automotive Command (TACOM) and Navy Integrated Warfare Systems (IWS) Program Office.
Prior to joining Technomics, Andrew was a cost analyst at LMI (formerly Logistics Management Institute). Andrew earned his B.A. in Mathematics and Economics from the University of Virginia in 2007 and is a Professional Cost Estimator/Analyst (PCEA).