Evaluating and Minimizing Cost Estimating Bias Early in Program Development

Developing robust cost estimates early in the acquisition life cycle is difficult. Mission needs are driving numerous concept architectures that are circulating within the minds of stakeholders, unknowns flourish throughout the solution space, budgets are being formulated and everyone wants immediate answers. Within the thought leader communities' there is optimism (for a perfect solution), risk aversion and/or ignorance, new or enhanced system capability and latent estimating bias. Arriving at a balanced, low, or non-bias solutions and cost estimates that accurately reflect system scope, critical technologies, risks, schedules, and realistic and defendable costs that provide a viable program and stand the test of time are difficult to develop. Recognizing biases among stake holders is not always apparent and can impact objective cost estimate outcomes (will cost) from desired outputs or driven by budget-based estimates that are in the minds of thought leaders (should cost). Obtaining robust cost estimates requires following a rigorous process culminating in an initial point estimate, then expanding it with uncertainties, risks, and other factors that result in, simple averaging, analogies and perhaps the mean of a Monte Carlo solution result among other probabilistic solutions. In all cases, the ground rules, and assumptions, often justified, may be biased with expert opinion, a drive to meet notional budgets, a need to implement the latest system design approach or other influences. As a result, there may be implicit flaws that do not surface until much later in the lifecycle, leading to cost growth and schedule delays. This paper investigates common psychological bias types in decision making, assesses root causes, and evaluates resulting impacts to cost estimates; then we provide methods to identify and apply corrective action that can minimize bias within the initial cost estimate. We will assess historical cost growth in Department of Defense programs; then investigate some common behavioral model tools such as “Judge-Advisor System”, “Weight-of-Advice”, Averaging, and Overconfidence characteristics that can simultaneously influence an estimate outcome. Two examples will provide methods to implement tools that result in an approach to support development of non-bias, high quality, and defendable cost estimates early in program formulation. Furthermore, we will provide an approach to support clear alignment of mission scope, technology development and fielding schedules to the forecast cost estimates. Future work will investigate less common and unconventional biases, the Analytical Hierarchy Process, Bayesian models, and human versus algorithm advice that may influence robust and enduring cost estimates throughout the life cycle.