Converting Point Estimates for Cost-Risk Analysis

The estimating process for most cost estimates follows the approach of determining the cost of each step in the production process, which is considered the "most likely" cost, and summing the cost of each of the steps, or "rolling up" the costs to estimate the total cost of the product. An amount of profit is then determined based upon the total cost to result in a quote to the perspective customer. It is generally accepted that the actual costs are typically higher than costs used in preparing the quote as the estimate is often low as some items have been omitted from the estimate, the estimating data used is not up-to-date, or the costs increase during the time between the estimate and the production of the product and delivery to the customer. A cost-risk analysis would help the supplier to better evaluate the risk to achieve the cost utilized in the preparation of the quote. The solution to the problem would be to perform a formal cost-risk analysis. The formal cost-risk analysis would require that probability distributions be developed for each step of the process, developing correlations among these distributions, and sum the distributions statistically usually via the Monte Carlo simulation process. This would be a very expensive undertaking and may be cost effective for the aerospace and defense contractors, but it would not be cost effective to most medium and small businesses. A simplified model was developed by Stephen Book for those who want the benefits of a cost-risk analysis, but cannot afford the cost or time to perform a formal cost risk analysis. The model has been modified to develop a Cost "S" Curve from the traditional point estimate value based upon the triangular distribution and using three parameters, H/L ratio, the percentile value for the point estimate and the percentile value for the most likely cost. This approach eliminates the need for the traditional triangular distribution parameters of the high with a specified percentile, the low with a specified percentile, and the mode. It is difficult to get estimates of the high and low values associated with percentiles, whereas the H/L ratio is easier to obtain for estimates. The results from the model include the lowest cost, the most likely cost, the median cost, the mean cost, and the highest cost estimate as well as the cost values over the entire range in five percentile increments. The model has also been modified to develop a Cost "S" curve from the traditional point estimate value based upon the normal distribution using the H/L ratio and the percentile for the point estimate. The primary assumption made is that the distance between the high and low values is six sigma. The results from the model include the lowest cost, the mean cost, the highest cost as well as the cost values over the entire range in five percentile increments. The normal distribution is preferred for estimating the mean of a sum of components, but it is not necessarily a good estimate about the distribution.