There are three basic methods used for cost estimation — the industrial engineering, analogy, and statistical approaches. The statistical method is considered the most useful for government cost analysts, whether the purpose is long-range planning or contract negotiation. An estimating procedure must rely on a database that includes cost, physical and performance descriptions, and a development and production history of previous equipment programs. Cost is predicted by estimating relationships based on parametric explanatory variables such as weight, speed, power, frequency, or thrust. Often these variables are examined through the use of statistical procedures. Experience has proved the value of the learning-curve concept. Its basis is that each time the total quantity of items produced doubles, the cost per item is reduced to a constant percentage of its previous cost. A thorough knowledge of the learning-curve phenomenon is indispensable to persons involved in cost analysis.