Cost estimating methods: analogy, parametric and bottom-up
Analogy adjusts the actual cost of a similar past system. Parametric applies a statistical relationship between cost and characteristics such as weight or lines of code. Engineering build-up, also called bottom-up, estimates each lowest-level WBS element and adds them up.
Analogy
An analogy uses the actual cost of a similar program and adjusts it for differences. GAO lists it for when few data are available, for rough-order-of-magnitude estimates and as a cross-check. It is quick and cheap but rests on a single data point.
Adjustments should be objective and tied to factors such as size, performance, technology or complexity. GAO's example: an estimator who raises a cost by 20 percent because the new item seems more complex, without saying why, has made an unacceptable adjustment. The uplift must link to a parameter, such as 20 percent more processing capacity or weight. GAO's engine example scales $5.2 million by thrust, 16,000 lb over 12,000 lb, to about $6.9 million.
Parametric
The parametric method relates cost to physical or performance characteristics through a statistical relationship built from many past programs. GAO calls it a top-down approach, used when little is known beyond a few key characteristics. It suits budget estimates and design trade studies.
The new item must fall within, or not far outside, the data behind the relationship. A relationship built on software of 10,000 to 250,000 lines of code should not be used to estimate a program of 1 million lines.
Engineering build-up (bottom-up)
Engineering build-up estimates labour and materials at the lowest level of the WBS and rolls them up. It needs a detailed design and a parts list, is slow, and is easy to audit. GAO says it is normally used in production, when the configuration is stable and actual cost data exist.
GAO adds that the method depends on where the program is in its life cycle, and that several methods are usually used over a program's life. Other methods include expert opinion, extrapolation from actual costs and learning curves.
Where the exam catches people
Believing an analogy adjustment can rest on judgement alone: it must be tied to a measurable difference between the systems.
Believing a parametric model works at any scale: the new item must fall within or close to the model's data range.
Believing bottom-up estimating suits an early concept: it needs a detailed design and is used mainly once the configuration is stable.
Believing parametric estimating is bottom-up: GAO calls it a top-down approach.
Believing analogy draws on many programs: it relies on one comparable system, while parametric uses many.