Copyright © 2009 Corvus International Inc. All Rights Reserved
Phillip G. Armour
Communications of the ACM - Personal information management, 2006
The height of what is now called "Mount Everest" was first measured between 1847 and 1849 (not only did it take time to walk around the mountain and measure it from different angles, there was some confusion about whether people were actually looking at the same thing). The survey was conducted by the then Surveyor General of India Andrew Scott Waugh.
There is an apocryphal tale that Sir Andrew was horrified that the calculated height of what the British then called "Peak XV" was exactly 29,000 feet.
The first printed map showing Peak XV as "Mount Everest"
showing survey lines of sight. Copyright (c) Royal Geographical Society
Divide by 100
For various reasons we humans accord both a great significance to numbers which can be divided by larger powers of our base unit 10. So becoming 100 years old is a big deal. Transitioning our calendar from 1999 to 2000 is also a big deal. Along with the perceived significance of 10n comes a certain assumption of precision. When someone says "it is a million miles away" we don't assume that it is precisely 1,000,000.00 miles, we assume it is somewhere close to that.
How we estimate
Software projects are often estimated by trying to identify all the tasks that the project team is expected to do and then summing them (using critical path analysis to determine schedule--if we are sophisticated). This is rather like measuring a mountain by counting boulders:
We could size a mountain by walking around and adding up all the boulders, rocks, and pebbles that make up the body of the mountain. The sum of boulders times the density of rock would give us the weight and general size and weight, the rest angle (how they lean together) would give us some idea of the size.
It would be very tedious to do this, which is one reasons surveyors don't do it.
A better alternative would be to take several measurements of distance and angles between points and to the peak and use arithmetical formulae to deduce the height of the mountain, and this is what surveyors do
Task-based versus Scope-based estimation
Task-based estimation is rather like counting boulders. There is nothing wrong with it, except it can be rather time consuming and tedious. So why do most project managers do this? Well, it is simply because, to a project manager, a project is a collection of dependent and independent tasks. Scope based estimation is more like the surveying method. From one or more assessments of the final scope of the delivered product, we use formulae to derive the scope (effort, budget, duration, staff) of the project.
But this does require that we can somehow derive the "size" of a system before we have built (or sometimes even specified) it. The answer is: with difficulty. However, the fact that we don't know how "big" a system will be when it is built is simply a function of the fact that when we estimate and plan a system we often don't know exactly what it is supposed to do and how it is supposed to do it. Therefore the "variance" in size is related to the variance in the system scope so if we can figure out how certain or uncertain we are about what the system has to do, we can figure out how likely or unlikely we are to produce something in a given amount of time and budget. There are tools available (see for instance SLIM-Estimate(R)) that can calculate this probability.
Mount Everest Redux
When Sir Andrew Waugh realized that his team had measured the height of Mont (sic) Everest (which he so named in 1856 after his predecessor George Everest (pronounced Eee-ver-est, not Ev-ver-est)), as being exactly 29,000 feet, he purportedly was concerned that people would think the measurements were "inaccurate". Then, according to the tale, he added an additional two feet to give the impression of accuracy that the round number would not. For many years the height of Mount Everest was quoted as 29,002 feet. The extra "2" being an ironic testament to the surveyors' art and the precision of their measurements... but at least he didn't count boulders.
Mt. Everest is currently measured at 29,035 ft. Is this evidence of scope creep?