Saturday, April 20, 2013

Efficiency and Completion Time

For a long time I've been puzzled by why it seems to take much longer to do something than I first guess. I think I may now have an answer.

A typical task has three components: preparation, action, and luck. If we know exactly what to do, have the resources we need, and have no bad luck, then we can accomplish a task in a minimum amount of time. There is no preparation, and we will be 100% efficient in completing the task. If these conditions are not met, then in the same amount of time we will only accomplish a fraction of the task, with some of that time taken learning, acquiring resources, and dealing with the impact of more normal luck. In my experience, the first attempt at doing something within an ideal timeframe (what I call an "iteration") will result in the task being, at best, about 70% complete. If you're familiar with the normal probability (or "bell") curve, that's the area under the curve that's roughly within plus-or-minus one standard deviation of the mean.

Sometimes we don't even know that we haven't completed the task. With writing, for example, reading what I've written often uncovers problems with what I wrote. When the remaining amount of the task is identified, and if I have the opportunity to work on it, I may be able to knock out 70% of it on the second iteration, which began with the review that uncovered the remainder. This still leaves 9% of the original task left undone. In many situations, the 91% that I've accomplished may be good enough; for others, even that isn't acceptable.

On my third iteration, I will typically spend most of my time preparing: finding out what's left to do, and then getting what I need to do it. Once again, with typical luck, I'll at best complete 70% of what's left, driving the total up to more than 97%. In most situations where someone else decides what I'm working on and how long it should take, anything more than two iterations is a luxury (and I often have to take the second iteration out of my hide), with three iterations being the absolute maximum.

I am considerably worse at bowling than at writing and editing. I recently played five games following a 13 year hiatus. The first game, which counts as an iteration, was better than I expected: 68. From this starting point, which was no doubt the result of my previous experience, my efficiency averaged less than 2% (attempting the maximum score of 300, with values per game varying from -6.5% to 22.7%). If the model holds true and my efficiency doesn't change, I'll need to play at least 322 additional games to consistently score 299 points.

The amount of time involved in action (working directly on the task), what I call "effort," is simply the reciprocal of efficiency. My 70% best case corresponds to 1/0.7 = 1.43, or 43% more time than if I was working at 100%, while the productive part of my bowling amounts to more than 60 times what it would take to bowl a perfect game.

Interestingly, the actual time, in iterations, is closely approximated by a linear function of effort, whose coefficients vary with how much of the task we expect to achieve. For example, to achieve 99.7% of a task (plus-or-minus 3 standard deviations on the bell curve), the number of iterations is approximated by multiplying effort by 6 and subtracting 3. A more modest goal of 95.5% (2 standard deviations) takes 3 times effort minus 1.5 iterations. The lowest amount I've seen professionals accept is 80% (part of the so-called "eighty-twenty rule"), which interestingly is nearly as close as the iteration approximation gets to a pure multiple of effort: 1.5 times effort.

Both my major professions, test engineering and technical writing (the editing part), involve identifying the parts of tasks that have not been completed by other people. Based on that experience alone, I expect this model to apply to everyone, with efficiencies comparable to mine. Without in-depth scientific research to back it up, I can only propose it as an hypothesis, and explore some potential consequences and questions that derive from it, if it is true.

Education is an obvious area of exploration. Should education be redefined as a means of enabling people to perform multiple iterations of the components of tasks they will encounter elsewhere, so they can use their innate efficiencies to achieve acceptable starting points for those future tasks (much as my bowling games built on experience from years ago, which built on component tasks of walking and throwing)? Is efficiency innate, or can it be modified (and if so, is this task-dependent)?

How does this affect planning and execution of complex projects that have multiple dependent and independent tasks being completed by people with different efficiencies and access to resources? What are the implications for waste from such projects, at scales up to and including global civilization, especially on the survival of everyone and everything impacted by it?

No comments: