Most of us have been bred from an early age to relish success. Our parents enrolled us in Little League, piano lessons, and college so that we would succeed. Now, as adults, the thought of failure is repulsive, even though our managers pay lip service to risk taking and the likely failure that it entails. In fact, in many cases, our CEOs are the presidents of our local Success Clubs. After all, they probably got where they are due to a long string of successes.
When it comes to developing products, there is a problem with success — we don’t learn much from it. When we dissect product development, which we also call innovation, we find that it is simply a series of trials that can either fail or succeed. These trials go by various names: experiments, analyses, sketches, prototypes, and tests. In other words, we try something out and see if it works. Based on the outcome of the trial, we try something else.
Progress occurs when we learn from the trial. To make faster or more effective progress in product development, we need to arrange our trials so that we learn from them faster or better. The interesting thing about trials is that we learn the most when we cannot anticipate the outcome. If we arrange an experiment so that we expect it to succeed and it does succeed, we haven’t learned anything. Similarly, if we expect failure and the outcome is failure, there is no learning. We learn the most when we don’t know beforehand what the outcome will be (see figure).
“Optimal” product development is a sequence of carefully arranged trials in which the outcome of each trial is uncertain beforehand. But we can dig deeper and do even better than this. For example, sometimes success and failure behave quite differently for the various objectives of product development, such as cost, time, etc. As a result, it is better to bias our sequence of trials to take advantage of this behavior.
Consider a common case in which success is simply the lack of a failure. To determine success, we have to prove beyond a reasonable doubt that our item won’t fail. This can take a long time, so it may be much faster and cheaper to run our trials biased toward failure. We run quick failure tests and inch up toward the success threshold, rather than running long tests on the success side.
These strategies depend very much on the nature of the trials. In some cases, failure is cheap and fast, and then the approach just described makes for rapid progress. But other trials are very different. If you are flight-testing helicopters, you probably want to bias your trials toward success.
We could take this further, but there are two basic take-ways:
(c) Copyright 2013 Preston G. Smith. All Rights Reserved.
