Level II

The Importance of "Error" in "Trial and Error"

Like it or not, observations are the only way to learn about the world. Experiments are controlled observations that help us to focus specifically on the parts of the world we want to understand.

Many people refer to experimentation as "trial and error," and with good reason -- "trial" refers to any run that directly answers our research question, while "error" refers to all the trials (most of them) that don't answer our question. Most people also, unfortunately, think the "errors" should be forgotten and only the "trials" should be used to further our knowledge. This is a big mistake.

The "errors" give us critical information that can help us succeed -- we know what doesn't work. As strange as it may seem, it is possible to predict a "trial" -- a success -- from a set of "errors!"

Design of Experiments is the most powerful way to perform trial and error experiments. It simply tells you, in advance, which trials will contain the most valuable information. From this you can create a model to predict your Sweet Spot -- the best trial possible. This model summarizes the information you've gathered, not only from the successes, but also from the "errors." In fact, in many Designed Experiments the majority of the runs (sometimes all of them) are "errors," yet the information you obtained can still be used to find a Sweet Spot.

Article Labels

    There are no labels assigned to this post.