Fail fast is one of those ideas that is hard to argue with. After all, if “it” ain’t going to work, the sooner you find out the sooner you find out the sooner you can stop doing “it” and try something that might work.

While that advice seems obvious, I have always had trouble with it. Perhaps because of the tired, but true, observation that, “everything takes twice as long and costs twice as much as you think.” Or, perhaps because these decisions involve judgment calls. If you get to the point where you know for absolutely sure that you failed, then you probably let it go too long without pulling the plug.

David Cancel has a post on the subject of his interview with Gail Goodman (CEO of Constant Contact) that has a number of great insights but here is one on the subject of failing fast:

The trick is to give experiments enough time to prove themselves. Too often a focus on failing fast leads to false positives. Three months is not enough time to figure out a sales model; you need to give it time.

The question really is how much time,  when do you know, and how do you figure all that out? Determining when you have failure is very hard to do. Among other factors, there can be a huge emotional screen between you and the right decision.

Here are a few observations:

First, get outside input from people who don’t have an axe to grind. Listen to what they say, but remember what Winston Churchill (I think) said, “War is too important to leave to the generals.” In the end, if it is your business, you have to make the call. But you need to make the call based on a realistic (not optimistic or pessimistic) view of the world. It seems to me that the key is to get to that realistic view of the world.

Second, think of yourself as working from a thesis. For example, a thesis might be that customers will leave their current providers and use my service because current providers are changing their focus away from these customers. If time passes and this is not happening, ask yourself why. Perhaps the model you developed in 2007 is not working in 2010 – there was a little matter of a global recession that may have changed the underlying conditions upon which you built your model. Perhaps it is time for a mid-course correction.

Third, investors are, by their nature, impatient. Their tolerance for delay is low. As soon as things seem to be taking longer than planned they will be thinking that the model does not work. Soon, they will be pushing for a change. Keep in mind that they may not be wrong. They also may not be right. Having to convince skeptical investors is a good discipline. 

While you need to avoid false positive in your fail fast strategy, you also need to be able to spot when you have an unreasonable commitment to a strategy that has a low probability of success (hopefully before you have spent the time and resources to conclude that it has zero chance of success).

This means that you will have to leave some percentage chance of success on the table, but that is what judgment is all about.