Everyone is talking about becoming more tolerant of failure these days. You read it everywhere. Failures are really learning opportunities. You learn more from failure than success. You have to be willing to fail if you wish to innovate. Experimentation entails some failures; without them, you won't create anything truly bold. You have to make it ok to fail.
Is it really true? Should failure be acceptable in your organization? HBS Professor Amy Edmondson has noted that there is a spectrum of reasons for failure. Some failures are truly preventable. They result from people knowingly deviating from procedures or not paying close enough attention to specifications. These types of failures are blameworthy events. They are not acceptable. At the other end of the spectrum are intelligent failures. These failures result from the experimentation process. In these situations, people are trying to learn through hypothesis testing or exploratory experiments. These types of failures, according to Edmondson, are praiseworthy events.
Are all experiments praiseworthy? Of course not. Some experiments are well-designed and carefully implemented. Others are hastily arranged and not well-designed. You want to encourage people to make smart bets. They should be engaging in enlightened, disciplined trial and error in the iterative process. They shouldn't just be haphazardly trying new things. Failures that emerge from a well-designed experiment are praiseworthy events. We should not send the message, though, that all testing is good testing. We should encourage people to put some thought into how they execute tests, experiments, and prototypes. Are they designed in such a way to elicit highly useful user feedback? Was data collected in an unbiased way, or were people simply trying to confirm what they already believed? Did people pick the right sample, and were appropriate controls chosen if a hypothesis was being tested? These types of questions need to be asked when people fail during the iterative process. We want to tolerate certain types of failure because we want to maximize learning. However, poorly designed tests and experiments do not lead to effective learning.