Failure is big in the news lately, because people are getting smarter about it. I’ve written before about my admiration for Google’s philosophy, as expressed in this quote from Senior Business Product Manager of Google News, Josh Cohen, in a recent Atlantic article:
“We believe that teams must be nimble and able to fail quickly.”
Now, we get that failure and innovation are joined the hip. But I still hear failure spoken of as a kind of necessary cost, a purgatory one goes through in order to enter the heaven of success. It’s like the old song:
Pick yourself up,
Dust yourself off,
Start all over again.
But let me suggest that failure is not an unwanted consequence, but a necessary ingredient of innovation. And you’re not starting all over again after a failure.
I learned this lesson years ago, when our company decided to develop our own method of “genetic programming.” For those unfamiliar with it, genetic programming is computer programming inspired by biological evolution, in which the program itself evolves and eventually programs its own solution.
A couple of years back, we invested a great deal of time exploring how we might use genetic programming to better understand why survey respondents answered certain questions in specific ways. If we could answer this, we would gain critical insight into the “drivers” behind certain decisions.
Genetic programming uses computer programs that manage themselves, in a way. Think of a “master” program that works with “agent” programs. The master tells the agents to go take a stab at a given problem. The agents aren’t really equipped to solve it, so the master analyzes the agents’ attempt, and figures out which parts were useful, and which weren’t. The master then reformulates the agents and gives them another try. Repeat over and over again until the problem is cracked. Typically it can take several hours of straight processing time to get to the answer. But amazingly, they always find it.
The key to this whole process is the programs’ ability to analyze the failed attempts at solving the problem – to figure out what got them closer to an answer, and what didn’t.
That’s the cycle: formulate, attempt a solution, fail, analyze the failure, formulate again. It’s an evolutionary process, and here are two salient facts about it:
- The faster you can move through it, the faster you arrive at a solution.
- Failure is an intrinsic part of this process.
And that’s the real point. Success at innovation is not a golden ring to be grabbed, or a target to be hit. It’s the result of an evolutionary process, and failure is a necessary element of it.
The guys at Google are not dumb. They’re a company of programmers, and they understand that genetic programming is a paradigm of the innovation process. Which is why they embrace failure, and why we benefit from their successes.
That’s not to say you throw a hundred ideas at the wall and hope to hit something. You choose a few starting points, and when the failures come (and they will), you analyze the failure and recombine the elements – not to start again, but to continue the evolution.
How did our genetic programming experiment go? It failed. But in the process we were led to another mode of prediction – Bayesian Inference, which proved a great success. Chalk up another one to failure.
Thompson Morrison is CEO of FUSE Insight Labs, which develops Automated Online Interview technology to help marketers listen more efficiently. He also authors The Radical Ear, a blog about listening that drives innovation.