When Herman Hauser created a team to create the ARM processor, now the processor that runs most of the mobile devices you know and love, he remarked:
When we decided to do a microprocessor, in hindsight, I think I made two great decisions. I trusted the team and gave them two things that Intel and Motorola had never given their people: the first was no money and the second was no people. They had to keep it simple.
When you start a project the normal course of action is to try to get as many people as possible so you will be able to have enough resources to accomplish the goal. In Lean Enterprise, the authors advocate another way, what they call the Lean Startup Cycle:
- Working out what we need to learn by creating a value hypothesis
- Decide what to measure in order to test that hypothesis
- Design an experiment, called the minimum viable product
- Build the minimum viable product to gather the necessary data from real customers to determine if we have a good product/market fit.
The authors continue:
The trick is to invest a minimum of work to go through this cycle. Since we are operating in conditions of extreme uncertainty, we expect that our value hypothesis will be incorrect. At this point we pivot, coming up with a new value hypothesis based on what we have learned, and go through this process again.
I love this idea. Instead of wasting years of time and money on an idea, let’s realize that we don’t know everything and use the scientific method to get us to the right idea. Let’s not be afraid of failure; you can’t have innovation without failure. The key to innovation isn’t avoiding failure; the key is reacting to failure earlier than the competition.
This method also calls into question the annual budget cycle. If all you have is once chance a year to respond to reality, you’ll end up clamoring for more people and avoid the simple, early, scientific method laid out above. A better way is to start small, provide some runway to solve a problem, and scale only when the solution shows its value in the real world.