Astro Teller on Innovation (by Singularity University).
Astro Teller gives a great talk about his rules for managing an organization that is focused on big innovation. As the head of Google[x], he’s had lots of opportunity to test these theories.
This may be one of the best manuals available today for learning how to manage an innovative organization. The entire video is well worth watching, but I’ve pulled out the key lessons Teller came prepared to tell to the class, and my interpretation of his statements on each. Please assume that all the good stuff comes from him and all errors stem from my harried note-taking.
1. 100 * Innovation(1) > 1 * Innovation(100)
Taking lots of little bets is better than one or two large bets. Running more experiments makes you better - trying, testing, and seeing failure is important. A good strategy for ensuring this is to make people show you the failures (10 bad ideas for each good one, 2 bad prototypes for each good one).
2. !Understand Details + !Delegate -> !Innovation
To manage innovation, you have to accept that most breakthroughs will appear counter-intuitive to you at first. To avoid killing the true breakthroughs you must either spend enough time with the innovator to understand the details or you must delegate decisions to an external metrics or decision making framework.
3. New Tools + New Metrics -> Innovation
Organizations don’t spend enough time investing in building new tools. When you vet ideas from your innovators, ask them to tell you what tool was the primary enabler for them coming up with the idea and one idea to improve that tool. Then, during review, if it’s a good idea - focus on making it part of the next milestone for that company.
4. Rebuild = Innovate
When forced to rebuild something from scratch, you often rebuild it in a better way because you have a much better understanding of what you want to build. In an average development project, you spend most of your time figuring out what you want to build - if you assume you’re going to have to blow up everything you’ve done and rebuild once you know what you want, it doesn’t take that much longer and it allows you to build something far better and far more scalable.
5. Effort(Innovation_v2.0) < Effort(Innovation_v1.2)
The effort to invent something that is a completely new version of something is less than the effort of improving something. The existing product is some smart person’s best effort at solving a challenge within their constraints, if you just try to improve it a little bit, you stay mostly within the same constraints and you have to really fight hard for small optimizations and wins.
We think by asking for 10% cheaper we increase our odds of success, but in reality we’re better off issuing audacious challenges (“make it 1/3rd of the price”). Instead of issuing this as a challenge, one good tactic is to tell them a competitor has done it. Then people search for how it could have been done.
6. Innovation(story) = Benefit(vision) + Product(vision) + How(vision) + WhyNow(vision)
If you want to tell me how we’re going to change the world, you have to start with a big problem (and a big corollary benefit). Then you have to identify what the product or service is that could solve that problem and cause that benefit. Something about making that product or service a reality has to be really hard, because otherwise it would have already happened, so talk about what’s hard and how you think you can solve it. Finally, you have to identify what has allowed you to have this perspective shift and solve this problem — you have to have some sort of secret weapon that
7. P(Innovation | Bob) > P(Innovation | Idea_Z)
In the external innovation market (venture capital), it’s so obvious that you don’t bet on the idea, you bet on the entrepreneur. In companies, this is completely lost: companies vet and bet on ideas, evaluating the person driving the innovation only after they like the idea.
8. Idea Transplant -> Innovation
This is the easiest way to find a technology shift. Import ideas that were really successful in another field (e.g. PageRank understood information by graphing the connections between web pages) and try them as a hypothesis in an entirely new area (e.g. can we understand the connections between how companies buy and sell from each other and try to predict stock prices based on those changing patterns?).
9. CleanUp(painful work arounds + rule breaking) -> Innovation
Where people are torturing themselves, there are great ideas for innovation. When people are contorting themselves to do something, you can open up the flood gates by making that goal easy to accomplish.
10. Innovation(speed) α 1 / MediocreProjects(half-life)
The hard part of innovation is not coming up with great ideas. The rate limiting step of innovation isn’t coming up with ideas, it’s killing ideas. You want to kill your bad ideas and most of your good ideas, so that the great ideas have room to grow. Innovation directors should spend most of their time teaching their organization to kill more projects. If you can convince your engineers that innovation is stifled by not killing projects, you can get them to celebrate the process of innovation. You have to do this or else your great people will become frustrated and will quit.
11. Technology Judo > Technology Kung Fu
Engineers are taught Kung Fu, they are taught to take hard problems and solve them. Engineering does not train you to pick the problem, it trains you to solve it through hard work. The best innovations are driven by changing your perspective and understanding where technology or social behavior makes a new approach work. A good exercise for this is when faced with a hard technical problem, try to think of ways to reframe the question where the technology problem becomes easy (good for brainstorming). Change the problem, don’t just solve the problem.
12. Δ(Assumptions) 10% > Innovation > Δ(Assumptions) 1%
You can’t have big innovation without breaking some of the top assumptions held by most experts in your field, however, breaking lots of these assumptions at the same time is an indicator of overly optimistic thinking on whether your idea works or whether it’s something that will happen on any reasonable timeframe. Requiring one miracle to make it through to a breakout success is okay, but don’t require more than one.
Here are three rules that were on his slide that I don’t think he actually defined in the talk. I’ve given my best guess at the intended definition of these - but I could be wrong. If I missed anything, please let me know:
13. Δ(Assumptions) -> Innovation
If you identify places where smart people have different assumptions about something, the difference between those assumptions may lead you to a new idea or a deeper kernel of truth.
14. !Story -> !Innovation
Being a good idea, or a good story, is not indicative of being a good innovation. Don’t trick yourself into ignoring the hard work of proving the idea is actually good.
15. Innovation(someone) > Innovation(everyone) > Innovation(no one)
You can’t innovate by consensus. Exploring the ideas of one person (or a small group of people) is better than trying to get group consensus early on in the innovation process.
Finally, these ideas stuck out as being useful, but weren’t part of Teller’s formulas.
* Isolate innovators from those who have any vested interest in those innovators being wrong.
* The cost (opportunity and otherwise) of dumb projects not getting killed is much higher than any other cost.
* Innovation tends to happen best in small groups with a strong structure that protects their ability to think differently.
* Steve Jurvetson has a good rule for evaluating new innovations, you want 3-4% of the smartest people in any field to think this is the smartest innovation they’ve ever seen and everyone else to think it’s the dumbest. If you can’t get any one to agree, you’re wrong — if everyone likes it, you’re too late.
* As a manager, you’ve got to be immediately upfront about what the constraints are. This is the key to not frustrating innovators.
* The amount of room people have to fail is the same amount of room they have to succeed.
* If you look at all of the different use cases for your products, often innovation is about examining the fringe use cases and seeing if you can create something that creates a huge new market.