I’ve been systemically learning about organizations that were inordinately effective at producing innovations in the 20th Century: DARPA, Bell Labs, PARC, etc. Despite the celebration and focus these orgs have received, several under-discussed similarities stand out to me. Most of these points involve uncomfortable truths and incentives that we often gloss over in feel-good discussions.

This is an incomplete list that needs more nuance, examples, and contains some contradictory principles. Despite that, I’m publishing it because not enough people are thinking about and discussing these points.

Small, Ad-hoc Teams

Successful innovations seem to start from situations where there is a pool of really smart people who organically come together to experiment and try stuff out in small groups. Successful projects tend to grow “bottom up” or “pull” style. Contrast to organizations that start projects by pulling together a huge number of people and resources from the outset and silo individuals into specific projects.

Low Barriers to Communication

Spontaneous conversations and the ability to ask experts questions with very little overhead is a theme that comes up over and over again. That seems obvious. The non-obvious piece is that small interaction frictions add up and need to be explicitly fought. Culture plays a big piece: in successful places from 15th century Florence to Bell Labs, the explicit understanding was that even masters should entertain questions from anybody.

Instigators, Visionaries, and Operators

Instigators are the folks who say “This is interesting! You/we should pull on this thread.” Visionaries are the folks who point the way, pull people into the project, and aim them in a certain direction. Operators are the ones who make sure things actually get done. The key points are 1) These can be different people and the barrier to them working together needs to be low (see previous principle) 2) Each role needs praise - often there is a power/credit dynamic that favors one category over the others. That imbalance incentivizes people away from certain roles and stifles innovation.

Aligned Timescales

Over and over, I run into innovations that die because different parties operated on different timescales. For example, VCs need to return funds in ~5 years. If something is going to take 7 years before it even starts making money or getting users, that doesn’t fly. Long-term projects in public companies are constantly in danger of getting the axe because top executives are focused on yearly or quarterly progress.

To enable innovations, all parties need to actually have the same time scales - not just pay lip service. Because when push comes to shove, you’re going to act according to the time scale you’re incentivized to be on, not the one you pay lip service to.

Making Failure Acceptable, but not Too Acceptable

Current Silicon Valley culture pays a lot of lip service to failure, but in reality it’s subtly penalized; you aren’t ostracized, but failure does make it harder to do the next thing unless you hit certain milestones before you failed. A real innovation org needs to support you just as hard on your third try as your first.

At the same time, you need to incentivize not-failing. This looks like having a number of at-bats after which you are subtly or strongly pushed out.

Distinction Between Research and Development with Organic Transition

Research and development are very different beasts. When people conflate them it leads to misaligned expectations. Bad expectations lead to projects being crippled, killed, or producing subpar results. Research is an unscheduled activity that leads to unknown results. Development is a scheduled activity that leads to known results.

The switch should only happen when people are convinced that you can get to a known result in a known amount of time. The handover is tricky because generally you want different people doing research and development. At the same time you want the handover to be organic and for the two teams to both be in the loop throughout the process. A common failure mode is for the research team to suddenly lob things over the wall and never think about it again.

Only Innovate One Thing at a Time

Each new process, idea, device, or tool is more prone to failure than an old one that has had time for people to work out the kinks. Combining multiple new things creates multiple failure points. Multiple failure points compound and make it so that trying to innovate in several ways at once is less likely to make an impact on the world.

It’s All About the Money

When innovations are never going to generate money, it actually shortens their time-horizons because they need to make their beneficent funders feel good every six months or a year. Thus, innovations need to be targeting capturable value on some time horizon.

Coupled to a Money Factory

Innovations take money and don’t generate money for a long time. Expecting monetary results too early (sometime ever) is death to real innovation. Thus, innovations need to be coupled to a stable money factory. The stability is important because the stability of funding is just as important as the amount. People’s incentives go haywire if they have money but don’t know that it will continue.

Aligned with the Money Factory’s Mission

This one is subtle and crucial. If the incentives are misaligned between the money factory and the innovation org, the friction leads to a downward spiral or just keeps the innovations in eternal limbo and from actually coming out into the world. Misaligned core incentives will cause these problems regardless of best intentions or words.

The goal of the money factory isn’t whatever mission statement they’ve made up for themselves. The real goal is the thing that they’re being measured on or brings in money at the end of the day. The Manhattan project was aligned with the US military’s goal of crushing Japan. Bell Labs was aligned with AT&T’s goal of making phone calls better. PARC was not aligned with Xerox’s goal of putting more ink on paper and that’s why the tech there was in limbo until it moved into apple.

Next Steps

I want to backtest these principles against other innovation organizations over the past century or so. Ideally, you can go through the principles like a checklist and correctly predict whether an organization will be effective at producing innovation. If the principles prove predictive, we can start applying them to organizations that want to produce innovation today and possibly create new organizations around them.

If you’re interested, let me know: 1) The top innovations of the past 150 years (don’t focus just on computer things - your life would be pretty miserable without LEDs) 2) Your top examples of innovation organizations today