Monday, April 28, 2008

innovating innovation

One of the best parts of the Second Life experience was its impact on how I think about innovation. From the first conversations with Philip about how to build Linden and our embrace of user-generated content to ESC and Rivers Run Red, innovation has been central to the process. And I do mean innovation – the commercialization of knowledge – not just invention – the creation of knowledge. It’s not enough to come up with interesting new ideas; those ideas need to be taken to a market.

As I zero in on what to do next, innovation is once again a central discussion, so I wanted to write up my thoughts on how I would approach creating an environment to maximize innovation. To maximize the commercialization of knowledge.

(nota bene: “Commercialization”, in this context, doesn’t necessarily mean “for profit.” It means that it has been released into a market. Case in point would be a new standard or piece of open source code released for free in order to create later opportunities.)

Innovation is very much a random walk. As much as we want innovation to be like throwing darts at the bull’s-eye, particularly in disruptive innovation it is extremely difficult to recognize a priori that an idea will be good. Moreover, most disruptive innovation emerges at the intersections of (largely) disjoint communities of practice and of different social networks.

These two realities – no certainty of direction and the need for heterogeneity – have implications if you want to maximize innovation. First, it means you want to create situations where you are able to try lots of different ideas. Second, you want as much diversity as possible in how and who tries. Third, it means you want to recycle ideas as the people and expertise around them changes.

Some of these implications are easier to leverage than others. Y Combinator, for example, does a very good job of optimizing for the first one and a reasonable job on the second. By drawing from a large set of submissions and then pushing participants through a short development cycle, Y gets to do a lot of experimenting and draw from a broad set of people. By focusing on such junior, hungry teams, Y creates massive incentives for those teams to bring a project to market in order to get funding or to be acquired.

However, there are tradeoffs. The micro scale of the funding means that participation is biased towards younger, more junior entrepreneurs. The cyclic approach of not being an incubator means that projects are strongly incented to succeed enough to get the next round of funding, rather than having the opportunity to fully explore an idea, decide there are better options and start over. Finally, Y doesn’t directly participate in funding projects at the next level or supporting more expensive experiments, constraining the exploration space.

(None of these are criticisms of Y, by the way. Paul, Trevor, Jessica, and Robert have adopted a specific strategy that is generating lots of interesting ideas, is clearly a blast, and may end up being financially successful.)

It simply may not be an optimal strategy for innovation, so there are a few aspects that I would approach differently.

You need greater diversity of participation, particularly of experience, temperament, and expertise. You want to be able to build teams that can blend years of expertise with youthful fire, impetuousness with wisdom. Filtering participants down to only those who can dispense with income, health insurance, and non-Ramen caloric sources robs you of the ability to leverage this diversity. Instead, I would argue to make them employees, to give them the scaffolding and support -- salaries, health care, vacation -- required to take huge risks, to experiment freely.

(Of course, this adds cost. Worse, it risks creating a comfy environment without innovation, but I think you address that through cultural and other means.)

Innovation needs a high failure rate. Paul et al are justly proud of the incredibly high percentage of Y teams that reach demo day with a product and that later to go on to achieve funding. However, what if the incentives that drive this performance – the Y team picking projects likely to launch in 3 months, teams not having an easy second chance, tight finances – mean that they are also more risk averse than they should be? If you change those incentives and support greater failure, the initial project ideas can explore a far larger set of ideas, resulting in more failures, but also more learning. Philip had a great saying about the benefits of “noble failures” which I think was dead on. You need to celebrate failures, capture the experience of them, and then preserve that information so that a later group can decide to riff of the failure, to build knowledge and try again.

About now, some readers will be commenting that this looks like an incubator and that all incubators are failures. Yes and no.

This does look like an incubator, but an incubator in the Bell Labs, ARPA, PARC, or Stanford grad school sense, not the modern “will trade space for equity.” Existing teams don’t need incubators, so the idea of providing a home to a set of them doesn’t seem like a good one to me. However, incubating ideas – where you bring bright, motivated, diverse, interested people together, give them challenges, and then get out of the way – has a long history of producing world-changing innovation. So, historically incubators weren’t failures, it’s just that we’ve changed what we mean when we talk about them.

The downsides, of course, are cost and comfort. If you have to employ everyone, to give them competitive salaries and benefits, you have a much higher burn rate. Worse, you must ensure that employees take great ideas out of the incubator to go start them. On the comfort side, you need appointments, contracts, or term limits, combined with a culture that your goal is to join startups. You can probably incent this as well – unpaid parts of your contract get transferred to the startup so startups recruit people, greater ownership of startups you helped launch if you also launch one, etc – but transparency and experimentation is needed here.

The cost is still a challenge, especially if you insist – as I think you should – on focusing the value generation on the launched ventures rather than the incubator. The incubator builds knowledge and expertise but should not be trying to IPO itself, since this strongly misaligns incentives with maximizing innovation.

Because of this, an attempt to innovate innovation may require very different – dare I say innovative? – approaches to funding in order to have enough runway to have a chance to succeed. It might be best applied within a larger company rather than as a stand-alone incubator. Consider a large company with a need for disruptive innovation but suffering from the “raising mice in elephant cages” problems common to large corporations. Rotating employees through a Skunk Works – and potentially letting them mix ideas with academics, outside experts, or interns – might form the kind of catalyst needed to break out of the innovator’s dilemma. As employees came and went from other groups and divisions, a Skunk Works would act as an innovation virus, spreading innovation processes and ideas throughout the organization. More importantly, by committing to experimenting with innovation, funders or the supporting corporation can avoid the micromanagement and hyper focus on short-term gains so deadly to innovation.

When we think about markets or technologies that seem moribund and unable to change, disruptive innovation is probably looming. The challenge is how to avoid Christensen's "Innovator's Dilemma" and drive that innovation rather than letting it happen around you. The answer may be to look backward -- to PARC, to Bell Labs -- in order to reinvent a path forward.


Oren Raboy said...

nice post cory.

I agree that different types of innovation demand different setups to prosper and some (particularly those that require more time/money/technology/market-dev) don't have a good platform to work on.

another area of funding is academia and public-funding (i.e. government)-- with all the pros/cons that come with that.

TrevorFSmith said...

I've split my professional career between startups and PARC so I have some experience with different strategies for innovation.

Many sorts of technical research simply cannot take place within the context of an organization like PARC because they place such a large value on patents. There are strict rules about public disclosure in patent law so running a live experiment (e.g. iterating a public web app) is at odds with researchers' professional reward structure.

That said, areas of exploration which require multidisciplinary teams, deep pockets and long timelines (e.g. biotech infrastructure, recombinant networks) have huge potential value and can take root in healthy industrial research groups.

If you're interested in systems which evolve more quickly in a social rock tumbler like the web, then Y Combinator style funding makes complete sense. Alternately, if your questions require wise teams and long runways then teasing out some of the PARC strategies and giving them a fresh coat of paint makes more sense.