The most interesting company in tech: Valve

You hear it in technology companies all the time, especially at firms that have survived from their days as a startup to become a bigger firm: we want to remain entrepreneurial. To feel like a startup. Nimble. A place that entrepreneurs want to work. A place for builders to build (a phrase Jeff Bezos always used to describe what he wanted Amazon to be as a company).

But it has always felt a bit disingenuous. You couldn't fully escape the top-down corporate imperative, though they might have wanted to provide the illusion that you had.

[The Google 20% idea in recent history sounded like the most promising attempt, perhaps a more practical evolution of a earlier incarnations, for example research divisions like Xerox PARC or Microsoft Research]

But then I read about Valve Software, and it sounded like a company was actually taking all this lip service to heart and pushing this concept to its most logical extreme. What Valve has implemented as their "corporate" structure makes them, to me, the most intriguing company in technology, if not in business.

Here is the Valve Employee Handbook (PDF) which had the Internet buzzing a while back. In summary: Valve is a completely flat company, with no hierarchy, and everyone has to find their own project or start their own project and recruit other employees to the cause. You have no boss, no one can tell you what to do. Some companies have occasional hackathons or hack weeks; Valve is run like a perpetual hackathon. Google had 20% time; Valve Software lets every employee have 100% time.

As the Valve Economist-In-Residence Yanis Varoufakis explains in this long and fascinating blog post, the way Valve is organized is an attempt to turn corporations into more responsive, efficient entities by introducing real market forces.

Interestingly, however, there is one last bastion of economic activity that proved remarkably resistant to the triumph of the market: firms, companies and, later, corporations. Think about it: market-societies, or capitalism, are synonymous with firms, companies, corporations. And yet, quite paradoxically, firms can be thought of as market-free zones. Within their realm, firms (like societies) allocate scarce resources (between different productive activities and processes). Nevertheless they do so by means of some non-price, more often than not hierarchical, mechanism!

The firm, in this view, operates outside the market; as an island within the market archipelago. Effectively, firms can be seen as oases of planning and command within the vast expanse of the market. In another sense, they are the last remaining vestiges of pre-capitalist organisation within… capitalism. In this context, the management structure that typifies Valve represents an interesting departure from this reality. As I shall be arguing below, Valve is trying to become a vestige of post-capitalist organisation within… capitalism. Is this a bridge too far? Perhaps. But the enterprise has already produced important insights that transcend the limits of the video game market.

Varoufakis refers to Valve as a spontaneous order firm. What replaces market price signals in the Valve model is individual time allocation. That is, every employee can freely choose how to spend their time, which project or projects to devote themselves to (the Google 20% time model taken to its extreme). Contrast that to the traditional corporation, with people's work allocation imposed from the top down, through the organizational hierarchy.

He concludes his blog post:

Whatever the future of Valve turns out like, one thing is for certain – and it so happens that it constitutes the reason why I am personally excited to be part of Valve: The current system of corporate governance is bunk. Capitalist corporations are on the way to certain extinction. Replete with hierarchies that are exceedingly wasteful of human talent and energies, intertwined with toxic finance, co-dependent with political structures that are losing democratic legitimacy fast, a form of post-capitalist, decentralised corporation will, sooner or later, emerge. The eradication of distribution and marginal costs, the capacity of producers to have direct access to billions of customers instantaneously, the advances of open source communities and mentalities, all these fascinating developments are bound to turn the autocratic Soviet-like megaliths of today into curiosities that students of political economy, business studies et al will marvel at in the future, just like school children marvel at dinosaur skeletons at the Natural History museum.

A few reactions...

I'm not sure why intrinsically a time allocation model would be superior to a market-price driven model, but at the very least it would only seem to have an advantage if the individuals were very smart. My hypothesis is not that this model is inherently superior, necessarily, but that it provides a critical recruiting edge which, in a market with constrained talent, is a massive advantage. That in turn provides Valve with the necessary star talent to make the time allocation model flywheel spin. The innovative games (e.g. Portal and Portal 2, Half-Life) and business models (Steam) Valve has produced may simply come from that superior labor pool.

Secondly, spontaneous order firms may work best in a business like Valve's, the videogame business, which, like the film business, is a hits driven business in which every incremental game creates its own new market. They compete in a far less zero-sum game market than, say, Apple does in mobile phones. When you think about the coordination it would take for Apple to shift itself to a twice a year release schedule for iPads and iPhones, coordinating its product development, supply chain, marketing, and retail efforts across hundreds of countries globally, the concept of them becoming a spontaneous order firm seems impossible.

Third, I can think of many companies where a model like this wouldn't work because in those companies, people who are more senior in the hierarchy genuinely believe in their own superiority over the folks beneath them on the org chart. They'd likely be exasperated by the day to day work decisions coming out of a spontaneous order firm. This is not an indictment of the Valve model, just a check on the realistic speed with which such a model might realistically spread to other companies.

Fourth, at all the tech companies I've worked at, which are all more traditionally hierarchical, I wouldn't characterize them as strict Coase-ian "islands of conscious power [corporation] floating in an ocean of unconscious co-operation [market]". Most tech companies I know are obsessed with gathering price signals from the marketplace, and that data permeates the firm.

My first job at Amazon consisted of assembling, every month, a 100+ page report called the Analytics Package which had metrics, external and internal, on every aspect of our business. It would take me almost the entire month to compile, I'd have to translate each of them into graphs Tufte would approve of, and then I'd write prose analysis to accompany each package to highlight the most interesting signals. I had to generate hard copies of this, and every month I'd make good friends with the copy repairman as one copier after another broke down under the load of cranking out hundreds and hundreds of pages of information. Nowadays, most startups I know of have reporting portals that can generate such data in beautiful manipulatable charts on demand.

Lastly, a model of time allocation might be more susceptible to the cult of charisma? In my experience charisma and competence or intelligence are not always tightly correlated. The most dangerous person in a company is the charismatic fool.

Within the confines of a more traditional firm, though, I suspect there's much to learn from the Valve Software experiment, and so I'm really curious to see how they evolve over the next five to ten years. How much larger can Valve grow with this business model? Is it a more efficient model for gathering price signals from customers? How well does the model hold up against bad eggs, like the mythical brilliant asshole or just someone incompetent?

Let's examine one issue in more detail.

In companies, politics often crop up. This is especially common as companies grow larger. Politics are damaging to companies because they can lead to local instead of global maximums (wins for a local fiefdom or manager instead of for the company as a whole).

My experience is that politics is rooted in perceived mismatches between a person's own sense of worth and external signals of that worth, from explicit signals like one's title and salary to softer signals like the time spent with the CEO.

When a company is small, the politics tend to be minimal since many startups either are completely flat or have little to no hierarchy, everyone gets lots of time with the CEO, and everyone's marginal contribution is massive and easy to detect. In a larger company, the pathways for recognition get clogged. Suddenly the CEO you used to see  all the time you only see once in a while. Hierarchy is put in place to try to minimize coordination costs, but suddenly everyone is judging their self-worth against where they're positioned within that org. chart which is inherently a ranking system generally tied to compensation.

Valve's model has the potential of upending those political costs. There's so little hierarchy that mismatches between internal evaluations and external markers of value or less common. Since the company's surplus is divided up each year based on contributions, theoretically compensation is more closely and efficiently tied to value generation rather than getting out of synch purely based on factors like seniority or tenure.

In the end, it may be that all of Silicon Valley, rather than Valve Software, is the most interesting spontaneous order unit to study. The common complaint about Silicon Valley is the competitive labor market, with the average tenure at less than 2 years. California does not look kindly on non-compete agreements, it's a labor-friendly state, and so people carry ideas with them from company to company all over the region. They are all putting the time allocation model to practice, and while it makes recruiting and retention a pain in the ass, it leads to the region being among the most generative business ecosystems in human history.