I like to think about the ongoing “remote vs HQ” debate in a more abstract way. It is really a question of decentralization vs centralization. Steve Sinofsky’s worthwhile thread last week reminded me that this is not a new plane of innovation or debate for companies, corporations and startups. Corporate structure is a technology itself that goes through waves of innovation.

Sinofsky points to Alfred P. Sloan’s post-WWII GM as a quintessential centralized “HQ” corporation. I don’t totally agree, but before we get there we need to go back a little further in history to understand the nature of centralization in corporate structure.

The first companies were decentralized

500 years ago, “companies” were small. There was no mass manufacturing and no efficient supply chains to distribute mass produced goods with certainty. So there was no benefit to producing goods centrally — nor yet the machines to do it. Neither yet were there efficient means to order goods from a distance or communicate between buyers, sellers and distributors. Postal systems were emergent at the time but not reliable and certainly expensive.

As such, when the first larger companies started to emerge, they didn’t produce goods. Rather they were trading companies that operated in the reverse way, bringing highly distributed commodities from the “new worlds” to more centralized consumption, largely in a few cities in Europe. Think the Hudson Bay company. Analogs also existed in Asia and the Middle East. These were decentralized organizations, relying on a relatively new “innovative” global sailing industry and the early vestiges of public and private postal systems for communication.

The next wave of large corporations — railroads — were also decentralized and also focused on trade, not production. Steam engines are often cited as the major innovation that enabled railroads. True, but telegraphs were just as important. If you are moving at higher speeds and measuring time in hours not weeks, postal communication is inadequate to coordinate a decentralized organization. Railroads would not have worked as businesses without telegraphs to enable them.

Efficient supply and distribution led to centralization

It is true that as large corporations started to encompass manufacturing, as well as distribution, that more centralization was seen. This is because of the cost benefits of mass production once efficient supply and distribution chains are established. Once you have to have a big building, why not put everyone else in it, especially when the best alternative to meeting in-person is just the mail or phone. Sinofsky also argues that centralization was driven by the top-down military norms of WW2 and the many people who returned from war to run big companies. I don’t think so. The pentagon has about 26,000 people working there to run a military of 1.4 million people around the world. That seems pretty decentralized. Also, to point to GM as a centralized organization ignores the complex tiers of decentralized supply chain that GM built to support it. Centralized in the early days, auto companies rapidly decentralized as processes and technology allowed supply chains to decentralize. The iron ore in the back door, Ford out the front door dogma of the Ford Rouge River plant didn’t last very long.

The more work can be specified, the easier decentralization is

Let’s take a step back to generalize the conditions that allowed for decentralization of the auto supply chain. There are two components: processes for communication and specifications for an acceptable unit of work. Think of the first as simple inter company processes like “send this form to order more” or “call this number to report a problem” and intra company processes related to taking those external interactions and performing on them. With the evolution from mail to telegraph to phone to EDI to API, there is less and less friction to both inter and intra company communication and processification. Communication is no longer a major impediment to decentralization in most cases.

The second condition for decentralization is the definition of specifications for a unit of work. In auto or any production, those are the drawings and quality specs for a part and the SLAs associated with lead time, delivery and payment. Put all this together and the more a unit of work or “job” can be specified and the less friction there is to communication, the more decentralized companies can be.

Now there is a difference between the unit of work for a product company and that for a service company — and moreover a difference between a unit of work in a production company and the internal service organizations (design, sales, etc) that support it. In the limit, you could hypothetically create communication processes and specify or standardize every single service activity or product related task to fully decentralize a company made up only of individual 1099 workers.

There are two issues, however, that frustrate this horizon. First, it turns out that standardizing and specifying certain types of service activities is very hard. I could contract MBA students to do market sizing for all of my diligence needs with pretty consistent success, but I can’t contract an MBA student to make a reliable recommendation on whether to invest in a certain company. The latter requires nuanced context that is hard to specify and requires a lot of iteration on research and inquiry. So instead, we’re hiring two investors. And because there is so much iteration in a diligence process and things move so quickly, we want our team to be in the office to avoid communication latency and confusion.

It’s hard to decentralize innovation and decision making

The second reason that the hypothetical all-1099 company is hard to achieve is that a company has evolve and innovate. If communication processes and units of service and product are always fixed, the company does not evolve and competitors will win. It’s that process of innovation itself that is highly dependent on informal means of communication and unstandardized efforts of work and experimentation, often in the form of judgment and decision making. It is hard to fully decentralize those parts of a company.

It’s that process of innovation itself that is highly dependent on informal means of communication and unstandardized efforts of work and experimentation, often in the form of judgment and decision making. It is hard to fully decentralize those parts of a company.

This is why you see a lot of corporate leaders pushing for a return to the office. They feel like day-to-day operations — which are easier to standardize in form and communication — are addressed well remotely, but innovation and decision making are suffering. There is no right answer here, but my view is that while routine roles with very clear standards and KPIs can be decentralized (and remote) a lot of judgment based roles will still benefit from some or frequent in-person interaction. I know mine does.

So can your sales people, customer success and developers continue to be successful remotely? Probably. But making good business decisions and staying competitive will probably require some continued in-person interaction at the leadership level and in less standardized functions (product, corp dev, biz dev, strategy, etc).

And, of course, culture matters

‍Now here is the most important part saved for last. So far we have not discussed culture at all! It is not the case that everyone wants to live in a cabin in Montana or a condo in Miami, Slacking with people they’ve never met for 8 hours a day. Work is part of life, and for most, life is social. Some talent will be drawn to decentralized companies but some will not. Startups need to define the culture they want — including along the dimension of centralization — and then focus on talent that fits that culture and mission. And if you choose to solidify as a decentralized organization, remain vigilant that decision making and innovation are not impeded.