Structure Follows Context

I like mirroring principles in business a lot. My two favorite ones in business are Conway’s Law (product structure follows organizational structure) and Chandler’s Law (structure follows strategy). In conversations about business in recent years, I’ve been adding two more principles to complete a loop of sorts: market structure follows product structure and strategy follows market structure. The whole thing is what I call the data-driven death spiral, and is the reason I’ve become a partisan on the question of product-driven versus customer-driven thinking.  It operates through unimaginative leaders navigating entirely on the basis of market signals, which ultimately leads to businesses chasing their own tails. The only way a maturing business can break out of the death spiral is through the actions of a very strong leader. One capable of injecting a stiff dose of imaginative authoritah from the top.

dddspiralThat said, I’ve been sensing that my model is incomplete in a significant way. The biggest mirroring effect is the one it is easiest to miss: structure follows context. A context is the evolutionary environment (which is not the same as the competitive environment) within which a business grows, and which they shape to serve their needs as they grow. A city is the classic example of a context, but there are other kinds, such as ancient trade routes, or github (for purely virtual software teams). Contexts host businesses, but are not themselves primarily or necessarily businesses.

A context  is the sum of all history rolled up into a present-day operating environment, like a canvas with an evolving painting already on it. A new business must be painted onto some such canvas, just as software must be compiled for a specific machine. Only dictators have the luxury of razing a living context, creating a blank canvas (a dumb thing to do in almost every case).

Let’s look at the example of Seattle to see what I mean.

Seattle as a Canvas

From the point of view of a business, a context is simply the collection of environmental factors that potentially affect its operations, simply because pathways of influence exist, rather than because they are logically relevant. These pathways for non sequitur influences are what make contexts more valuable than blank canvases. Serendipity is a function of what’s already on the canvas. If there’s nothing there, there is no serendipity either. Good contexts create more serendipity than interference, but either way, structure follows context.

Think of this view of a context as a generalization of the concept of a playing field or board, with the difference that more than one logically unrelated game could be playing out on it. In large public parks for instance, you can see may informal games of frisbee, catch and touch football going on at once. These games can impinge on each other because they share a context. There might also be other people around who are not playing any game, but are picnicking, passing through, or maintaining the grounds.

A business context is not the same as the competitive environment or industry structure (the sort of thing Porter’s five forces are about). An industry or competitive structure is limited to logically relevant elements in the environment.  A context on the other hand can include things that are just coincidentally proximal in the environment. Neither is a subset of the other, and a business inhabits both.

As an example, consider Seattle, a city with a clearly marked context. It is easiest to understand it in evolutionary terms. A very short 10-point business history of Seattle would read as follows.

  1. The action began with exploration related to the northwest fur trade. This created geographic knowledge.
  2. The geographic knowledge gave rise to the more capital intensive timber trade, which drove the creation of basic railroad and water transport in the Puget Sound region, creating an infrastructure.
  3. The existence of infrastructure allowed a broader community to emerge beyond the timber trade, leading to more infrastructure: railroad connections to regional coal mining areas and the continent-wide railway network, and urban amenities.
  4. The existence of a connected community allowed Seattle to become the base for the Klondike gold rush of 1897.
  5. The community exploded as a result, leading to more port infrastructure and the emergence of the fishery industry as more than a small-scale lifestyle sustenance activity. Aggressive top-down development efforts such as the Denny regrade (which leveled an inconvenient hill and created a lot of real estate reclaimed from the sea) were triggered by the growth.
  6. This maturing social and logistics infrastructure in turn enabled a ship-building boom through World War I, which upgraded the economic base from commodities (timber, coal, fisheries) to manufacturing, and created a technological class. A major university began to develop, partly as a side effect.
  7. This allowed Boeing to take root in the area, accelerating the growth of an educated middle-class community, with a large service sector. By 1962, the image of the city as a technological hub was secure, and cemented by the 1962 World Fair, marked by the construction of the Space Needle.
  8. Containerization created a contest between Portland and Seattle, which Seattle won for a variety of more subtle contextual reasons that I won’t go into, stabilizing what had been a boom-and-bust economy somewhat, through the co-existence of regional and global economic activity.
  9. This leveled up the region as a transportation logistics powerhouse, and drove the rise of auxiliary logistics industries around everything from dry dock facilities to the cruise industry.
  10. The middle-class technology base created conditions ideal for the growth of the IT industry: Microsoft and Amazon.

The point to note about this story is how each chapter in the story created the serendipitous conditions required for the next chapter to unfold, but did not determine the nature of that chapter. Unexpected new things happened, which shaped the course of events at least as much as history. You could say coincidence turned into correlation and correlation into causation, through the impact of major human decisions made between chapters.

This story is clearly context-dependent. There is an obvious and evolving relationship between the environment and things that happened in that environment.

But is the story also path-dependent?  Let’s take a little detour to understand what that means and why the question is important.

Context-Dependence versus Path-Dependence

Path dependence is a strong claim: it means that where you go is primarily a function of where you’ve already been. The past determines the future with a sort of grim inevitability. A story that is path-dependent is continuous, each step leading to the next, with no sudden jumps or breaks.

A story that is context-dependent but not path dependent will have short segments that are path dependent, but also contain jumps that move the action to new paths. The old path may continue, but may not retain its status as the main show. The jumping-around behavior cannot be predicted from the shape of the old path (but can be influenced by it via environment modifications). Here’s a picture to help you visualize the four possible situations.


The picture is not entirely accurate, since previously traversed path segments shape the context for future segments by modifying the environment, but it would take an animation to show that. If you want to get pedantic, you could say that context-dependence in my sense is a weak-sort of path dependence mediated by a responsive environment, but let’s keep things clean. In our informal, non-mathematical sense, path dependence means the trajectory is continuous, with no jumps.

With that caveat, the first picture shows path dependence: an economic context containing a data-driven death-spiral, converging on a moribund end state. Ghost towns and abandoned regions reflect the end state of the first picture. Detroit is threatening to turn into a terminal, path-dependent ghost town.

The second picture shows a path that periodically jumps around but still conforms to the “geography” of the context, its hills and valleys. That roughly describes Seattle’s economy.

The third shows a path that has an internal logic but ignores context: this is how large-scale centrally planned economic activity can look within a local snapshot (in a very literal way when it comes to railroads being laid across a landscape, through the use and abuse of mechanisms like eminent domain).

The fourth picture shows a path that depends neither on its own past nor the context. This is a completely capricious economy, one that jumps around in ways that have no relation either to the past path or to the context. I can’t think of a real-life example. Possibly North Korea looks like this internally.

(Sam Bhagwat had a great post earlier this year applying the same metaphor of shifting landscapes and paths to a related theme).

Context-Dependence in the Seattle Story

The thing to note about the Seattle trajectory (the second kind of trajectory illustrated above) is the extent to which each step depended on exploiting surplus resources and broad capabilities created in a previous step, but in somewhat unpredictable ways, via an imaginative leap. The unpredictability and imagination are what make the story context-dependent but not path-dependent.

The transportation infrastructure was created to serve the timber industry, but became a critical piece in the game to occupy an advantageous position in the general transportation industry. The maritime industry grew around shipping timber to San Francisco and miners to Alaska, but allowed a large-scale fisheries industry and cruise industry to develop as well.

Each time, the region broke out of a functionally fixed understanding of itself and reimagined its economic identity. The process is continuing in the digital world: Amazon moving from ecommerce to the cloud business via AWS is a move that is very similar to the railroad being built for one purpose and then being used in a reimagined way, only within a single company instead of across multiple generations of companies. In contexts that have a strong digital element, context dependence dominates path dependence even more strongly, to the point that companies plan around strong context dependence (an idea that the software industry understands as “platform effects”).

Material factors involved in the evolution of a context are at least somewhat predictable. It makes sense that more uses would be found for shipping infrastructure for instance, or that a railroad network built to move timber would expand to move coal. It is important, however, not to overstate the predictability with hindsight. It seems obvious now that the cruise industry would grow in Seattle due to the existence of a shipping industry and port facilities built for timber-industry, mining-industry and military needs. But it was not at all inevitable that Seattle would carve out a position in cruise-ship tourism given competition from Vancouver and Portland, or that the Space Needle would be constructed, creating an iconic focal point for tourism. Or for that matter, that the Pacific Northwest would have even developed as a major cruising destination without the effects of the Klondike gold rush.

But the most unpredictable element in the evolution of the environment is the human factor. The development of the Seattle community in terms of diversity and size, as a generalized human capital base, allowed for the most unexpected and creative moves. This included winning the race to be the logistics hub for the gold rush, turning the region into a tourist destination to grab some of the cruise industry, and using a skilled population to jump from ship-building to plane-building and container-shipping and from there to software.

Another way to think of this is as follows: while material resources can only be repurposed in limited, somewhat predictable ways, human resources can both actively behave and be passively repurposed in unpredictable ways. An existing railroad makes it easier to lay a telegraph cable, since right-of-way agreements already exist. But no direct value can be added to steel-making, outside of supply and distribution. By contrast, children growing up in a community can end up doing something entirely different from their parents.

Software tends to be somewhere in-between human and material in its repurposability. Amazon’s ecommerce business created generic computational infrastructure that now runs a lot of the startup economy.

This unpredictability in how the environment created by one era will be exploited to create the next one means the evolution of the environment is context-dependent without being path-dependent. The path can jump about unpredictably in a broader context-landscape through creative and imaginative leaps.

This is the primary role of the lightning bolt labeled authoritah in the first picture: flashes of authoritarian insight allow an evolutionary path to jump to a new restart location on the landscape. Instead of being trapped in a local minimum via the data-driven death spiral, the path can move to a new basin of attraction. Sometimes all it takes is for a businessman like Jeff Bezos to decide to locate facilities in the area. Sometimes it takes a bunch of citizens deciding to bulldoze a hill, a gold rush nearby, or the building of bridges and locks.

How do we know this sort of development path is not inevitable?

One way is to consider a natural A/B test in a very similar part of Europe: Scandinavia. Not only is Scandinavia similar to the Pacific Northwest geographically, it is also similar demographically, since Norwegians and Swedes made up a lot of the initial immigration to the Seattle region.

In Scandinavia today, there are two very similar countries: Norway and Sweden. Sweden has become a technological powerhouse, while Norway, which has richer natural resources (in particular oil) has remained a significantly simpler economy.

At the risk of annoying a lot of locals here, there’s an A/B test even within the greater Seattle area. Amazon is a company that  jumps to new paths with almost boring regularity, while Microsoft today seems to be on the brink of a terminal death spiral. I have my fingers crossed that they pull out, but I am not very hopeful.

The Curse and Blessing of Resources

Norway is usually held up as an example of the idea that the curse of resources is not inevitable. Being lucky enough to have significant petroleum reserves does not doom a country socially and culturally.

But it does not unambiguously bless a region either, as the lack of energetic development in Norway relative to Sweden or the Pacific Northwest shows. The Norwegians have been wiser than both their southern neighbors (the Dutch, who lent their name to the Dutch disease, a rough synonym for the curse of resources) and the oil-rich Arab states, but haven’t entirely avoided the ennervating effects of having a lot of natural resources.

Here’s my theory of how resources work: to keep a region economically thriving, unpredictable jumps need to be made roughly once in the working lives of every new generation, otherwise cultural memories of risk-taking as the norm fade. A single generation that does not experience the pressure to reboot the economy from a new starting point is enough to create a too-rooted culture, and drain a region of vitality. If there are enough resources to secure the future of many generations, the region goes from being context-dependent to being strongly path-dependent.

Path dependence, the data-driven death spiral, and the curse of resources are all mostly the same thing: evolutionary trajectories that unfold continuously in a context, without significant random jumps.

Unpredictable leaps can only come from individual humans within a developed community that feels secure but not too secure. The capacity of a community to throw up individuals who can see the available leaps and make them happen is the real high-value economic resource. This might take the form of building companies, driving through ambitious civic projects, or making significant regulatory moves that open up opportunities for regulatory arbitrage relative to political neighbors.

How do you maximize the capacity?

First, you need to somehow give up or take out of play as much of the surplus resources as necessary to create a sense of economic urgency for every generation. This is obviously hard for biological reasons, since the natural human instinct is to secure resources for descendants as long as possible.

Second, you need mechanisms that allow those who can see and make context-dependent leaps to self-select into a context. Ideally, this would be paired with neighboring contexts that attract those blind to opportunity and looking for comfort.

This mostly means more open domestic and international borders, and a better regulatory environment for businesses relative to neighbors, which may share many of the same context-dependent opportunities for leaps. This calibration is important: the state of Washington does not need to become the most business-friendly political region in the world. It merely needs to do better than Oregon to the south and British Columbia to the north. You don’t need to dismantle the welfare state and turn the context into a red-blooded Libertarian paradise. You merely need to create the right sort of minimal migration gradient across all your borders.

When both conditions are satisfied, you get the blessing of resources: increasing capability to spot and make imaginative economic leaps in the context. Imaginative risk-takers select into your context, while risk-averse individuals leave.

When both conditions are violated, you exacerbate the curse of resources rather than alleviating it, via adverse selection. Risk-takers leave, the risk-averse stay.

Accumulating risk capacity is a blessing, since it arrests the progress of the curse of (material) resources, breaks toxic path-dependence (but not valuable context-dependence), and renews the lease on life for an economic context.

Context Follows Structure

Let’s summarize the story so far.

Context matters. Businesses do not operate on a blank canvas except in North Korea.

Wealth is created by periodically decoupling economic paths from the past, but not from the context, via unpredictable, at-least-once-a-generation leaps of the economic imagination. Failure to make these leaps creates path-dependence and a data-driven death spiral for both the context and the businesses within it.

The curse of resources is a predisposition to the latter path. The blessing of resources is a virtuous cycle of increasing capacity for making unpredictable leaps. This can be secured by taking the curse of resources out of play as much as possible, and creating a local positive selection effect that draws imaginative risk-takers with an eye for context-dependent opportunities, and repels the risk-averse.

Let’s say a little bit more about the unpredictable jumps part. A jump has a from location and a to location. What characterizes the two? I think it leap is primarily a leap in perspective.  Specifically, perspective on functional capabilities.

Many who understand and appreciate the point about functional-fixedness (“when there’s a hammer in your hand, everything looks like a nail”) fail to see that you cannot break out of a functionally fixed state in the abstract. You have to develop a new functional fixation to break an existing one. To stop seeing everything as a nail, you have to start seeing your hammer as (for instance) a long stick with a heavy metal grip.

This takes more than risk-taking and simple willingness to abandon a fixed view. It takes imagination.

A sufficiently imaginative shift in perspective changes the meaning of an economic context, thereby creating a sharp turn in the future evolution of the context itself. Unlike the path, the context evolves continuously (Seattle is not suddenly going to turn into flat, land-locked Nebraska no matter what Amazon, Microsoft or Boeing do), but it can take sharp turns in response to the leaping path. Here’s a visualization.


To a sufficiently imaginative population, a timber-hauling railroad suddenly looks like a ready-to-go coal-hauling railroad. The further development of the railroad makes it look more like a coal-hauling network and less like a timber-hauling one. Developed fishing and cargo shipping industries suddenly look like a half-built cruise-tourism industry. Acting on this perspective makes the waterfront more of a cruise-ship waterfront and less of a cargo shipping/fishing waterfront.

As a result of these perspective shifts, usually due to individuals, the entire context is steered, not just individual businesses or sectors. In the short-term, between jumps, the context shapes the path. In the long term, as a result of jumps, the path shapes the context. The more broadly imaginative and risk-tolerant the resident population, the more path and context start to merge into one.

As a result of dynamics at these two different time-scales, the mirroring effect of structure following context is much looser than Conway’s Law or Chandler’s Law. There is a great deal of temporal distortion. But ultimately, you can look at a business and see its environment reflected in it, like in a fun-house mirror.

The Creative Class

The idea that structure follows context highlights the problem with Richard Florida’s idea of the creative class (which has fallen into disfavor in recent years). It is not that human factors do not matter. They matter in this model in a specific way: you want the kind of community that will throw up imaginative risk-taking leaders or attract them from the outside. I doubt that this capacity correlates particularly well to the creative class. Florida’s psychographic model is at once too broad and too narrow. It overvalues marginally useful traits like a fondness for good coffee and undervalues important traits like curiosity about the opportunities in a context and authoritah. Attracting coffee snobs to Seattle does not mean you are also attracting those who will get curious about how Starbucks operates. Attracting theater lovers does not mean you’ll attract people who get curious about container shipping.

More broadly, most of those within the creative class are rank-and-file types who will follow once a leap has already been taken.

But there is an even more serious problem. You can’t develop an economic context independently of ideas for how to develop it. This is why the creative class model focuses on peripheral context-independent elements such as an arts scene, cafes and nightclubs. Those things can basically be created anywhere through the injection of money.  Those things are also essentially sideshows, not main economic acts with deep context-dependent roots. Main acts exploit features of the context in an intimate rather than superficial way. The decision to build a cruise-ship terminal is a deeply context-dependent one. The decision to build a symphony hall or football stadium is not. The decision to build a casino is probably the most context-independent move it is possible to make, and to my mind, having experienced living in Las Vegas, a motif of end-times. When a casino sprouts up in a region, it is probably time to leave. It is a sign that the residents have run out of ideas.

To spot the next main acts, you need a lot of imaginative risk takers looking at the context with fresh eyes, and spotting specific new directions to drive towards. But you only need a few of them to succeed. It’s a calculated numbers game, not brute-force “cultural” or “industrial” development. This means attracting those who are interested in the particulars of a context rather than its generalities. This means putting the particulars front-and-center in plain sight (something those glossy hardcover guides in hotel rooms largely fail to do, since they usually have some sort of nightlife or opera scene on the cover, rather than feature stories about local industries).

The increasingly digital nature of economic contexts  makes the creative class even less relevant as a causal force. You could be in a cafe anywhere exploring the economic possibilities of a distant city.

It should be possible, using virtual mechanisms, to attract the attention of imaginative risk takers anywhere in the world. Releasing large, open datasets that provide a view into an economic context is one way to do this. An older way is to provide friendly regulatory conditions that can be exploited even by those who are geographically remote, as Delaware, Panama and most recently, Estonia have discovered.

Regulatory affordances and data are a way to create a remotely accessible API of sorts for geographic regions. They allow software to eat a part of a context. There is no reason why other parts of a context cannot also be similarly exposed.  In fact, it seems to me that something is a valuable part of the context only to the extent that its digital extension creates new value. There is no meaningful digital extension possible for many things that are assumed to be important in the creative class model, such as nightclubs and great cafes, because they present no particular context-dependent advantage. By contrast, a detailed virtual 3D tour of the Port of Seattle might inspire more entrepreneurship than an entire neighborhood full of cafes and nightclubs. You could make the statement more extreme: the more time visiting business travelers spend time in local nightclubs and cafes, the less likely it is that they will spot a local opportunity that will make them want to stay and take an interesting risk.

Can purely digital contexts, with no geographic footprint, ever achieve such serendipitous context-structure couplings? Are purely digital cities possible?

I am not yet sure. A context is only a context, as opposed to a competitive environment or an industry structure, if it has multiple largely unrelated things going on within it, with pathways for serendipitous influence among those unrelated things. Second Life attempted to be precisely such a pure digital context. In a more controlled way, Facebook is such a context. Github is a step or two beyond both.

But I don’t think any of these purely digital environments are as yet as comparable to geographic contexts in terms of fertility. Seems like we’ll get there though. I don’t see any fundamental reason it cannot happen. This is why I’ve taken to calling those who insist geography will always prevail geographic supremacists. They’re not quite as bad as digital dualists, but not by much.



Get Ribbonfarm in your inbox

Get new post updates by email

New post updates are sent out once a week

About Venkatesh Rao

Venkat is the founder and editor-in-chief of ribbonfarm. Follow him on Twitter


  1. Your breakdown of Seattle’s path reminds me of Gray Brechin’s “Imperial San Francisco: Urban Power, Earthly Ruin”, a similar outline of the economic history San Francisco.

    Briefly, Brechin traces how hydraulic gold mining depended on metallurgy. The farmers downstream of the mines banned hydraulic mining, but the US Military doubled down on the metallurgy cluster to build ships and artillery. Those shipyards funded radio research, which was fertile ground for the characters in the founding myths of Silicon Valley.

    “Can purely digital contexts, with no geographic footprint, every achieve such serendipitous context-structure couplings” I’ve often wondered the same thing. Could there be a digital equivalent of Freud’s prewar Vienna, Hemingway’s interwar Paris, or even Wozniak’s Cupertino? The closest thing I can think of is a good irc channel, but that still falls well short of the mark.

    • Interesting. Yeah, SF is fascinating too. Not as up to speed on local history there, so I used Seattle as my example. But basically every city large enough to have a character I think has a story like this.

      There are digital examples, but they are as yet relatively limited/early stage.