If you’re like me, you’ve been avidly consuming supply-chain news and analysis lately. Semiconductor shortages and stuck container ships are the charismatic megafauna everybody is watching, but I’m primarily watching for more mundane little things that surprise me in some way. I’m hoping they will serve as clues that point to new ways of thinking about supply chains, because I think our existing mental models are failing us.
A big reason is that we think supply chains are old and well-understood parts of the world (after all, ships still ply trade routes that have been in use for millennia), but as I will argue, they’re not. Supply chains as they exist today are as young and mysterious as the internet. The ongoing supply chain crisis is at least as novel an event in human history as the recent big Facebook outage. Arguably more novel.
And we won’t take supply chains seriously enough until we remystify them in a way that forces us to make up new frames that center their recently evolved novel aspects rather than their ancient and familiar aspects. Frames that encompass the entire vast scope of things that move along global supply lines today. Not just shipping containers full of durable manufactured goods, but bulk-carried materials, air-shipped perishables, seafood, livestock, piped materials, electricity, and so on. Viewed this way, the internet is just one supply chain among many, a bits-and-bytes-specific member of a cohort of technologies that date approximately to the 1960s.
Let’s look at a couple of examples of mundane little things that I found interesting.
Exhibit A: a shortage in transformers, among other things used by electricity utilities, means power outages that blow transformers in the near future may last much longer, because the transformers will take longer to replace.
Exhibit B: different chemical markets are reacting differently to the surge in natural gas prices. Ammonia producers are cutting production, while producers of a chemical called MDI (methylene diphenyl diisocyanate), which is also downstream of natural gas, are increasing prices. The Column (an excellent newsletter for the chemicals industry) notes:
The decision to cut production or increase prices has to do with current storage levels and whether the end market is willing to pay the price. Ammonia is used to make fertilizers and MDI is used to make polyurethanes. Apparently, polyurethane consumers can eat the margin.
There’s plenty more examples out there.
The condition of the beleaguered supply chains is like that of a patient being operated on by a drunk neurosurgeon making lesions for the lulz. The question is: are we at least learning while the patient is suffering? And if not, how do we install the right mental models in our heads, and associated instrumentation in the world, to enable learning?
Our Supply Chains are Young
The state of our understanding of supply chains, as revealed by commentary about the ongoing crisis is, as I said, rather unsatisfactory. All the reporting and analysis seems to adopt the posture that we are talking about a crisis of mismanagement in a well-understood old technology rather than a crisis of understanding in a poorly understood young one.
And the lay perceptions are even worse. Nationalist types genuinely seem to think we can trivially rewind to an era where countries had primitive, largely decoupled industrial bases manufacturing finished goods largely domestically. An era when trade was a single-digit percentage of GDP, and largely restricted to luxury goods and a few commodities. Somehow we’ve managed to convince ourselves that a series of revolutionary leaps over a couple of centuries amount to a small evolutionary movement.
The supply chain crisis is in some ways more unprecedented than Covid itself, given that containerized supply chains, and the world of distributed, networked, computationally coordinated production they enabled, are only a few decades old.
This is the first crisis of this magnitude to hit them.
To find a comparable crisis in history you have to go back to World War 2, with U boats sinking transatlantic shipping. And that was in an era when global trade was less than a third of today’s levels if I’m not mistaken (as a fraction of GDP) and still in the ancient mode of breakbulk shipping.
Beneath all the play-by-play updates and situational responses (both planned and market) to phenomena still mentally modeled as “shortages,” there’s a sense in which we are learning about the nature of the beast for the first time.
If you look pass shallow polemics about the evils of globalization and the fragility of aggressively lean operating practices, there’s a genuinely new science of supply chains emerging here. Supply chains as they actually exist, in fielded form, across the entire planet, not as they exist in engineering textbooks, limited ethnographies, or ideological imaginations.
Computation, Circularity, Situatedness
I’ve been a supply chain nerd for as long as I’ve been writing online, though with no particular expertise in the subject beyond what most engineers pick up routinely. Supply chain issues routinely crop up in my consulting work, but I’ve never worked directly on them.
But I’m definitely a fan. Looking back, I’ve written a surprising number of things about supply chains.1 Enough that maybe I should consider making an essay collection out of them.
Looking at the evolution of my thinking, I notice three major themes that help get at what’s mysterious about supply chains: computation, circularity, and situatedness.
First, I’ve tended to approach supply chains through computing metaphors. Containerization is like TCP/IP. The backend is like the cloud, but for stuff. The “front end” is like client hardware, comprehended through UX metaphors. Value getting unbundled, migrating to the most cost-effective global production loci, and the rebundled, is like distributed computing, complete with CAP-theorem style dynamics.
There are important differences. In particular, a lost packet can be retransmitted, but a lost container cannot, a difference that makes atom and bit networks hard to unify mathematically. But the close correspondences between supply chains and computing are, I suspect, very important. The metaphor is usually deployed in the other direction — computer networks understood in terms of shipping — but I find seeing supply chains in computing terms to be more fertile.
All this stuff is new. Supply chains didn’t always work this way. They started working this way with containerization, GPS and so on. To think of modern supply chains in terms of say, medieval Venetian merchants funding individual voyages to Aleppo to buy stuff brought by Arab traders from China along the Silk Road is like thinking of computing in terms of abacuses.
Second, I’ve tended to approach supply chains through full-lifecycle mental models that include circulation, the waste stream, recycling, reprocessing, and so on. This is surprisingly uncommon outside of environment/sustainability circles, even though the importance of circular thinking extends far beyond environmental concerns.
My interest in the anthropology and sociology of garbage goes back to a 2010 binge-read of several books on the topic, but the world at large is still not particularly interested. One indicator: the supply chain crisis is headline news everywhere today, but a similar crisis in the waste stream a few years ago, triggered by the National Sword policy in China (tldr, they stopped taking in Western recycling) was barely noticed. One narrow aspect, empty containers being in all the wrong places right now, has received some attention, but through the lens of a container “shortage” (there is no shortage; they’re all just in the wrong places and it’s not yet worth it to move them empty).
This aspect has now become an increasingly important one, and the so-called “circular economy” is now a fashionable theme among policy makers and regulators. With good reason. There is now so much trade, our trade streams are like geological or weather cycles, in terms of how they transport base materials around in space and time, and transform them in entropic terms. We have to literally think about, for example, the soil-nutrient impacts of exporting beef from Argentina to Europe, and “water exports” embodied as agricultural produce exports.
Third, and perhaps most important, I’ve tended to approach supply chains with a highly situated mindset. This is partly an outgrowth of simply enjoying travel to weird places where hidden parts of the supply chain operate, their contexts fully visible, but partly also due to the sense that it is an important lens on supply chains that engineers in particular tend to neglect. It is all too easy to get lost in abstract stock-flow models, and esoteric discussions of bullwhip effects, while losing sight of the fact that supply chains exist in meatspace. Last week we talked about how embodiment makes robotics richer than AI. Something similar is true of supply chains and situatedness. Talking about supply chains in the abstract is like talking about AI. Talking about situated supply chains is like talking about robotics.
Stuff moves from place to place on the surface of the planet, through places, in heavy diesel-guzzling vehicles. That’s what flows are.
Stuff sits around in huge warehouses and container yards in vast, vaguely dystopian regions like the Inland Empire east of Los Angeles. That’s what stocks are.
Taken together, these three aspects of supply chains: computation, circularity, and situatedness, point to a great deal of phenomenology that’s either missing, or marginal in our mental models.
Much of what is mysterious about supply chains has to do with one or more of these aspects. So how do you center them?
Centering the Mysteries
The computation, circularity, and situatedness aspects of supply chains are fundamental aspects that are often treated as marginal implementation details.
To model these aspects poorly, or worse, to fail to model them at all, means you deeply misunderstand how they operate even from an engineering perspective. If you’re an engineer, criticism from social scientists and environmentalists is the least of your problems. Your basic engineering models go wrong in a spherical cows in a vacuum sort of way. Your thinking likely suffers from the kinds of fundamentally religious blindspots that affect macroeconomists.
But this is not obvious during “normal” times when supply chains behave well, do their “jobs,” and appear to validate your models.
A crisis like the current situation reveals how little you actually know, and how little your models actually capture.
Let’s survey a few of these modeling lacunae.
The Chinese New Year seems to have a big seasonal impact on supply chains due to the fact that the back half of most supply chains lies in China, a consequential topological correlation created and confounded by history and geography.
How do you model that?
The ports of Los Angeles and Long Beach are the two busiest container ports in the United States, but they are also situated in an offshore/onshore oil drilling region, and it appears that the oil spill last week might have been caused by the dragging of a ship anchor. Which may have something to do with the fact that there’s a logjam of container ships waiting to unload, moored off the ports right now. Which might well add to the ongoing delays in port operations.
How do you model that?
Co-extensive with the nominal supply chains of the world, there exists a “dark” supply chain associated with smuggling and trafficking, and a big part of supply chain thinking has to do with securing the “light” side against the “dark” side. And by the way, the dark side doesn’t just steal bits and pieces on the margins. It might lock up an entire major shipping network via a ransomware attack that then takes a major hi-tech mission and a big dose of luck to fix. And the attack might be more than just a criminal operation — it might be geopolitical maneuvering by a state actor.
How do you model that?
Supply chains move stowaway organisms around the world, resulting in invasive species effects and widespread change in the evolutionary futures of ecosystems, which in turn affects human ecologies and their productive capacities. In other words, supply chains reprogram and couple local ecosystems, creating a sort of artificial “biosphere internet,” making the global supply chain more like the nervous system of a Gaia-like planetary meta-organism rather than a passive transport medium.
How do you model that?
And of course, supply chains cross borders, triggering geopolitical shenanigans, tariff wars, and accounting nightmares around the world. This is a whole large category in itself:
The Suez canal sues the Japanese owners of the Ever Given for the revenue lost during the blockage. How do you model that with a stock-flow model?
China holds belligerent military exercises in Taiwan’s backyard, scaring people whose lives revolve around TSMC (Taiwan Semiconductor Manufacturing Corporation, basically the only company that knows how to make leading edge 7nm and 5nm chips right now). How do you model that?
Politicians want to impose tariffs on high-tech manufactured items that cross borders multiple times based on simplistic “country of origin” labeling that can go badly wrong. How do you model that?
How do you model any of that?
These are not mere colorful implementation details of no significance.
They make the difference between models that are usable in practice, versus models that are only good for homework assignments.
They make the difference between vague invocations of “market forces” fixing things and actually knowing how to surgically intervene to unsnarl a snarled supply chain.
Engineering or Husbandry?
Many engineered artifacts can be viewed largely in terms of their designed function without much loss in understanding. If you’re designing a truss, material properties and stress/strain calculations tell you almost everything you need to know about how it will perform in the field. You can go from paper-napkin sketch to CAD design, to prototype, to production artifact, via a largely one-way flow, with very little iteration, and not go too wrong.
This is not true of supply chains. Even though many of the pieces are designed and put together the way other engineering artifacts are, the effects of those behaviors are different. And they evolve over time.
Supply chains “learn” the stocks and flows they handle (which themselves change of course). Architects talk about “how buildings learn.” We rarely talk about how supply chains learn, even though vast numbers of working supply chain professionals — purchasing agents, freight forwarders, truck drivers — spend their entire lives doing the actual learning, in the form of supply-chain lore.
Right now, for instance, world leaders are ponderously talking about how to hedge against China in alternate supply chain arrangements, but few have stopped to consider how supply chains actually emerge, learn, and grow over time. Or how we ended up wired to China in the first place.
Boromir voice: one does not simply make a new de-Sinified supply chain.
The thing is, a supply chain is mostly an emergent entity rather than a designed one, and its most salient features often have very little to do with its nominal function of getting stuff from Point A to Point B. That’s just the supply chain’s job, not what it is. What it is is a homeostatic equilibrium created by billions of sourcing decisions made over time, by millions of individuals at businesses around the world making buying and selling decisions over time.
Engineering emergent systems like this is more like the taming or domestication of animals than like the design of regular engineered systems.
Cows, for example, can be understood as milk and meat-making machines in a narrow economic sense, but they are primarily animals, and must be understood and cared for as animals, not just because that’s the humane thing to do, but because it’s the only way that actually works. It is not engineering, but husbandry.
Perhaps the right way to think of our dealings with supply chains is in terms of supply-chain husbandry? Certainly, they seem to present many of the alive-and-kicking features cattle do.
This alive-and-kicking nature is one reason supply chains make such great backdrops for movies. Supply chains are more like characters than props in the stories they feature in.
Marlon Brando’s On the Waterfront is set in the pre-containerization days when longshoremen ruled the docks. Chinatown is about Southern California’s water supply chain (yes, water supply is also a “supply chain”). The Amitabh Bachchan movie Coolie is about luggage porters on Indian railway stations.
Yet, though they can serve as great backdrops for human-centered stories, it is a mistake to think of supply chains in human-centered ways. Their primary animal nature manifests on an altogether larger scale than the scale of the animals (including ourselves) we are used to dealing with.
So just as it is a mistake to think of supply chains primarily in engineering terms, it is also a mistake to think of them primarily in social science terms.
Supply chains are a new class of engineered-emergent artifact, one that includes a few other globe-spanning things like the internet, the air travel system, and low earth orbit, that exist at a level of Gaian phenomenology, terraforming, and planet-scale husbandry. We only ever catch local glimpses of these things. The wholes are too big to fit in a single human mind, and the physical embodiments are too vast to capture even on a single map, let alone in a single photograph.
We have to understand these beasts, in all their evolving, learning glory, while living within their bellies. Abstract slicing and dicing of the phenomenology, via aspects like computation, circularity, and situatedness, can only get us so far. To finish the picture, we have to develop a sensitivity to how we inhabit these beasts at a human scale. Which brings me to a fun metaphor I want to share with you: Mattervision.
Supply Chains as Mattervision
When I was a kid, growing up in pre-liberalization India in the 80s, middle class families had “showcases” in their living rooms,2 to display curios and souvenirs acquired on travels.
Objects acquired abroad had a particular cachet, in a world where most things in the home were locally, regionally, or domestically sourced, and our showcase moved up in the world when my father brought back a treasure trove of things to put into it from his first business trip abroad in 1979. It was an age when “foreign” objects were still fundamentally mysterious, and the events and behaviors that brought them into our lives still seemed like special adventures rather than routine comings-and-goings.
Traveling abroad was rare, and for the middle class, almost exclusively something a few lucky members of the middle class did for work, not pleasure (“going to foreign” in the idiomatic Hindi rendering). If you got to go, it was closer to going on a treasure-seeking voyage than a routine kind of domestic trip.
We didn’t get TV until in our town until 1985, when I was 11, so evening entertainment until then often meant visiting family friends to gossip, and in lieu of watching TV together, chatting about the host’s showcase of curiosities.
The showcase was a sort of extremely low-frame-rate matter television — it was mattervision. The picture was solid and 3d though. A glimpse of an emerging material metaverse of global trade that India was not yet plugged into.
Today, such showcases are unfashionable, and India is firmly plugged into the global economy. Middle-class foreign travel is routine for the globe-trotting software developer class. By contrast, it was so unusual back in the day, when my father first returned from his trip abroad in 1979, he recorded a long talk on a 90-minute cassette tape (the tape player was one of his acquisitions) that we would often listen to, and also play for visiting friends and family. It was something like a partial soundtrack for our living-room mattervision.
Today foreign objects have lost their exotic appeal. They are no longer conversation pieces, let alone worthy of recorded soundtracks.
But at the same time, the entire home has been transformed by the globalization, and global penetration, of supply chains. The whole house, instead of just the living room “showcase” has become a sort of immersive matter TV.
You could say we now live inside our showcases.
Take a look around you. The supply chain isn't some distant back-end reality that stops at some warehouse outside of town. You are inside it. It flows through your home. And your neighborhood. And your town. Imagine you have AR glasses on, and everything around you is tagged with where a little map showing the places and routes involved in putting it together and bringing it to you.
That’s mattervision.
I find the UX metaphor of the supply chain as mattervision to be very helpful. Delayed shipments are like picture distortions. Stock outs are like dead pixels. Weird supply patterns are like banding artifacts. Misregistration of sound and video is like shipments getting broken up into multiple shipments. Quality problems are like graininess on the screen. Getting two things when you only ordered one is like a ghost image on your screen. Tuning your TV to different “stations” (a thing we used to do before cable and streaming) is like “tuning” in to Amazon versus the local grocery story. In some places one channel is stronger, in other places, another channel comes in stronger.
This metaphor gets at an important aspect of living inside the flows of the supply chain. It’s not raw material that we transform into waste through the act of living. It is an always transforming, flowing thing that we are part of.
It takes some temporal imagination to see the material conditions of your life this way, because so much of the flow we inhabit is so slow to change and move. It oozes past your attention rather than zipping past. Things pause in our living rooms on their way from factory to landfill or reprocessing center long enough that we see them as fixed elements in our environment.
But it’s all a transient flow, part of an endless situated circulation that computes our world into being every minute. The metaverse is already here, it just hasn’t been digitized yet.
To remystify supply chains is to appreciate the alienness of the beast we’ve cajoled into emergence around ourselves, and re-enchant the modern world itself.
If we can do that, perhaps we can begin to truly understand the messages hidden within the current crisis.
This is as good a place as any to compile a chronological list. I may have missed a few, but this is a pretty good record of my evolving thinking on the subject.
A 2009 review of Marc Levinson’s classic on container shipping, The Box
A 2009 review of William Langwiesche’s The Outlaw Sea
A 2010 travelogue, An Infrastructure Pilgrimage, about a visit to Bailey Yard
A 2010 survey of books about garbage (the waste stream is the other side of the supply chain)
A 2012 travelogue, Glimpses of a Cryptic God about a tour of the Panama Canal
A 2013 travelogue-essay on Aeon, American Cloud, developing a “cloud” and “client” metaphor
A 2016 issue of this newsletter about opportunities in the last mile
A 2018 issue in this newsletter, The Zeroth Mile suggesting a way to think about supply chains in terms of entropy
A 2020 issue of this newsletter, on Circular Economies
I’m told was also a common practice in the US in late-Victorian and Edwardian times.
Your writing and ideas are phenomenal. I very much enjoy and appreciate your way of thinking and perspective. Thank you for all you share with the community and for the inspiration.
On the question of how does one model complexity, from the anchor dragging one to all the others, one should look at cadCAD, the simulation tool developed by Blockscience. Then, understand how crypto tokens can help in unravelling the complexity. Before all that, watch Michael Zargam's talk at MIT. https://youtu.be/HldQF_MJN_Y