#15: Maintenance by design
How to think about designing people, teams, organizations, and supersystems to be ready for uncertainty.
I’m an author, organizational sociologist, strategy professor, unsuccessful furniture maker, and Xoogler—this is yet another of my attempts to make sense of the state of not-knowing. The ideas below are only partially baked.
After leaving Google in 2008 on the cusp of a global financial crisis, I spent a poorly paid but transformative summer helping to run (= be a clueless minion at) a lavishly equipped woodshop and furniture teaching workshop at the Anderson Ranch Art Foundation at the top of Colorado’s Roaring Fork Valley.*
Monday through Saturday, we ran courses in a range of different woodwork skills for 10-15 students. These were taught by visiting craftspeople, with shop staff helping out and keeping the studio operational. For the advanced courses, we’d watch students who were experts in their own right doing amazing things to wood. For the beginner courses, we’d leap about stopping students from sawing their arms off (true story).
Every day, we started up the shop around 7.30am and closed it down around midnight. During those hours, the shop was in use by students and staff working on their projects. The assorted equipment—power tools, hand tools, workbenches, finishes, sawhorses, clamps, vacuum molders, brushes, etc—was probably in use about 12 hours a day.
This meant lots of entropy and real-time efforts to fight it by putting stuff back where it was supposed to go and fixing or replacing broken and worn-out components. We also cleaned up each night as part of shutting down. Nonetheless, intensive use degrades even an assiduously managed complex and fragile resource.
So: every Sunday morning the shop crew did a maintenance morning. The goal was to return the shop to the state in which it began the week, before we all headed to the attractions of the nearby town of Aspen (beer, pizza, a library, a progressive thinktank).
We sharpened every chisel and handtool in the shop (with benchgrinder and successive water stones), cleaned out, vacuumed, and oiled every powered machine and checked their blades, emptied the sawdust collectors, restocked all the consumable bins and shelves, rearranged the workbenches.
Looking closely at the shop and its components every Sunday morning promoted a kind of attention to problems and the whole system that was impossible to summon during the rest of the week.
Four types of maintenance
The most common way to think about maintenance is as a process of finding and fixing broken stuff—maintenance as the routinized search for problems. This allows many small fixes (easier and usually cheaper) instead of a big one (harder, requiring more downtime, more expensive). Maintenance at the woodshop was largely of this type: aimed at catching and fixing what was broken.
But there are at least three other ways to think about maintenance.
The first goes beyond merely fixing what’s found to be broken or about to break—maintenance as surveying what is suboptimal and improvable by being rebuilt. Refactoring code is a common example. If you do it right, nothing about how the service works externally changes. Internally, though, duct tape and baling wire is torn away and jerry-rigged bullshit is replaced. Things run cleaner, cooler, less buggily, and are better documented.
One more step beyond that is thinking of maintenance as a way to trigger unexpected ways of looking closely at a complex system—maintenance as investment in nonparameterized system awareness. A daily yoga practice is a form of maintenance, as is a weekly stand-up project team status meeting, or a quarterly board meeting for a startup. Doing the same thing at a regular interval provides opportunities to recognize when the system is beginning to fray in ways that simple problem-finding wouldn’t catch.
(Some people I know invite people over for dinner every month as a way to periodically inspect their supplies and force the use and cleaning of the apartment, as well as every piece of flatware and crockery. Admirable.)
But the highest form of maintenance is designing whole systems to require and enable meaningful attention when their operation conditions change—maintenance by design as whole system design for environmental perception and response.
Maintenance by design
Maintenance by design means building systems (individuals, teams, organizations, supersystems) to intentionally be sensitive to change, to display the effect of change transparently, and to be malleable in response to that effect.
A surprisingly good example of this is wheat.
Over the last two years I’ve spent a surprising amount of time trying to understand agriculture in general and wheat in particular. One of the more exciting innovations here is an old approach to wheat cultivation in which a field is sown with a genetically diverse population of wheats instead of just one variety of genetically undiverse wheat. This latter monoculture approach is how nearly all modern wheat is grown. The photo below is of a few different wheat varieties being trial-grown for possible inclusion in a population—a future issue of the newsletter may be about this remarkable project.
A population wheatfield has a huge variety—hundreds of different kinds—of wheat; a conventional wheatfield will have a single variety of wheat growing in it.
A diverse wheat population embodies some qualities of maintenance by design. It is highly sensitive to the growing environment: soil and weather. The specifics of the soil in which it is grown and the weather during that growing season will determine which of the wheats in the population will do better and which will do less well. The same starting wheat population will look different and produce differently in each field in which it is planted, and each field will look different and produce differently each year.
And though the weather might not be predictable, the genetic diversity of the wheats in the population buffer against that automagically: each type of wheats simply grows better or worse depending on how the weather is at any particular moment during the season.
As a result, one eminent developer of wheat populations in the UK tells me, population wheat fields aren’t prone to complete failure like monoculture wheat fields can be—in fact, that their yields are relatively stable from year to year despite dramatically different growing conditions.
Maintenance by design (creating a population wheat field) highlights even small changes in expected operating conditions by quickly adjusting its behavior observably but in small ways (promoting or suppressing different wheat varieties depending on soil, climate, and weather). These micro-adjustments don’t cause the system to fail, and reveal how the operating environment has changed. The system itself can then be easily restructured or redesigned to accommodate this new information.
Maintenance by design makes systems less—not more—fragile. By making the system more susceptible to maintenance, it forces the other three kinds of lower-level maintenance to happen. By making the system easier and cheaper to adjust, it allows the other three kinds of maintenance to make adaptive changes. Maintenance by design makes systems antifragile by giving them enhanced environmental perception and enhanced adaptation ability.
Maintenance by design has at least two principles:
Design the system so it shows clearly and swiftly how unexpected changes in the environment affect it without becoming incapacitated. Another way to put it is that when the system works unexpectedly because something in the environment has changed, it should do so quickly, non-catastrophically, and transparently—instead of opaquely storing up failure debt which then erupts catastrophically.
Design the system so it can be easily and undisruptively modified to respond to unexpected system behavior. Another way to put it is that when the system’s expected behavior changes, it’s inexpensive to restructure the system to accommodate the change.
Aspects of maintenance by design are captured by many concepts with great currency, such as designing for graceful degradation, designing for modularity, and agathonic design. And, though I do say so myself, the entire second half of my forthcoming book is about maintenance by design for innovation teams.
However, the best, most elegant, most compelling—but also most gnomic and perplexing—statement of Maintenance is in Christopher Alexander’s extremely weird but definitely worthwhile book, The Timeless Way of Building (WorldCat, Amazon US, Amazon UK)**:
In the commonplace use of the word repair, we assume that when we repair something, we are essentially trying to get it back to its original state. This kind of repair is patching, conservative, static.
But in this new use of the word repair, we assume, instead, that every entity is changing constantly: and that at every moment we use the defects of the present state as the starting point for the definition of the new state.
When we repair something in this new sense, we assume that we are going to transform it, that new wholes will be born, that, indeed, the entire whole which is being repaired will become a different whole as the result of the repair.
In this sense, the idea of repair is creative, dynamic, open.
It assumes that we are constantly led to the creation of new wholes, by paying attention to the defects in the existing wholes, and trying to repair them. It is still true that every act helps to repair some larger, older whole: but the repair not only patches it — it also modifies it, transforms it, sets it on the road to becoming something else, entirely new.
In this framework, we gain an entirely new view of the process through which a sequence of acts of building generates a whole. (p. 485-486)
Much needed, but still rare.
Because so few systems embody maintenance by design, it’s easier to illustrate these two design principles with four examples that violate them both:
Supersystems: A global system of tightly interdependent firms, each with significant balance sheet exposure to complex derivatives. Such a system stores up and conceals failure debt (layers of successive junky derivatives where no one understands what the underlying assets are) under the veneer of high function (very profitable banking operations) until the whole system reaches a threshold and rapidly breaks down (the 2008 GFC). Modifying such a tightly interdependent system was difficult, extraordinarily expensive (requiring enormous amounts of quantitative easing in all the major economies that has continued to the present day)—and ultimately unsuccessful, since things have basically stayed more or less the same.
Organizations: A leading industry incumbent, in a very comfortable place. Its internal systems are set up to manufacture and sell its highly profitable product with great efficiency. Such a corporation famously finds it difficult to understand and neutralize the threat posed by changes in the industry that favor smaller, newer competitors who do things differently to service emergent customer segments or take advantage of emergent technology. (This could very easily be incumbent non-profits too.)
Teams in organizations: A project team working on an ultra-high-stakes, do-or-die product launch, with extremely high expectations imposed on it from top management. The launch must hit on the specified date, or else there will be hell to pay—so project management and interdependencies are tightly controlled. There is no possibility of presenting rough work in progress for low-stakes feedback—only polished, perfect product will do. Such a team is almost definitionally unable to see new and disconfirming information and make the micro-adjustments to accommodate it, even if these adjustments are what’s likely to make the team successful.
Individuals: The goal-oriented person who is dead set on a particular career and begins investing heavily, early in life, in that endpoint. He (it’s always a he right?) takes all the right classes, does all the right internships, gets the right first job, goes up from rung to rung in the correct sequence at the appointed times. How does (or can) this person deal with any incipient revelation (or discovery) that the end he is working so hard to achieve isn’t the one he really wants?
These four examples highlight the pitfalls of not investing in maintenance by design. But these pitfalls materialize only when the operating environment cannot be predicted in advance.
If the environment were fully predictable, the system should be built instead to be highly efficient, precisely and wastelessly accommodating that environment. It would be the equivalent of a monocultural field of wheat, where the variety was chosen for its perfect fit with the completely predictable soil and climate in which it would be grown.
It’s when the future is uncertain that we should invest in maintenance by design, building expensively unpredictable systems which reveal when change happens and can be rebuilt easily to accommodate change.
Uncertainty in the environment is both growing and increasingly unavoidable. We see more evidence of this every day (demagoguery arising out of seemingly nowhere, stable polities breaking down unexpectedly quickly, long-established incumbent businesses falling apart within a few years, continents suddenly aflame for months, pandemics shutting down cities, etc). What we need is more maintenance by design—people, teams, organizations, supersystems that are designed to respond quickly and uncatastrophically to changes in the environment.
Unfortunately, though maintenance by design is self-evidently crucial for the world we live in today, we seem to barely invest in it at all.
It’s like some kind of weird cognitive disorder: most people love and cannot resist building (and being) these kinds of highly efficient, superoptimized systems which are ultimately fragile to unexpected change.
Those who resist may survive and prosper.
(And … if you’ve been reading this newsletter for a while, you’ll see that we’ve come back to the same realization—this time coming from a different direction—that there’s no escaping how appearing to be wasteful and inefficient is unavoidable if you want to prepare for uncertain futures.)
** Coming upon The Timeless Way of Building for the first time? I can recommend reading it as I did on my first approach: mug of tea, a couch in an empty room, winter morning light coming in through an east-facing window. Two things to note: Alexander does not use the term “maintenance by design,” and maintenance by design as a concept shows up only near the end of this quite long book but everything before and after is essential reading.
* Incidentally, if you’re looking for a short program this summer in wood, ceramics, metal, painting, or photography, look up Anderson Ranch. From my memories of the summer working there: the food is great, the mountains are stunning, the weather is superb, the workshop facilities are primo, the staff are excellent. They haven’t even paid me to write that.
As we can no longer really believe in modern democracy ...
You can find previous issues of The Uncertainty Mindset here. You can find me on the internet at www.vaughntan.org, on Twitter @vaughn_tan, Instagram @vaughn.tan, or by email at <uncertaintymindset@vaughntan.org>.