A fence? But why?
A sensible theory about Blockbuster was that they made so much money on late fees that they eventually went broke. Everybody hated late fees, but Blockbuster kept charging them — at thousands of stores! Then at hundreds of stores! Then at dozens, then one, then none. It’s easy to ask why they didn’t see this coming. Surely, their customers told them they hated late fees; a competitor once built its entire PR story on a (probably fictional) tale about spending $40 in late fees on renting Apollo 13.
Since Blockbuster went under, the critics were right, but it’s a broken-clock criticism. Hacker News commenters and journalists are always wondering about what seem like pathologically dumb business decisions: Why do so many TV ads still get sold at upfronts? Why didn’t taxi companies get together and build a slick app in 2012 to strangle Uber in its crib? Why did music labels blow so much money on advances for albums that wouldn’t sell, and why do book publishers still do this? Why do gyms sell so many memberships in January that the squat rack has a five-hour wait? Why do recruiters get such egregiously high fees? 20%, really?
There’s a quote from G. K. Chesterton, every atheist’s favorite Catholic apologist, that goes like this:
In the matter of reforming things, as distinct from deforming them, there is one plain and simple principle; a principle which will probably be called a paradox. There exists in such a case a certain institution or law; let us say, for the sake of simplicity, a fence or gate erected across a road. The more modern type of reformer goes gaily up to it and says, “I don’t see the use of this; let us clear it away.” To which the more intelligent type of reformer will do well to answer: “If you don’t see the use of it, I certainly won’t let you clear it away. Go away and think. Then, when you can come back and tell me that you do see the use of it, I may allow you to destroy it.
You can rephrase it in more pointed and hostile terms: is it more likely that everyone is missing something, or is it more likely that you are? If you’re so smart, how come you’re not rich? Anybody can look at how an industry behaves, and say it’s stupid.
You can also be more middlebrow and say it’s not stupidity, just an institutional imperative: sure, everybody involved knows that doing X is dumb, but nobody wants to rock the boat. If you did not-X, you could get in trouble, whereas nobody gets fired for doing X. That’s really attributing these behaviors to stupidity with more syllables.
And that’s not what GKC would want us to do, not at all. As it turns out, most of these weird practices are explicable without turning to stupidity as a default explanation. The path from “If you’re so smart” to “Oh, you are rich, carry on,” passes through an articulation of why a smart person might behave in a superficially stupid way.
A Brief Tour of Ostensible Incompetence
Take the music labels: the toy model of their economics in the 90s and early 2000s was: they blow a bunch of money on paying unknown artists to make albums, most of which never sell. To recoup these bad investments, they have to take well-known acts, bundle together two or three hits with a dozen forgettable tracks, and then sell the whole package for $20 per CD. Meanwhile, they’re constantly suing their best customers for using Napster, Kazaa, and Bittorrent. It’s craziness all around.
The problem with this view is that it misunderstands what a label is: a music label is a specialized venture capital firm that’s in the business of losing lots of money on advances in order to find the next Nickelback. To find that Nickelback, you have to pay a for a lot of studio time for bands that are, somehow, even worse. But once you’ve found your nickel-plated golden goose, you need to recoup the investment. And how do you do that? By charging whatever the market will bear.
The story of the record industry is not a story of evil and stupid monopolists. It’s a story of savvy venture capitalists whose entire business model fell apart the minute iTunes found a way to dilute their equity stake in hit bands down to nearly nothing.
Gyms are a totally different story. Here, what’s going on is a form of customer segmentation. A gym can divide its customer base into two categories: suckers, and walking billboards. A sucker is someone who makes a New Year’s Resolution and precommits by signing up for a full-year gym membership. They go regularly for a few weeks, desultorily thereafter, and by Spring the squat rack is blissfully free again.
(If you must make a New Years resolution to get fit in 2019, try this one: burpees and sprints every day in January, and don’t sign up for a gym until either a month passes with no missed workouts or someone says “Hey, have you lost weight?”)
Your walking billboard customer is different: they’re going to the gym regularly, month in and month out, no resolutions required. Aren’t they annoyed when it’s crowded? A little bit, but if you’re conscientious enough to sustain a gym habit for an entire year, you’re probably conscientious enough to set your alarm early. I’ve never waited for someone to finish a set before 6am.
Now let’s talk about recruiters. A recurring theme when you talk to programmers is that recruiters are ludicrously overpaid, and also annoying and incompetent. There are two problems with this:
- The median recruiter who contacts you is not very competent, because some recruiters target a wide funnel and some target a high conversion rate. The recruiter who spends one minute vetting each LinkedIn profile can contact 480 people in an eight-hour day. The recruiter who spends an hour reviewing Github profiles for every cold email she sends will hit eight people in the same time. So a small number of lazy recruiters (intellectually-lazy, at least) are responsible for a disproportionate share of programmers’ interactions with recruiters. Fortunately, this brand of headhunter burns out fast; unfortunately, there are always more of them, and they do eventually get lucky.
- The money: let’s switch perspectives, from the recruiter to the client. You are the CEO of a small, venture-funded startup. You hear some cool tech scuttlebutt from a friend of a friend of a friend: the product you’re prototyping is a great idea! In fact, your biggest competitor is adding it as a free feature in six months! Congratulations: you’re doomed. Unless you can launch fast, that is. If a recruiter can reduce the time it takes for you to double your engineering headcount, such that you can get paying customers before the competition has finished launching their product, it’s worthwhile to pay up. Better to spend $30k apiece on hiring programmers when you need them than to lose precious time and only get the talent when it’s too late to use it.
When I first heard about recruiters’ fees, I balked. But imagine wasting your $10 million Series B because you wanted to save 1% of that by hiring the old-fashioned way.
Crazy, or Crazy-Like-A-Fox? A Framework
You can make a meta counterargument here: I’m implicitly shifting the critique but making the same argument. Instead of saying that companies are stupid, I’m arguing that the people observing those companies are stupid. Fair!
My lodestar in these matters is to remember that everyone responds to incentives. Eventually. Individual employees respond to the incentives that management sets; management responds to the incentives the board sets; and the board, i.e. the shareholders, respond to a simple incentive: if they mess up, they lose their money.
People making decisions have skin in the game, and observers do not. Furthermore, journalists have skin in a different game: they get paid to generate clicks, and surprising outrage sells better than fundamentally boring stories about why this particular corporation’s stupid and/or evil activity is in fact perfectly benign. I’ve previously described some white-collar service professions, like banking, consulting, and accounting, as the tofu fields; they absorb the flavor of whatever business they interact with. Journalists might be the Miracle Berry field, instead: they absorb a touchy, hyper-critical attitude towards whoever they write about. (There are exceptions! I cherish them all.)
Why don’t experts step in and clear things up? Sometimes, they do. But the media make it hard, and social media makes it harder still. There is no thousand-word explanation of why something is reasonable that does not contain a 280-character quotable bit that makes it sound abominable, and you don’t win any points for guessing which of these the Internet is more eager to read and share. The tyranny of the pull-quote keeps us more ignorant than we should be. (Notice that I didn’t try to explain the quirks of a business I’m actually involved in? There’s your reason.)
Why Don’t We Just Automate This?
Chesterton’s Fence is a powerful framework, because once you reverse-engineer something you can figure out why it works and how it breaks. If X happens, not because people are stupid, but because of Y and Z, you know that watching for the moment when Y or Z changes will point you to opportunities. (Many of the successful tech startups of the 2000s could be described as “which dumb idea from the 90s is a better idea now that everyone has an Internet-connected computer in their pocket?”)
Of all the weird corporate aberrations, perhaps the biggest is: why are there so many knowledge-worker tasks that are done manually, rather than through software? Get data, put it into a pivot table, figure out what the data says, make a decision — that could be ten lines of code, but in many cases it takes phone calls, faxes, copying, pasting, Ctrl-Z, paste-as-values — whoops! Excel crashed again! — copy again, paste-as-values again, click this, save that. Ugh.
Consider the incentives of a manager who has some task that’s been done by full-time employees, and could be done automatically instead. The tradeoff is not high marginal cost versus zero marginal cost: it’s known marginal cost versus uncertain capital cost. A bird in the hand is worth two in the bush when “two” is the expected value but the range of outcomes is somewhere between ten birds and negative six.
Never attribute to cowardice that which can adequately be explained by the Capital Asset Pricing Model: a low but certain return is worth more than a higher but more variable return, and people vary in their risk tolerance. People at big companies tend to have low risk-tolerance: if they’re right, their bonus is about 1% of the upside they’re responsible for; if they’re wrong, they’re 100% fired.
Managers have low risk-tolerance because big companies have low risk-tolerance: part of what investors are paying for is stability. Sure, one big business process overhaul could be beneficial, but what about a dozen? A hundred? What if you’ve fired half of your headquarters staff before you realize that — surprise! — the consulting fees you’re paying to maintain your system exceed the salaries you were paying in the Bad Old Days.
These things happen.
I don’t purport to explain why companies will never automate, just why they haven’t done it all at once. Software’s Long March through the Institutions is driven by a gradual drop in the standard deviation of the cost of big software projects. This dynamic actually presents a long runway for the SaaSification of the white collar job: every time something new gets implemented, the range of possible outcomes shrinks, so more automation is possible. The companies with a comparative advantage in estimating costs will win, and their wins will compound — this applies both to outside service providers and to companies that take their automation in-house. The outside service providers have two advantages, though: one, they don’t have a legacy process to defend, and two, they’re often earlier in the business life cycle, at a time when they can afford to be — are, in fact, required by their investors to be — cash-flow negative.
(As a sidebar, one of my favorite case studies in knowing what you got wrong involves this very subject: Warren Buffett invested in IBM in 2011, and sold out at a loss earlier this year. When he invested, he described his due diligence process: he talked to the CTOs of Berkshire companies. That was a sensible thing to do based on how the enterprise software industry worked at the time: big projects lasting several years, with juicy maintenance contracts afterwards. What Buffett had no way of knowing about was the Bring-Your-Own-X trend. The way companies used to choose version control software was that there would be an RFP, companies would pitch to the CTO, and eventually there would be a big fat contract. And now the way it works is more like the CTO is talking to some devs, and says “By the way, what’s a ‘bitbucket’?” and they explain it, and he blanches for a bit and says “And how much does it cost to have a version of this that doesn’t put all our data on somebody else’s server?” and pays it. SaaS companies: your minimum viable product is something an engineer can spin up fast for a side project; the “enterprise” pricing tier on your website should start with an area code, not a dollar sign. You’ll like the calls you get.)
G. K. Chesterton: Reverse-Engineer Extraordinaire
There’s a naive version of G. K. Chesterton that says we should always hew to tradition: if burning witches alive was good enough for my ancestors, then by golly, get me some lighter fluid. But really, Gilbert is suggesting that you reverse-engineer the reasoning. Traditions made sense for a reason, and once you can articulate that reason, you know why they don’t make sense any more.
A lot of investment research works this way. The basic process is:
- Make a very short list of the variables that affect an asset’s price.
- Make a short list of the things that predict those variables.
- Predict those things better than the other guy.
In his essay, In the Beginning Was the Command Line, Neal Stephenson has a brief interlude about cars (it’s part of a longer interlude about cars as a metaphor for operating systems, while the whole essay is an extended interlude about operating systems as a metaphor for everything):
One of my friends’ dads had an old MGB sports car rusting away in his garage. Sometimes he would actually manage to get it running and then he would take us for a spin around the block, with a memorable look of wild youthful exhiliration on his face; to his worried passengers, he was a madman, stalling and backfiring around Ames, Iowa and eating the dust of rusty Gremlins and Pintos, but in his own mind he was Dustin Hoffman tooling across the Bay Bridge with the wind in his hair.
…
Sure, the MGB was a lousy car in almost every way that counted: balky, unreliable, underpowered. But it was fun to drive. It was responsive. Every pebble on the road was felt in the bones, every nuance in the pavement transmitted instantly to the driver’s hands. He could listen to the engine and tell what was wrong with it. The steering responded immediately to commands from his hands. To us passengers it was a pointless exercise in going nowhere–about as interesting as peering over someone’s shoulder while he punches numbers into a spreadsheet. But to the driver it was an experience. For a short time he was extending his body and his senses into a larger realm, and doing things that he couldn’t do unassisted.
Really good salespeople, executives, and analysts are like that MGB driver: they feel every little detail, because the underlying context makes all those details vivid. It’s a surreal experience to go to a store after reading the owner’s last couple annual reports and conference call transcripts: every detail is at the intersection of heated arguments, careful analysis, wishful thinking, office politics, and information asymmetry. It’s like seeing an extra dimension.
95% of the time, the Chestertonian view is a bit of a bummer: you see what looks like a great arbitrage between bad incentives and good ideas; you do a bunch of research; and ultimately you discover that there’s no arbitrage at all, and the world is a little bit more complicated than you thought. It could be worse: you don’t have an opportunity to make money (yet), but at least you learned something.