A few years ago, a scammer managed to steal over $100m from several big tech companies. The scam worked like this: he set up a company with the same name as one of their suppliers, he sent over an invoice, and then he got paid.

Two quick observations on this:

1. While $100m is a lot of money, it was not enough to have a meaningful impact on these businesses. Nobody had to update their valuation model to reflect this loss, and presumably most investors were oblivious to it. It just showed up as slightly higher operating expenses, but in the context of the target companies, that meant slightly-less-stellar-but-nonetheless-impressive operating expenses.

2. Any organization that consistently misplaces $100m will run out of money and die.

It’s easy to see how some individuals could fall for such a scam, but it’s obviously not the kind of mistake that everyone makes. But as an organization scales, those with high attention to detail can’t scale their attention, so they’re forced to divide it up. Of course, there are exceptions—a surprising number of them choose to scale the attention of a handful of people well past reasonable levels: a few sources claim that Elon Musk personally interviewed almost all of the first thousand employees of SpaceX; Michael Milken placed himself in the middle of the trading floor at Drexel Burnham Lambert, and would listen in on other traders to manage risk; Bill Gates was still writing code that shipped as late as 1989, when he was already one of the wealthiest people in the US; the person he displaced as the wealthiest American, Warren Buffett, likes to talk about things like the unit economics of selling a can of Coca-Cola or the relative margins of different fabrics for sale at Nebraska Furniture Mart.

But most organizations are not run by that specific kind of obsessive; sometimes instead of memorizing all of that information and being able to work as an individual contributor, it's easier to inspect your collection of 200,000 notecards detailing the meetings you've had ($, WSJ) so you can call the right person and ask them for help instead.

That's a particular model of organization, and probably one that's not optimal even for the people who are capable of it: sometimes being able to symbolically do extra work is itself a form of high-powered delegation; by 1989, the point was not that Bill Gates would write especially impressive code compared to the median Microsoft employee, but that nobody wanted Gates to scoff and insist that writing the code was the best use of his time.

The actual way delegation works is to resist the entropic force that makes some problem look like a rounding error. This shows up everywhere: Salesforce won't miss its numbers this quarter because one particular salesperson is a little slow. McDonald's won't need to write down the value of its intangible assets because one bathroom isn't cleaned thoroughly. Very few companies have died entirely because of excessive travel expenses (though it's often an early sign of other failures to control spending).

Anyone making decisions at this level can be fairly confident that there is no direct, urgent reason to do the best possible job. And yet, if you indicate to a Salesforce salesperson, however subtly, that you might someday be a source of annual recurring revenue, you will hear from them a lot. And if you visit a McDonald's it will generally have pristine, gleaming bathroom facilities (at least if you adjust for the foot traffic a typical location gets).

Delegation is, essentially, a relentless quest to make the individual decisions that are a rounding error for the company feel like weighty and important matters for whoever makes them. That can't be done entirely by fiat; there's some flexibility required because circumstances often change in a way that a) does change the optimal behavior for a given employee, but b) doesn't reach the level of importance that would require attention from the CEO or board of directors.

What makes these organizations work is that they're consistently breaking high-level incentives down into granular ones that actually affect people's behavior without locking them into some approach that doesn't make sense. It's not a good idea for a company with a sales team of thousands to set a quota for firmwide cold calls, for example; the level at which that kind of quota should be set is the level at which someone can see whether their team is getting better results from cold calls, cold emails, events, requests for referrals, or any of the other tools in the salesperson's kit. So the devolution is that highly specific key performance indicators matter at the lowest level, and as they percolate up they get more general and abstract, until they roll up to numbers that make sense across almost every industry: revenue, some indication of margins, some measure of return on investment, and some proxy for making sure the business actually generates cash in a timely fashion.

This is most visible in franchise companies. From an investor's perspective, a franchise business looks very low-risk: it's capturing a fairly fixed piece of the upside from a brand, without all the messy operational intensity of buying or building a location, staffing it, and operating it. But from an operating perspective, it's a nightmare. Every time Starbucks opens a new location, it's betting a brand worth tens of billions of dollars on one store, and, really, on every worker in that store. Any bad decision at the lowest level can threaten the brand equity of the entire business.

This may be why franchise-based models eventually slow down. After a while, the accumulated value of the business is so high, and the marginal benefit of one location so relatively low, that it doesn't make sense to risk so much brand equity on one more spot. Meanwhile, the existing store base is, presumably, continuing to compound the value of the brand, so a store-growth model slowly shifts to a same-store-sales based one.

The implementation of this is, in practice, signing a monstrously detailed franchise agreement, with the agreement devoting lots of space (starting on page 57 in this case) to enumerating all of the additional training sessions that will be required to cover material not mentioned in the agreement itself.

Any big company is, in historical terms, a miracle of human coordination. It's astounding that on any given day, 2.3 million Walmart associates spend their workday more or less the way Walmart CEO Doug McMillon wants them to. That's more direct influence on human behavior than historical heads of state could command. And entropy constantly pushes against this coordination working. The profits a company produces, and the share price that represents the expected future sum of those profits, is the source of organizational negentropy that justifies the otherwise herculean task of keeping everyone on task.

A Word From Our Sponsors

Read a book a day with Shortform’s POWERFUL nonfiction book guides.

Have a shelf full of unread books? Shortform can help you learn (and retain) the world's best ideas in a fraction of the time with their super-detailed book guides.

Take your understanding of important subjects to the next level with Shortform’s comprehensive coverage of must-reads in every nonfiction genre - from finance to self-improvement.

But don’t mistake Shortform for the other summary services out there that just give a brief overview. Their book guides include chapter breakdowns, insights, and even exercises.

Start your FREE Trial and get 20% off at shortform.com/capital.

Read More in The Diff

The question of how organizations scale and what happens when they fail to is a running theme in The Diff. Some posts that touch on this include:

• Just One Thing or Every Single Thing? On the payoff from CEOs who sometimes do the work of individual contributors.

• Is Mental Math Part of the Current Meta? More on how mastering the simple, easily-outsourced stuff has a surprising payoff.

• The Industry Demographic Transition Lottery ($): What happens when an industry stops growing and starts aging? And what happens when this reverses?

• Companies and Entropy looks at large organizations as a constant effort to avoid decision fatigue without delegating the essentials.

Help us learn more about you!

We're building a media kit and need some anonymous aggregated subscriber data. If you work in finance and have time for a 3-question survey, that would be very helpful:

Join the discussion!