The Future of Remote Work is Not That Remote

Plus! Multilateralism, Unilatéralisme, Wanamaker Updated, Domestic Fabs, and More...

This is the weekly free edition of The Diff, the newsletter that tracks inflections in finance and tech. Last week’s subscriber-only posts:

Subscribe now and get instant access to the full library, and future paid-only posts. Coming next week: posts on private markets, reserve currencies, and much more.

The Future of Remote Work is not That Remote

A specter is haunting real estate owners. The specter of eternal work-from-home. A CNBC survey last month found that 42% of those surveyed were working from home, though given that the respondents all cited some sort of “work,” it omits the 36.5 million people who are suddenly not working at all.

Landlords worry that the situation is permanent: people who were forced to work from home will adapt and get better productivity, and nobody will pay $1.45m to live in the Bay Area or $59 per square foot to work there.

But the winners and losers aren’t linear. Some people hate working from home (a broadly descriptive term for these people is “parents”—I love my kids, but at the office nobody ever interrupted a conference call to ask me for a hug). And offices exist for a reason, or several. Investors in Zoom are partly betting on the end of the office as an institution, but Zoom has $87.8m in operating lease obligations on the books, so clearly they weren’t betting on full-remote any time soon.

Covid-19 and the attendant shift to work-from-home will not end the office, but it will force companies and employees to rethink the office. And when they do that, they’ll be optimizing against several constraints.


Knowledge work has a built-in agency conflict: output is a function of skill and focus, skill comes from legitimate interest in a field, and that legitimate interest means that spending your entire workday on an Internet-connected device is incredibly distracting. It happens in finance: give someone who loves investing access to a Bloomberg terminal, and they can spend all day researching some esoteric microcap stock or weird macro hypothesis instead of doing their day job, which involves getting this quarter’s EBITDA estimate from +/- 2% down to +/- 1.5%. It happens in software: why import a library when you can spend ten times as long elegantly reinventing the wheel?

The especially difficult part for hiring managers is that people who are passionate about the job can blow away expectations at job interviews, and then disappoint at the job itself. Conventional wisdom suggests that money is a good way to resolve the mismatch, but it’s not perfectly motivating, especially for the best employees. Money has declining marginal utility, while pursuing passion projects basically doesn’t, so employers aiming for a purely economic relationship get far less than 40 hours a week of output from a 9-to-5-er. But money plus shame plus camaraderie delivers more benefits; when your boss and your peers can literally look over your shoulder, you’ll work a little bit harder and save the fun stuff for nights and weekends.[1]

The output of knowledge workers is extremely skewed based on focus. The productivity tiers seem to be:

  1. <10% focused on the job at hand: meaningful risk of getting fired.
  2. 10-50% focus: “meets expectations,” gets regular raises.
  3. 50%+ focus: superstar, 10x engineer, destined for greatness.

There’s a lot of wiggle room in tier two because of range restriction. For a company that hires employees based on a given level of output, some will be smart but relatively lazy, and some will be less smart but relatively hardworking. But almost no one can maintain day-long intense focus on something they’re not genuinely excited about, and the existence of Linux, , blogs, novels, and Wikipedia is a testament to the gap between what people love to spend time working on and what they can get paid to do. (This is also one reason company productivity per worker declines so much as the company grows—the motivation goes from “this is the only place I can be to change the world in the specific way I care about” to “The 401(k) matching is nice.”)

Since part of the trouble with measuring knowledge worker output is that they’re getting hired for knowledge, i.e. things they know or know how to do that nobody else does, measuring results relative to potential always involves some guesswork. The closest you can get to a definitive answer is a process of elimination: discourage all obviously unproductive behavior, and the most interesting way to spend the workday is on actual work.


The boss-to-peer relationship matters, but the peer-to-peer relationship matters more. The narrative around successful companies usually focuses on the CEO, both because it’s a better story (you can’t write a story about Facebook with 45,000 protagonists) and because the CEO’s efforts are more visible. But if you talk to CEOs at growth companies, they usually say the task they spend the biggest chunk of their time on is recruiting. Teams matter.

And it takes less effort to treat someone as part of a team if you bump into them regularly. Zoom meetings and phone calls are better than nothing, but there’s something to be said for random serendipity. There are interactions that are a) valuable, b) not worth interrupting someone over, but c) worth talking about if an interruption happens at random. And those interactions are less likely to happen virtually—the break room coffee machine should be budgeted as an R&D expense.

One very specific way colocation affects morale is the peer-to-peer version of measurement. Walk through an office and you get a good sense of what levels of focus and absenteeism are acceptable. This is a form of leverage; if the CEO takes two-hour lunches and leaves at 4pm on Fridays, that’s way easier to notice in person than on Slack. But business is pretty Darwinian. In the long run, the only behaviors that matter are the behaviors at the best companies, because everyone else ends up working for them or switching to a different line of business.

This is intangible, but it’s important. A remote-only company can have a culture, but it doesn’t have a vibe. And vibe is a force multiplier.

Bandwidth and Latency

Until this season, Y Combinator told startups to do two things:

  1. Spend 100% of their time coding, talking to customers, going to the gym, or asleep. For pre-launch startups, that means 100% of their work could be done from literally anywhere.
  2. Only do it from the Bay Area.

There’s one important reason for small companies to locate in the Bay Area: meetings. In-person meetings are higher-bandwidth than remote meetings, although it’s unclear why. Maybe posture and fidgeting are subconsciously informative, maybe there’s a subtle Uncanny Valley effect, maybe handshakes and fistbumps transmit otherwise-secret information. Nobody knows for sure. (I don’t doubt that there are studies, but I do doubt that they replicate.)

There’s a correlation between how important someone is and how hard it is to get a meeting. So there’s a certain set of important meetings that, if they happen, have to happen right then. If the company you’re trying to partner with suddenly has a free half-hour on their schedule, do you really want to be a flight away? You could, of course, live somewhere else and fly in once a week for meetings; the market-clearing price of SF real estate has been partly determined by the cost of round trips from there to Vegas. But the missing meetings are essential.

Where To Next?

I’ve been talking a lot about the economics of locating in the Bay Area. New York, LA, Chicago, and a few other cities have similar clusters. But very little of this argument applies to smaller cities. For a tech startup in my hometown of St. Louis, there are reasons to be local—good schools, access to investors. But the more important that startup’s next meeting is, the more likely it is to take place in New York or Mountain View.

Commercial real estate will be impaired for as long as there’s not an effective vaccine for Covid-19, but we’re already entering a period where travel is slowly coming back, and in-person meetings—mask up, fist-bumps only, no shared food—are happening. Even though in-person meetings aren’t happening much now, people are planning for them to happen soon; 2021 is far enough in the future that we can all pretend it will be fairly normal, aside from the economic depression. If remote work is an option, it will give companies more flexibility in timing the expansion of their office footprint. Instead of doubling their square footage, they can halve their in-office workers. So from an employer’s perspective, the supply elasticity of office space just went way up; as long as there’s a free conference room somewhere, there’s no need to expand.

Big cities have another center of gravity: couples. New York and SF are very hard to afford on a single income, so they’re cities full of couples who both work well-paid jobs and split the rent. This means that relocating to a cheaper area requires the approval of four parties: you, your partner, your employer, their employer. If you both work for Twitter, it’s easy. If one of you works for a big law firm, on a trading desk, or for Apple, it’s just not going to happen. And the companies itching to send employees back to the office are also the companies that will be bringing back face-to-face meetings first.

Cities have network effects, and these effects are superlinear. If Covid-19 imposes a shock on every city, the ones with the weakest network effects will unravel, but the rest will survive. The net result is still negative for commercial real estate, obviously—and WeWork, a highly levered real estate play whose model involves the densest possible seating, is the worst affected. It’s also harmful for luxury commuting options; the daily-Uber-commute cohort may be a weekly-Uber-commute cohort in the future. And it’s awful for cities that don’t have a critical mass in some industry (there’s some evidence in the literature that the network effect is stronger for service economies than industrial economies, though Detroit in the 20s-50s and Shenzhen and Nagoya today make me skeptical). The density of the network effect has a paradoxical side effect: if office capacity is less of a limit, but people still want to work in the same city, it can mean more of them living nearby to work from home.

It’s true, in a sense, that the future of work is remote. It’s just that the future of remote work for the most important companies will involve paying through the nose so you can work-from-home in a home conveniently close to the office.

[1] Being my own boss has been a cathartic experience. In my last real job, I’d occasionally run up against a tight deadline and, due to my own poor project management, I’d end up keeping my direct report in the office for an absurdly long time. Now, I’m both the beleaguered junior working until 2am and the terrible boss insisting that the day’s not over until the project’s done. It’s nice. (Except when the baby decides to wake up for the day at 5.)

A Word From Our Sponsors

Here’s a dirty secret: part of equity research consists of being one of the world’s best-paid data-entry professionals. It’s a pain—and a rite of passage—to build a financial model by painstakingly transcribing information from 10-Qs, 10-Ks, presentations, and transcripts. Or, at least, it was: Daloopa uses machine learning and human validation to automatically parse financial statements and other disclosures, creating a continuously-updated, detailed, and accurate model.

If you’ve ever fired up Excel at 8pm and realized you’ll be doing ctrl-c alt-tab alt-e-es-v until well past midnight, you owe it to yourself to check this out.


I have a new piece out in Marker talking about how Lyft and Uber were converging on comfortable duopoly status, and how that’s still attainable post-pandemic.

(A few days after I published that, Bloomberg came out with a profile of Dara Khosrowshahi and his dealmaking prowess. Worth reading as a preview of what’s to come with Uber.)


The WTO hasn’t been able to settle disputes for months, and, sooner than expected, there won’t be anyone in charge, either. One element of the US/China conflict has been China’s efforts to lobby for control of multilateral organizations like the UN, WTO, and (notoriously) the WHO. From the US perspective, this looks like China trying to take over; from the perspective of every other country, it looks like essentially US-run institutions being put under new management.


But naming yet another European may not go down well with other blocs. Azevêdo’s immediate predecessor Pascal Lamy was French. There is a growing feeling among trade diplomats that it’s Africa’s turn to lead to the trade organization. Several African names are circulating, but a unifying candidate has not emerged so far.

China has had an increasingly active presence in East Africa over the last decade. The US’s response has been halfhearted.


It’s a fun thought experiment to ask roughly how many independent countries there are in the world. The UN has 193 members and two observer states, which is a start, but we can take that number down a lot: most countries are not in a position to determine their own fate, diplomatically, militarily, or economically. The most parsimonious answer is that a country is only independent if it a) has nuclear weapons, and b) acts like it does.

France has spent a lot of the postwar period trying very hard to be as independent in practice as it is in theory. France has nuclear weapons, may or may not have tried to undermine the Bretton Woods system, didn’t participate in the war in Iraq, and has otherwise insisted on not being a US satellite. Two recent headlines confirm this pattern:

51% of Ad Dollars Spent, Not Wasted

A recent study showed that only about half of online ad spend ultimately reaches the publisher. Given how complicated the ad tech stack is, and how much data advertisers could use but don’t have, that didn’t strike me as especially high or low. This writeup is a good summary of the adtech industry’s perspective on the matter.

It turns out John Wanamaker’s quote—“I know half the money I spend on advertising is wasted, I just don’t know which half”—was true. And it’s a stirring example of market efficiency at work: you can waste 0% of the money you spend on advertising, as long as you spend half your budget on avoiding waste.

Domestic Fabs

Taiwan Semiconductor plans to spend $12bn to open a new fab in the US. I highlighted the story a few days ago that the Trump administration was putting pressure on TSM and Intel to do exactly this. As I said then:

Intel’s US plants mostly make Intel’s own devices, but contract manufacturing has positive spillover effects: if more supply chains run through the US, more adjacent businesses—whether R&D and design upstream or electronics manufacturing downstream—can be located in the US.

In probably-related news, the ban on US firms working with Huawei and ZTE has been extended another year, and despite using clever loopholes, the companies are suffering.

I view domestic fabs as the important reshoring story right now. They’re knowledge- and capital-intensive, they have positive externalities, and they’re exceptionally hard to get right.

Google Privatizes the Web

Successful consumer Internet companies have a two-stage model: privatize an economic commons, and then maximize its value and tax it. The link graph wasn’t owned by anyone, but now Google’s algorithm defines the link economy; the social graph wasn’t owned by anyone, and now it lives on Facebook. But the upside is that these companies do, in fact, exercise careful stewardship over the assets they own. For example, Google is blocking resource-heavy ads at the browser level. These ads are a negative externality: annoying to the user, profitable to the publisher and advertiser, and nobody’s problem in particular. But Google owns enough of the web that anything that increases usage on the margin is a measurable benefit to them.

Restaurant Software; Restaurants as Software

In my V/L recovery series, I argued that the two versions of recovery are both corporatized, and the choice is between front-end and back-end corporatization. The front-end version is that small businesses fail and big chains take over their locations and get their customers; the back-end version is that small businesses prevail, with the help of large SMB-focused software and service companies. I’m still undecided on which we get. On the front-end side: McDonald’s has laid out its reopening rules. While McDonald’s sells hamburgers, it’s really a software company; its main asset is the source code to running a McDonald’s, which gets implemented through buildings, kitchen equipment, and workers. That software can adapt at scale much faster than the wetware competition. Meanwhile, OpenTable says 25% of US restaurants will fail, but points to improving dining trends at states that are opening up.

As you can see above, I’m experimenting with advertising in The Diff. If you’d like to reach an audience of 5,200+ subscribers (open rates >50% for the weekly free edition, subscriber count +49% M/M), reply and we’ll set something up.