Welcome back to The Diff. Here are the subscribers-only posts you missed this week:
Positive Spillovers from the Pandemic looks at some of the long-term positive inflections from Covid-19. There are lots of short-term adjustments, but lockdowns, work-from-home, and a scramble for vaccines have been a forcing function to change other factors permanently.
When the world's top economic think tank was located in Redmond, Washington: connoisseurs of corporate memos know that Microsoft in the 90s produced an incredible volume of thoughtful analysis, ranging from essays on how to value technology companies to thoughts on the bandwidth required to transmit smells. This essay looks at why, with examples.
Fin Du Tech: Trading Liquidity Risk for Model Risk: banking regulators don't mess around, and when a company evolves into a bank, it can suddenly find itself in violation of rules it didn't know applied. In this piece, I look at how banking regulators think about the risks they're trying to prevent—and why this is bad for pure-play fintech companies, but good for big tech.
One of the highlights of this week’s issues was that every post I wrote was inspired by conversations with subscribers. Everyone loves modeling a company as a flywheel, and as it turns out, The Diff’s flywheel is that it’s an extended tech/finance/economics improv session with a bunch of smart, interesting, and very opinionated people.
Subscribe now to join them.
This is the once-a-week free edition of The Diff, the newsletter about inflections in finance and technology. The free edition goes out to 19,421 subscribers, up 109 since the last edition.
In this issue:
The most successful companies are often synonymous with their founders, but that hides an important problem: a company is, in principle, immortal, but human beings, for the foreseeable future, entirely mortal. If a founder wants to build a truly important institution, it needs to achieve a bus factor greater than one. Bank of New York has been publicly traded since 1792, so it's outlasted many generations. BNY kept going even after its founder was shot by Aaron Burr, but some companies seem to disintegrate as soon as their original leader either leaves or gets distracted.
Which means that for the most interesting companies, success is not just a matter of a good founding, but of periodic refounding. This is true for all major organizations. The French conveniently use a versioning system for their governments, so we can explicitly see when Republic 4.0 was deprecated in favor of a more up-to-date platform. The US government likes to keep the old branding, so even though Truman presided over a government that was 13x as large in spending terms, as the one Coolidge ran, the two of them had the same job title. Stanford went through a re-founding under Terman, as did Harvard under Conant. Magazines, perhaps because the brand is a trophy asset but the writers are the actual assets, seem to attract many attempts at re-founding, most of which fail.
Refounding events at companies are a way to fight back against premature optimization. Whenever a company rides up a promising S-curve, it's because that company excels at solving a particular category of problem. Once that problem has been solved, the company's remaining problems are different. Sometimes, an organization can refocus without completely changing itself, but often it needs a more revolutionary approach. Refoundings don't necessarily require a new CEO, although they're often associated with it. They do generally need a painful catalyst, though.
Salomon: Ending the 80s
Michael Lewis made Salomon Brothers synonymous with the fast-money 1980s in his first book, Liar's Poker. In between (admittedly entertaining) scenes of gluttony, emotional abuse, juvenile pranks, and other debauchery, Lewis explains why the company was able to prosper so much: Salomon was willing to deploy its balance sheet and reputation aggressively any time there was an opportunity to make money. The company was perfectly willing to carry risk on its own books as long as it was confident in getting paid, and Salomon quickly granted more authority to traders who demonstrated an ability to make money.
(It certainly didn't hurt that Salomon was a bond house during one of history's greatest bond bull markets. Lewis notes at one point that the best way to identify future financial superstars in the 70s was to look at people who didn't fit in to the WASP banker stereotype. A nearly as accurate way to identify them is that anyone who got a job involving bonds when the federal funds rate was in the high teens, and kept it for a decade while rates dropped by more than half, did pretty well for themselves. Salomon, though, was one of the better places to be.)
Risk tolerance is often fractal. Some companies have it because they've judiciously assessed the payoffs of various decisions, and prudently concluded that the risk-adjusted returns available merit investment. And, in many more cases, it's because they have a temperamental love of ups and downs. This institutional hypomania is a great way to ensure that winning trades are appropriately large, but also encourages unseemly ways of winning.
This is exactly what happened to Salomon, a few years after Liar's Poker: a trader working there repeatedly violated treasury bond auction rules in order to buy more bonds than the firm's quota. At first, he openly gamed the rules; after getting caught, he submitted fake bids on behalf of real clients, without their knowledge. When the treasury department found out, Salomon was cut off from bidding for treasury bonds entirely—potentially fatal to a bond-trading firm.
Salomon's CEO and other senior employees ended up resigning, once it was clear that they'd known about the bidding problems for a while and had either been slow to act or had chosen to cover them up. Salomon, which relied on overnight funding for a substantial amount of its liquidity, could easily have gone into a forced liquidation.
Instead, they made Warren Buffett chairman. Buffett wasn't a trader or a banker, but he was trusted to do the right thing. His role at Salomon wasn't day-to-day management; it was to make the hardest high-level strategic decisions, and to represent the firm at a congressional hearing. (Sample quote, from when he explained the policies he expected to guide employees' decisions: "Lose money for the firm and I will be understanding; lose a shred of reputation for the firm, and I will be ruthless.")
Under Buffett, Salomon stopped doing business with Marc Rich, fired its political consultants, vetoed aggressive tax strategies, cut the bonus pool, and got rid of the company's law firm. After nine months as chairman, Buffett turned the role over to a successor, and Salomon was on its own again.
If the role of an investment bank is to get a good return on investment, Salomon's refounding was a mixed success. The company's stock was at $37 in 1991 before the scandal, and reached $76 the day the firm's acquisition by Travelers was formally announced. From pre-crisis peak to acquisition, it returned around 13% annually, a respectable performance but not extraordinary. On the other hand, Salomon's pre-crisis financial performance was nothing to write home about—compensation grew faster than revenues, so shareholders got the leftovers. And, of course, as a consequence of the refounding Salomon avoided committing major crimes and losing the US government as a client.
Amazon: A Store and More
Amazon is hard to evaluate, because the company seems to combine a series of last-minute opportunistic decisions with what appears to be a hundred-year plan to devour the world's GDP. The company started as a bookstore, because a digital bookstore could have three million SKUs; it's since expanded into basically everything else.
Any organization that functions at one scale will start breaking at some larger scale. As the Evernote's founder, paraphrasing Hiroshi Mikitani of Rakuten puts it, "[When you grow headcount by a factor of three,] everything kind of breaks. Everything. Your communication systems and your payroll and your accounting and customer support." At its IPO, Amazon had 151 employees (up from 11 the year before). Today, they're at 1.3 million. That's roughly eight cycles of everything-breaking. Suspiciously, the company remains unbroken.
One reason for this is the Amazon design pattern: everything is modular, so it can function as a collection of semi-independent companies with a few shared functions and one overriding vision. Early growth seems to have been ad hoc, but around 2002 (employee count: 7,500; everything-breaks count, 3) Bezos issued a mandate, famously described in this indispensable Steve Yegge essay: from now on, all interactions between different Amazon systems must be through APIs, and all of those APIs must be designed so they can be public-facing. Per Yegge:
This effort was still underway when I left to join Google in mid-2005, but it was pretty far advanced. From the time Bezos issued his edict through the time I left, Amazon had transformed culturally into a company that thinks about everything in a services-first fashion. It is now fundamental to how they approach all designs, including internal designs for stuff that might never see the light of day externally.
Three years of building internally-facing systems so they can hypothetically adapt to a future model is a very long time! Amazon was willing to dramatically slow its product launch pace in order to build a better platform for turning cost centers into revenue-generating platforms.
As a result, Amazon is a unique company, because every other company in the world thinks of problems in terms of build-vs-buy—is it better to pay someone else to solve the problem, or solve it internally? Whereas Amazon thinks of build-a-new-business-versus-buy: if the problem is hard to solve internally, at Amazon scale, then any solution that's good enough for Amazon is good enough for every other business on earth, and can probably be sold to them.
It's a software decision that leads to a cultural difference, since it turns every surprisingly difficult project into a potential source of upside. (It's not surprising that a company with a comparative advantage in doing things the hard way would have an employee who pulled four consecutive all-nighters and who later clarified that she did it voluntarily and loved her six and a half years at Amazon.) This corporate culture is not for everyone, some people certainly thrive there.
Building Amazon as a bookstore was a smart decision, but backing out the abstraction—Amazon as a collection of business Lego blocks that can be broken down and rebuilt for other purposes—was one of the best strategic choices of all time. It's unclear if something like this was always the plan, or if it's an example of a technical person insisting on creating the general solution, but in terms of market value added and consumer surplus generated, it's an extraordinary refounding.
Google and Social
In 2011, Google had several problems that were really one core problem. Business was fine—revenues were up 29% that year—but the company was losing employees to Facebook, and losing advertiser mindshare as well. Facebook could offer pre-IPO stock at what was then the hottest startup in existence (Facebook's stock was so attractive that more than one business was built just from helping early employees sell their shares before the IPO). And Facebook was also grabbing attention from advertisers. Since Google was the most successful Internet company, with slightly lower sales than Amazon, more than twice the market capitalization, and 15x the reported profits, the widespread pre-Facebook assumption was that the solution to making money online was search. Facebook's growth offered a counteragument: the way to make money online was social, and search would eventually be a set of poorly-customized, ill-targeted, semi-commoditized demand-harvesting tools, whose importance paled in comparison to social media's ability to generate demand instead.
Google did not exactly panic, but they did link every employee's compensation to the company's success in social. This was the era of Google+ (launched: June 2011; shuttered: April 2019). It was also the era of a Google+ button on seemingly every other service Google offered. The company also shut down Google Labs in an effort to direct more resources to social.
Google's pivot to social did not succeed at its implied goal of defeating Facebook. But it did lead to tighter integration across other Google services. For example, the company linked data across all of its services in 2012, instead of having separate silos for different sets of data. If Google is just a search company, that's a marginal benefit; better user-level information improves searches, but most of the information content is in the search terms themselves. But if the company is an advertising conglomerate, then this is a very powerful tool: it allows them to use high time-spent entertainment (YouTube), and high purchase-intent search (Google) to target ubiquitous display ads everywhere else.
In Google's case, the refounding as a social-first company didn't turn Google into a social media company. But it did make them a company with a social media site's attitude towards data: that every new datapoint makes ad targeting better, which subsidizes anything that can increase usage.
And, as it turns out, social was not an existential threat to Google. It was an important business that they weren't able to dominate (although to the extent that YouTube is a social media site, it's one of the most visited in the world). But their core business kept growing. For some categories, like casual games or niche direct-to-consumer products, social is a far better channel than anything else. And for other categories, like travel and personal finance, search is impossible to beat. What the social refounding showed Google was that their core business was more durable than it looked, and broader than it seemed.
Facebook and Privacy
Facebook is in the middle of an attempt at refounding. And, interestingly enough, it's inspired in part by a different refounding. From a memo by Andrew "Boz" Bosworth:
When I joined Microsoft in 2004, the only required reading was Writing Secure Code. The company had spent years getting dragged through the mud for the viruses prevalent on their platforms. While relatively few consumers faced material negative impact from malware, the possibility of it was a constant... By the time I joined the quality of new code impressive. Security got more attention than anything else in code reviews. Even the most junior engineer like me was held to the highest standards. They made product decisions that in many ways made their user experience worse but in exchange (rightly) gave consumers’ confidence that the system they were using was trustworthy... Today Microsoft is considered perhaps the most trustworthy software vendor in the world.
That's a fortunate example for Microsoft: a PR problem that can be fixed mostly with good software engineering.
One of Facebook's contemporary problems is also a PR problem that has some Facebook-favorable solutions at hand. In Facebook's case, the problem is that the company is synonymous with privacy violations, sometimes due to unfair coverage and sometimes due to deliberate design choices. In one sense, Facebook created the modern debate about privacy; by making sharing easy, it showed people that they don't actually want to share as much as they thought.
And since Facebook's business requires people to keep using its products, it can't ignore the PR; Microsoft might have been able to close enterprise deals by pointing to statistics on how rare malware was, but Facebook can't show a detailed and fact-rich PowerPoint to 3.3 billion monthly users.
By older standards, this will make Facebook worse. Using fewer signals to rank items in the news feed means showing people the content they're less likely to react to; moving more discussions into private groups and messages means that some users will get less content, and produce less in response. But Facebook's current view is that the company is at a local maximum: to hit the performance indicators it used to care about, it has to continuously erode user trust, which will eventually make growing those numbers impossible.
Facebook can do a lot with a privacy-focused business. Venmo notwithstanding, most people like to keep their finances and spending private by default, and present the results to the world in a controlled way. A trusted messaging system with identity built in is a good substrate for a payments platform, especially one in markets that don't have great non-cash payment options yet. If the Facebook refounding works, a company that earned its money from ads and didn't touch transactions will slowly evolve into a company that makes its money on transactions and has a legacy sideline in ads. (It wouldn't be the first time Facebook has made a pivot of that magnitude; when it went public, the shift to mobile was a risk factor, not the vast majority of the business.)
Refoundings in Perspective
Companies are legally immortal, but they're not immortal like vampires, more like the Turritopsis dohrnii that periodically ages in reverse and begins its lifecycle again. A big enough company has vast inertia, so its default state is to succeed for a while and then die from avoidable chronic illness brought on by lifestyle problems. A refounding is not a pivot; it doesn't mean giving up on the original business, just reframing the business in a way that's adapted to new constraints.
Some refoundings require a new leadership team, but some of them get their energy from an existing team. (In Facebook's case, the privacy-centric refounding involved losing, and then regaining, an early executive.) It doesn't seem possible to adapt an organization to refounding; to reach the point where it's necessary, the organization has to be over-adapted to a situation that no longer holds, and preventing that means preventing it from adapting well in the first place. But it's a useful pattern to keep in mind. The ideal time for a refounding is when a company looks unbeatable from the outside and vulnerable from the inside. That makes it an uncomfortable choice, but a necessary one.
Further reading: for much more on the Buffett/Salomon story, see Buffett: The Making of an American Capitalist and The Snowball.
Disclosure: I own shares of Amazon and Alphabet.
A Word From Our Sponsors
The Community Builders Fellowship is an 8-week remote program where fellows learn how to create communities, get them off the ground, and sustain the magic as they scale.
We aim to make the On Deck Community Builders the forefront of knowledge creation around community (frameworks, models, tools, best practices).
To do so, we are bringing together 100 active community builders and leaders, people who are actively growing an existing community with high impact potential and demonstrated level success (traction, impact, badass story). Most importantly, we are looking for people with ambitious goals and great challenges to tackle in order to take their communities to the next level—to make sure there are great conversations heard around the room. Fellows will include Li Jin, Sahil Lavingia, and Jeff Morris Jr.
More details about the program:
The program is structured around 3 pillars: people, knowledge and tools. Fellows will learn from other active community members in their cohort, as well as from seasoned founders, community leaders and domain experts with a lot of knowledge in the field. Last, but not least, they will gain access to a growing base of tools that will enable them to fast-track their communities.
The Other Bitcoin Standard
Balaji Srinivasan has an essay arguing that India should adopt digital currency, backed by a Bitcoin reserve. This is a very bold response to the Indian government's potential ban of Bitcoin. And the argument touches on many important topics—techno-nationalism, globalization in services, platforms' freedom to moderate, and of course monetary policy. Convincing India to create a currency backed by a store of value other than gold would be a challenge. The Economic Times mentions gold more frequently than any general business publication; right now there are four articles about gold on their homepage. On the other hand, it's politically astute to highlight a digital system's auditable-by-default nature for a country that remonetized a few years ago to crack down on corruption.
Overall, the piece is an interesting sketch of how a country could declare partial independence from the dollar system, and bootstrap independence from the US tech ecosystem. The tradeoff is that a country large enough to support multiple domestic tech giants would produce enormous buying demand for Bitcoin, if that's the asset it chose to back its currency; a smaller country could move more of its reserves into Bitcoin, but as a purely financial decision, not as the basis for pursuing technology sovereignty.
FT has a detailed look at the companies and countries pursuing clean energy ($). One important point is that, relative to hydrocarbons, renewable energy makes geopolitics less fraught; nearly every country has some way to generate its own power, so petro-politics in the Middle East and gas politics in Central Europe would be less significant.
Meanwhile, China has set a goal of zero net emissions for 2060, and plans to start trading emissions credits this year, but is adding coal power plants five times faster than the rest of the world combined. Because China's goal is so far out—a coal plant built today would be close to the end of its useful life in 2060—this is an implicit form of emissions finance; borrowing more carbon today in order to invest in industries that will produce negative net emissions in the future.
Exxon may add shareholder-activist-turned-more-general-activist Jeff Ubben to its board. Ubben is an experienced multi-industry capital allocator, and for energy majors a transition to low emissions is a capital allocation problem: turning relatively short-dated and high-variance assets tied to hydrocarbons into long-term and low-variance assets tied to renewables.
Loyalty and Information
Skift reviews the way airlines are keeping people in loyalty programs when there aren't many flights. One reason for this is purely financial, since keeping members is cheaper than acquiring them. But another is informational: people who earn lots of loyalty points are more likely to have mobile, remote-friendly jobs. Since air travel is, economically, a form of lead generation for credit cards, a key reason to keep people in the system is to figure out what the new route map should look like as travel demand returns to normal. If New York and San Francisco have lost some of their highest-spending residents, future route maps will reflect credit card spending's new center of gravity.
The Chip Shortage
Automakers have started giving out more details on the impact of the global chip shortage, which, due to demand swings, production lead times, and demand from other semiconductor-heavy products, has hit car companies the hardest.
The US and Taiwan are collaborating to divvy up the existing supply of chips ($, Nikkei). (Taiwan's economic minister says the US did not directly ask for help, but got it anyway.)
China's SMIC says that US sanctions have prevented them from growing capacity to take advantage of demand ($, Nikkei). Sanctions have made Taiwan and South Korea more central to the supply chain; China won’t stop assembling at exporting devices, but will have trouble building an internal supply of the most high-margin components.
I mentioned in yesterday's issue ($) that car companies may respond by committing to purchases earlier so they can lock in supply. FT says they are considering this ($), although it's a last resort. Instead, the plan is for a "better early warning system" for supply bottlenecks. An early warning system is reactive, and could make demand less stable. In effect, it means that whenever foundries are operating at a higher than usual capacity, car companies will order more chips, further squeezing capacity. And since this is known in advance, it gives other chip foundry customers a reason to lock up their own supply.
This story has been interesting because, as I wrote about late last year ($), chips are turning into the swing factor in more and more of the global economy, and production is increasingly localized to a few small countries with geopolitical risks. One of the most important stories in geopolitics in the last half-decade has been the US's declining reliance on overseas oil supplies, which allows the country to be less engaged with the rest of the world. As chips get more important, the specific causes and locations may change while the ultimate effect remains the same—that the US economy's supply chain inevitably leads to a more complicated foreign policy.
Stimulus: How Much is Too Much?
Larry Summers has a much talked-about piece arguing that Biden's stimulus plan is too aggressive. The key argument is that in 2009, Obama's stimulus was about half of the US economy's estimated output gap, while Biden's is at least three times the size of the output gap. One important consideration is that a Covid response program has a more definite end date, since the US population is approaching herd immunity at an accelerating pace. Since economic stimulus and inflation happen over time, a spending plan that's mostly tied to addressing a problem with a finite length is less risky. Summers is right that a large stimulus will mean 2021's data will show high inflation numbers, but a) inflation has been low enough for long enough that the Fed has committed to "catch-up" inflation, and b) the mix of consumer spending in 2021 will be impossible to compare to 2020, and hard to compare to other periods, so firm conclusions will be difficult to draw.