A Taxonomy of Drawdowns
Plus! Watercooler Shows; Smart Thermostats; Substitutes and Complements; Monetization; Apple Ads; Diff Jobs
Welcome to the free weekly edition of The Diff! This newsletter goes out to 40,782 readers, up 655 since last week. In this issue:
A Taxonomy of Drawdowns
Substitutes and Complements
A Taxonomy of Drawdowns
If you're trying to get a quick sense of where a company is right now, the best place to start is with their income statement, cash flows, and balance sheets for the last few years. But one great complement to this is to look at an all-time chart of their stock price. In looking at many hundreds of since-inception stock charts, one of my favorite long-term patterns has been the fact that some companies go through absolutely horrendous drawdowns, but still manage to turn out okay or even achieve above-average returns.
There's a broad taxonomy of these:
Some companies are industry and sector bubble survivors. Some of the big, established tech companies had this: Amazon lost 94% of its value in under two years starting in 1999, and Nvidia had a similar drop in the first three quarters of 2002.
Some companies are cyclicals, and are highly levered to their cycle. Mattress maker Sleep Number went through two giant drawdowns, losing 97% of its value from 1999 through 2001 and then 99% from mid-2006 through early 2009.
And some businesses have gone through pivots that didn't look like they would work out, but that ultimately turned out fine. Netflix was down 82% from mid-2011 to mid-2012, and IBM lost three quarters of its value—astonishing for a company that practically defined blue-chip growth—from the late 80s through the early 90s.
The Bubble Pattern
One of the most dangerous things that can happen to any business is for its category to get trendy after the company has started. When the industry starts to get cheap capital, competition heats up, and excessive capital selects for the businesses that are most willing to be too optimistic, not the ones that are too realistic. A company that's born in scrappy, capital-deprived times has to adapt to higher budgets fast, because the simplest way for competitors to catch up is to copy their business and add a bigger marketing budget, or to poach their employees.
A fascinating example of this comes from this 1999 tidbit: Warren Buffett gave a talk at a banking conference in 1999 where he noted that the aviation and auto industries had experienced rapid technological advancement without providing good returns to investors. Bezos heard the talk, and later asked Buffett for his list of failed car and plane companies. Bezos also compared 1999 to the Cambrian explosion, a period of rapid speciation and very rapid extinction.
A good manager going through a bubble might be tempted to sit the entire process out, and just let their bigger competitors burn money and flame out. But that strategy implicitly assumes that competitors will waste all of their money. A better approximation is that the competition will invest poorly, and won't get a good return on capital, but will, if only through blind luck, make a few good calls. If an industry can handle $1bn in investment and deliver a 10% return, it can also deliver a 5% return on $2bn, and overfunding implicitly lowers the hurdle rate.1
Companies going through this cycle will often raise more money than they're sure they need, and spend extra to run the same strategic plan at a faster pace. That's often a wasteful process—Amazon would have had better gross margins in the late 90s if they hadn't been worried that BarnesAndNoble.com would get to scale in the book business before Amazon got a chance to shift into being the Everything Store.
The upside to this is that it gives companies a continuous way to demonstrate to investors that the bubble is not completely overdone. Press releases are one of the key inputs to story stock valuations, and compressing the launch cycle accelerates press release production. But it also means that a company trying to survive a bubble is giving itself an unsustainable cost structure.
In total dollar terms, most of the cost of the bubble is paid a) in the late stages, when money is being raised and spent as quickly as possible, and b) after the bubble pops, but before everyone has had time to adjust. (You don't want to adjust too soon; some dot-coms got left behind during various 90s-era shakeouts (even during the staggering runup in the late 90s, Amazon had a 60% drawdown over a couple months in the first half of 1999). This also means that the best time to buy bubble survivors is when their finances look abysmal—when the growth story is ruined because they're cutting back on expenses so fast, and margins still look bad because some of those expenses are tough to get rid of without paying even more money. Amazon's sales growth was 313% in 1999 and 169% in 2000, but by Q3 2001 this had decelerated to +0.2%. And they were still losing money, with a -11% operating margin. This looked a lot like a company that had burned a few billion dollars in order to build a structurally unprofitable mail-order business—but it was actually a look at a company that had started out growing fast, accelerated its spending as a mostly defensive measure to prevent competitors from blocking it out of a future it had partly invented, and then brutally cut spending until it got to something sustainable.
The Cyclical Pattern
"Cyclical" is a broad term, and you can think of lots of working definitions, including:
Companies that have high operating leverage and no direct control over how fast demand for their products grows.
Any business where demand moves continuously and supply is added in discrete chunks, probably with a time lag.
A category that creates a persistent drag on systematic value investors' returns because cyclicals screen cheaply when they're at peak valuations and don't show up on value screens at all when they're losing money and have a negative net worth, which is often the time to buy.
One way to think about cyclicals is that the theoretical return comes from 1) potentially very high dividend payouts and rapid buybacks at the peak of the cycle, and 2) the distinct possibility that they'll go to zero before the next cyclical peak, or, if they survive, that they'll raise so much equity capital that shareholders who owned before the drop never recover. (The death-by-dilution problem is particularly acute in shipping stocks; TOP Ships has cumulatively reverse split its shares 1-for-18.9-billion since 2008, and delivered a -76% compounded return since its 2004 IPO.)
Even though cyclical industries are tough, they're survivable, and merely sticking around for a long time—and taking market share during each downturn, when competitors die out—can be a recipe for good overall returns. The oil and gas industry has had some extreme cycles before—not many industries can say they've gone through not just negative gross margins but negative prices as well—and it's also produced some fortunes. And those fortunes aren't just from rolling the dice compulsively, but from rolling the dice judiciously when the odds are good even though outcomes are uncertain. Berkshire Hathaway has a decent collection of businesses that are tied to the broad economic cycle, like Precision Castparts, the BNSF railroad, International Metalworking Companies, Clayton mobile homes, Shaw carpeting, etc. And their insurance business has its own cycle; early Berkshire annual letters often bemoan how irresponsible the rest of the industry is, and how hard it is for Berkshire to find good underwriting opportunities.
Surviving in a cyclical business can be a matter of global diversification and conservative financing, which is how the oil majors do it; they're hard to kill because the don't borrow too much and because they're involved at every stage of the energy industry, in multiple places, and sometimes a downturn in one place means an upswing somewhere else (a glut of oil is bad for Exxon's exploration and production business, but means cheaper inputs for their refinery and chemicals businesses). An even better way to survive in cyclicals is Berkshire's approach of owning a diversified portfolio of them, and being able to swoop in to make acquisitions at the bottom of one industry's cycle, while funding survival with the cash flows from other industries at different stages in their cycle. This was also part of the early formula for Blackstone: their first leveraged buyout was for US Steel's barge and railroad business, which was tied to the highly cyclical steel industry, and they closed their first fund the Thursday before the 1987 crash, which was indirectly one of the best market-timing calls ever.
The other way cyclical companies survive downturns is to use every financial engineering trick in the books to continuously restructure debt, defer principal payments, and hang on to cash. A poorly-capitalized cyclical at the bottom of the cycle is a knock-out option on the industry—it gets access to upside, but only if the company never runs out of cash and is forced into bankruptcy. If this kind of thinking sounds familiar, it's just the company's view of this famous Twitter thread on why shorting terrible companies is so painful. From the company's perspective, they aren't playing games with their balance sheet to burn short sellers. They're trying to realize the maximum value of their assets by staying in business long enough for the cycle to turn. The executives at these companies may be more optimistic than the average investor—they decided to get into steel, or homebuilding, or oil, or mattresses, when they could have done something else whose profits didn't rise and fall at 5x the change in GDP—but they're also better-informed, and may be able to detect the first signs that the cycle is turning at a time when investors are more pessimistic.
In general, single-industry cyclical stocks that barely survive a downturn do not go on to produce amazing long-term results for investors. They burn too much cash at the worst times, and while they can emerge with higher market share at the end, their industries rarely go through qualitative changes that make them less tied to the economy's ebbs and flows. If anything, the bigger a company gets the more it's sensitive to the economy, simply because it has fewer opportunities to smooth out growth by acquiring new customers once it's gotten most of the customers it's going to get.
The last great drawdown category is the hardest to analyze: some companies try to completely change the business they're in, often leaving a well-understood business for one with murkier prospects, and sometimes losing most of their market value in the bargain. IBM has had poor performance over the last decade, but to the company's credit it did survive a near-death experience in the 1990s, partly by getting out of more commoditized businesses.2 While IBM reduced its investment in the PC business, the company didn't pursue a strategy of radical decentralization, instead deciding that the key value it had was in helping other big companies centralize their IT spending with one big provider instead of dozens of smaller ones. This was basically slapping a service business layer onto what had traditionally been more of a product business, and over time that service business has seen its technology advantage dwindle. But during the 90s it turned out to be the right call. Over the course of the 1990s, including the near-collapse, IBM's total return to shareholders was 19.6% annualized.
Netflix might be the canonical case here. In 2006, they looked like a company that had figured out a nice business model, with sales growing 46% and operating margins rising from under 1% to 6.5%. It looked like a perfectly nice company that had finally hit the point where growth would pay off with high incremental margins. And so they insisted on launching a new line of business that was a) expensive, b) uncertain, and c) an existential threat to the rest of their model. That pivot went through some early wobbles, and Netflix shares were fairly flat from 2006 through early 2009. Then the stock went on a tear, rising roughly 10x in the next eighteen months, before giving up nearly all those gains in a few months as shareholders got worried that Netflix was too quick to jettison the reliable DVD business for the still-expensive streaming one. A shift from bundling DVDs and streaming to charging for them separately amounted to a 60% price hike for people who wanted both DVDs and streaming, and their US subscriber growth, which had been running at over 60% annualized, went sequentially negative in Q3 2011.
A few things happen when a company goes through this kind of event:
They get hit with employee turnover, because they've destroyed the value of recent equity compensation and raised concerns that the business won't keep growing.
This leads to worse morale, and makes it harder to hire and harder to retain—the cash cost of a given level of employee productivity ratchets up, and that's tough for margins.
They usually have near-100% shareholder turnover. Who sticks around? Nobody who liked Netflix for its DVD-by-mail business was going to be excited that all the free cash flow from that business, and more, was going to be burned creating a streaming company. But investors who liked the management might.
So in one sense, the volatility of these companies represents the volatility of betting on specific people, especially ambitious ones who can find ways to be discontented with running a high-growth, margin-accretive business when they know it could be a higher-growth business with structurally higher margins after a few bet-the-company moves. That's realistic, though; an individual who spends their entire career working "in tech" will be doing very different things over time, even if they're an individual contributor for the duration of that career. The industry changes fast, and sometimes change means the rapid economic obsolescence of previously valuable skills. As with cyclicals and bubble survivors, timing is hard, and the net winners tend to be the ones who are okay with being too early since the alternatives are either superhuman timing or being persistently too late.
You’re on the free list for the newsletter. For full access—and to read the back catalog—subscribe today!
Disclosure: long AMZN.
A Word From Our Sponsors
Tegus is the first port of call for M&A professionals and institutional investors ramping up on an industry or company.
Get access to a database of 35,000+ expert call transcripts, spanning 5+ years, or schedule expert calls through the platform for a fraction of the usual cost.
When thousands of research analysts are pooling their expert calls into an on-demand database, using Tegus is table stakes. It's the leading platform for due diligence and primary research.
See the power of a Tegus subscription, and get up to data parity with your competitors, with a two week free trial through the Diff.
Amazon revealed that their new show, The Rings of Power, got 25 million viewers in 24 hours, the first time they've disclosed viewership for their streaming service. It's interesting to consider how many streaming companies are trying to replicate, at great expense, the pre-cable network TV oligopoly. When there weren't many channels, but there was high TV viewing, one of the ways shows stayed popular was by being a topic of conversation at work the next day. In a more fragmented media environment, the success rate on this is likely to be lower, but Amazon has an astonishing number of levers they can pull: in-app promotion, on-site promotion, email marketing, standard press blitzes, ads on the packages themselves—if Amazon wants to, it can make something the most-marketed product in its category by fiat. It's useful to view anything extreme Amazon does as an experiment; if they're really able to draw that many viewers, they can make other shows the center of attention, too, making Prime a tiny bit more essential. On the other hand, the Super Bowl got an average audience of 112.3m viewers, and an estimated 208m total.
Xcel Energy, a utility, has an opt-in "smart thermostat" program: users get paid $100 to enroll and $25 annually, and they get the warm and fuzzy feeling of participating in energy conservation. They also get the literal warm feeling of being unable to crank up their AC on hot days, which they generally don't like. Rolling out a smarter grid has been a problem for a long time (here's a story from 2012, also about Xcel, also complaining that an early iteration of the smart grid system didn't live up to expectations). Like privacy online, energy conservation is something people like in the abstract but tend to really dislike when it entails inconvenience on their part.
Substitutes and Complements
AI will kill some jobs, but it's also creating some: there's now a marketplace for "prompt engineers" who can figure out commands to create specific visuals with Midjourney and DALL-E. As with lots of technologies, the substitution effect is easy to see—a computer does in seconds what a digital artist does in hours or days—but the complements are less visible. Since every job occurs in some kind of supply chain, when it gets automated there's some other job that gets much more productive. In the case of AI art, the net outcome depends on whether there's huge untapped demand for extremely specific artwork, constrained by the time artists require to make it, or whether we're all pretty satisfied with the art we have and will be happy to spend less on it.
Interestingly, the artist interviewed in this piece says their most popular work is these isometric block cities, described as "perfect for worldbuilding or games." Creating lots of art assets is one constraint on new games, both because they take time and because someone good at gameplay design is not necessarily good at digital art and animation. So one sector that might see growth is niche games, or mods of existing games. Some genres are limited by the need to invent new mechanics, but for other kinds of games there's a basic set of mechanics (e.g. "aim mouse, click to shoot") and almost everything else is map design and aesthetics.
Media incentives are shaped by whatever the best-monetizing ads they'll accept are, and in sports that increasingly means close partnerships between sports news companies and sports betting. Some of this is from chasing market share. Things aren't as frenetic as earlier this year, when you could theoretically pay for a New York vacation entirely by signing up for gambling promos while you were there, but the lifetime value of a gambler can be quite high, and people who are avidly following sports news are pre-qualified to be interested.
What sometimes emerges is a situation where most of the users are indifferent to a particular kind of advertiser, but when those users are weighted by the CPMs advertisers pay to reach them, the same advertisers become the main revenue source. This often happens in crypto, where signing up new traders (or Ponzi scheme victims) is lucrative, and in finance generally, where various kinds of asset-gathering dominate ads. And, over time, it affects editorial incentives, in a way that's not healthy for the business long-term but very hard to say no to in the short term.
Apple is hiring for 216 advertising-related roles, compared to 250 people who work on its ad platform already ($, FT) (though "Apple disputed the figures but declined to elaborate"). Apple's position in the ad industry is unique right now; they've wounded or crippled many large competitors, freeing up substantial budgets that will eventually go somewhere. And they've given themselves a privileged position to use data across apps for ad targeting. Even if Apple doesn't want to replace every dollar Facebook earns with a dollar of Apple ad revenue, they certainly have a runway to build up a substantial business. But the growth they get comes at a reputational cost, both among developers who are disappointed to see their ad revenue decimated and among prospective partners who will worry that their Apple-dependent margin will eventually be Apple's opportunity.
Diff Jobs is our service matching readers to job opportunities in the Diff network. We work with a range of companies, mostly venture-funded, with an emphasis on software and fintech but with breadth beyond that.
If you're interested in pursuing a role, please reach out—if there's a potential match, we start with an introductory call to see if we have a good fit, and then a more in-depth discussion of what you've worked on. (Depending on the role, this can focus on work or side projects.) Diff Jobs is free for job applicants. Some of our current open roles:
A successful systematic crypto fund is looking for engineers and quant researchers with Python or C++ experience (ideally both) and an interest in crypto market-making. (Remote)
A firm with a unique business model that gives them access to equity in promising startups is looking for a generalist to help scale their business (US, in-person, several locations in major cities).
A startup that helps companies' financial systems seamlessly communicate is looking for a mid/senior-level engineer, C#/.NET experience ideal. (Austin)
A startup using AI to tackle a hard problem in the dev tooling area is looking for multiple engineering and ML roles. (Washington, DC area)
A fintech startup is seeking people with deep experience in insurance, particularly life insurance. (US, Remote)
We're always interested in talking to people with engineering and product backgrounds who are thinking about their next role.
The exact form that this takes will vary. Usually, there are different elasticities for different kinds of investments, so you'll see a small impact on some things they invest in and a nonlinear increase in compensation for experienced people, and in the valuations that strategic-looking businesses get. Some of the best IRRs in venture come from funding a challenger business that panicking incumbents will buy at a steep valuation—especially when the size of the acquisition is used as a proxy for how seriously they're taking the threat. I have heard of at least one acquisition where the buyer negotiated the price up so they could give shareholders a nice round number for the size of their deal.
To IBM's discredit, the PC business was commoditized in part because of IBM's own decisions.