Inventing Demand

Plus! BNPL; Coopetition; Marketplaces and Ads; Last Mile, Last Users; Value and Rates; Diff Jobs

Programming note: this week I'll be coworking with my kids, who are six, four, and three years old. The posting schedule may be impacted.

Inventing Demand

In 1962, NASA had plenty of here-and-now problems, but, being an unusually effective organization, it found time to think about hypothetical future problems, too. For example: they had found a supplier for a particular component of their navigation system, and while that supplier made exactly what they needed—and, in fact, was the only company in the world that could make these components with sufficiently low weight and high reliability for spaceflight—it was a fairly new company, and a kind of faddy product. What if NASA built a computer, and found that after a few years it couldn't be replaced because the supplier had left the market or gone out of business entirely?

They hit on a fairly simple solution: they bought more than they needed, and used some of these tools for ground-based systems where they had wildly better specs than necessary, but where steady demand would keep their supplier in business. This bet worked out; the supplier was Fairchild Semiconductor, and the components were integrated circuits. Within a few years, they were well on their path from being so expensive and specialized that only Cold War paranoia could justify buying them in quantity to being ubiquitous.1

NASA's thought process is not an uncommon one for forward-thinking organizations, and it's becoming an important one for the entire economy. One of the broad problems the world economy has run into repeatedly over the last two and a half decades is that, in the aggregate and especially in specific industries, there's either a supply glut that, in combination with mildly protectionist policies, holds back global demand, or there's a persistent capacity shortage because everyone who can make capital expenditures is convinced that demand will evaporate by the time their new capacity comes on line.

The first of those scenarios applies to global manufacturing of low value-added products: electronics assembly, apparel, furniture, etc. China's decades-long investment frenzy made those product categories astoundingly cheap in historical terms—since 1998, overall prices in the US are up 81% but clothing prices are actually down 5%. But this also meant that for other countries to compete with China's labor market, they had to catch up to China's infrastructure, a task that gets more daunting every time the Chinese economy slows down and policymakers respond with another round of infrastructure investment. (Such as, to choose an example at random, right now).

The second scenario more commonly applies to US-based companies: the energy sector has been chastened by poor returns in the last decade, and has a very high threshold for making new investments. EOG, for example, raised its after-tax hurdle rate from 30% to 60%, and that's after assuming $40 oil and $2.50 natural gas. Big airlines tend to get punished when they try to grow capacity faster than GDP, even though travel takes an increasing share of GDP as economies grow.2 More recently, retailers like Target and Walmart have lost most of their Covid-era market value gains thanks to excess inventory from overestimating demand.

There are few easy answers at the aggregate level. Governments can run hotter and colder fiscal policy to balance demand, but they're always responding to data on a lag. And stimulative policies can't easily differentiate between encouraging investment to lead to stable overall growth and encouraging malinvestment that creates an ever-widening spread between reported GDP and actual wealth creation.

But at a more limited level, there are some promising examples of demand creation that do seem to work, at least if the goal is specifically to encourage supply today by creating credible demand tomorrow.

One way to look at this is that estimated demand is not just a specific point, but a statistical distribution. If the cost of a missed opportunity is unknown, while the cost of making an investment that doesn't pay off is obvious, then risk-averse companies may choose not to make the investment—especially because they can still achieve growth on a per-share basis by buying back stock (an even easier prospect when their cash flows are predictable because investment is light and the business they're sticking with are predictable).

Companies may conclude that undersupplying has a more favorable risk/reward, especially in an environment where fixed costs are rising. If every power plant requires a longer and more costly environmental review than the last one, even if it's technically unchanged, then the power plants with the highest returns on equity will be the ones that already exist, and a company that builds new ones with the same operating economics but higher and more uncertain fixed costs will be diluting their business. Airlines may face something similar, indirectly because of environmental reviews but partly because of the chunkiness of airport capacity: if an airport has two runways, then whether demand could saturate 100% of that or 149.9% of that, they won't expand. Airport departure slots can go for $30m+ each, and that cost rises nonlinearly as an airport gets more crowded. In healthcare, the generally rising cost of FDA approvals might also hold back investment in a similar way: if a drug required $1bn to get approval when it was first created, but would need $2bn today, the ROI on the older cohort is higher and the new ones have to hit a higher certainty threshold before they're worth pursuing.

Not every category is amenable to advance purchase commitments. Barrels of oil are fungible, and haven't changed in a long time. Cars aren't, and a commitment today to buy cars in 2030 would be an ambiguous one indeed. Some products, like vaccines, are indeterminate here: they vary in quality, so if purchase prices were consistent across different vaccines, OWS was relatively overpaying for the worst ones. But for a product with a measurable success rate, it's possible to create a formulaic plan that still differentiates between different success levels, like paying for a fixed number of doses at a price of $x multiplied by efficacy rate, where $x is naturally a premium to what the simpler formula would call for.

Purchase commitments can work especially well in two instances: first, any product with a steep experience curve gets cheaper or better as more of it is made, so those commitments amount to subsidizing future price cuts. Second, for a commodity product with volatile pricing, some of the price volatility is driven by the fact that no one wants to create new supply when demand is at a low ebb. By truncating the left tail of the outcome curve, this raises the risk-adjusted return on more supply. One place where these are combined is in the lithium market, which is a key input into batteries (a business with a very nice-looking experience curve) and a history of booms and busts (lithium carbonate prices declined 78% from late 2017 through early 2020, and are up 1,180% since then). That market currently has some other supply issues that make it harder to step up production ($, WSJ), but more certain future demand would mean more places where projects are viable.

One of the forces that puts stress on efficient markets is that volatility can wipe out skilled investors, even if their thesis is ultimately correct. That's avoidable for people buying financial assets, who can eschew leverage, but operating leverage is inescapable for companies that actually produce something, and the amount of operating leverage inherent in a business is only really apparent when things get bad. One function governments have in financialized economies is to be the volatility-absorber of last resort, keeping a firm grip on the financial domino whose fall would be catastrophic. And, increasingly, they're doing this through interventions in physical products rather than just financial markets.

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.

Elsewhere

BNPL

Mastercard recently launched a buy-now-pay-later option, and some retailers are complaining that they've been automatically opted-in. The main complaint is that the fee, while lower than typical BNPL provider fees, is higher than traditional credit cards. In other cases, retailers had exclusivity agreements with payment providers.

Mastercard is in a tricky situation. For any new lending product, the only way to figure out how to make money is to lose some first, by making loans and seeing which ones go bad. Adverse selection is ubiquitous in lending—the person most excited to borrow from you is someone who has been rejected with good reason by every existing lender. So the priority, for companies that can afford it, is to get some scale first and then tighten up lending standards later. That's especially important for a company like Mastercard, which views the BNPL category as a major threat. If their customer base adopts a homegrown solution, it will be harder for Affirm and Afterpay to make inroads selling the same thing.

For an earlier Diff look at BNPL, see this piece on Affirm from May.

Coopetition

Before Apple was shaking up the online ad business to the detriment of Meta and others, they were looking at ways to partner with Meta instead ($, WSJ). It's generally a good idea to assume this kind of dynamic for any relationship between big tech companies, whether it's cooperation or adversarial. It's safe to imagine, for example, that Apple is always running the numbers on creating a search engine (or partnering with Bing) when it renegotiates its Google deal, or that Disney's board saw at least a few Powerpoint presentations on the relative merits of competing with, partnering with, or buying Netflix on the way to deciding to launch Disney+.

One revealing note from the article is: "while Facebook’s products were among the most popular apps on the iPhone, they didn’t generate sales for Apple. This was a persistent frustration for some Apple executives, according to the people familiar with the matter." A lot of the meaning here hinges on whether the "people familiar with the matter" were on the Apple side or the Meta side; Apple certainly has made efforts to capture more of the wealth created by app usage, but it's natural to impute more of an adversarial angle to a company's behavior when it's a direct threat to your own company's profits. So what might have been a genuine offer for a partnership from Apple could have looked more like a threat to Meta. Naturally, that offer was partly a way for Apple to capture revenue it was helping to create but not getting a cut of—but, at least initially, the goal might have been to do this by adding upside and taking a cut, rather than increasing Apple’s share of a fixed amount of revenue.

Marketplaces and Ads

One useful model of the app store ecosystem is that it can thrive with both subscription and ad-supported models, but, as in the story above, it's harder to capture most of the value created through ads, while charging a cut of subscriptions is fairly easy. The ad-supported products can be worth keeping around, because they're broadly complementary (people buy iPhones to use apps like Instagram) and more narrowly so (some of the best ad ROIs are achieved by running ads in free apps for games with paid features).

YouTube is working on a "channel store" that will sell access to paid streaming services ($, WSJ). This is partly a hedge for YouTube, since there are so many companies with vast content libraries that have started to offer their own streaming apps. If Spongebob and Succession are locked behind somebody else's paywall, it's better to get a cut. But this is also strategically very helpful for YouTube, since it means that more streaming consumption decisions are mediated by YouTube's app, and in this case, the complementary ad-supported viewing is viewing that YouTube will get a cut of.

(The Diff previously covered YouTube's model in more detail in this subscribers-only post ($).)

Last Mile, Last Users

Some Facebook Marketplace orders will soon be fulfilled by DoorDash ($, WSJ). Part of what DoorDash has done well is using meal delivery to build a high-utilization last-mile logistics network for rapid purchases; like Amazon, they're big users of infrastructure that they also resell to others. This has been the plan for a long time. And, like Amazon, DoorDash's economics get better—especially relative to the competition—as their network gets close to full capacity. One interesting side effect of this is that for DoorDash, it's likely that the deals with the biggest incremental profit margins won't be the large ones like Facebook Marketplace, but smaller ones that add a few extra high-margin trips without contributing much at all to costs.

Disclosure: Long AMZN.

Value and Rates

A common stylized fact is that growth stocks are an interest rate bet: since more of their cash flows are in the distant future, they're more sensitive to how future cash flows are capitalized. This makes intuitive sense but historically, isn't especially true, and there are good theoretical reasons for this. This post breaks down the argument by looking at the rolling correlation between ten-year rates and the relative performance of expensive stocks (weak, but trending up for the last decade) and looks at the average empirically observed difference in earnings growth rates to find that the theoretical duration of value and growth cash flows are immaterially distant.

Which is true, as far as it goes, but there are a few recent counterarguments that make the growth-as-a-rates-bet argument stronger, in line with empirical observations:

Importantly, none of these are especially strong arguments that growth as a category will outperform: investors could be overestimating their ability to predict long-term growth3; growth sectors have plenty of casualties, driven more by technology and product cycles than the economic cycle; and a financial sector that collectively overweights investment in lottery tickets instead of adjusting to lower expected returns is certainly making a suboptimal choice. So the recent correlation between rates and growth performance might not be a good sign, but it's at least understandable.

Diff Jobs

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:


  1. This anecdote comes from the wonderful book Digital Apollo, a detailed history of the Apollo guidance computer but also an extended meditation on when and how human judgment can be automated—and how to get the humans involved comfortable with handing control over to a machine.

  2. How do we square capacity restraint with the fact that airline tickets got relatively cheaper throughout the 2010s? One reason was that oil was getting cheaper for most of this period, and another is that, thanks to more effective price discrimination, load factors—the percentage of seat-miles filled by paying passengers—went up from the low 70s two decades ago to almost 85% in 2019.

  3. Which could be partly a matter of career preservation. If someone's thesis revolves around what happens five or ten years in the future, it's hard to say they were wrong after a quarter or two. Maybe the market was wrong!