Sharing and Owning Standards

Plus! Institutional Bitcoin; Negative Prices; Facial Recognition; Recursive Product Development; Every Tech Company Becomes a Bank; Robinhood, Underwriter?

Byrne Hobart Byrne Hobart

Welcome back to The Diff. Here are the subscribers-only posts you missed this week:

To read these, and to get the full Diff experience, subscribe today.

Subscribe now

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 17,639 subscribers, up 143 week-over-week.

In this issue:

Sharing and Owning Standards

The only way to get a lot done is to spend essentially no time on unnecessary decisions. Some people manage this in their life—delegating the boring stuff, keeping a precise schedule, wearing and eating the same thing all the time, ruthlessly turning down interesting distractions. And companies do it, too, often through official or implicit standards.

Standards are an interesting economic good. They’re a public good, because by definition everyone has to know the rules in order to follow them. But they’re often provided by the private sector; governments can be a coordinating point for imposing a standard, but the rules themselves come from private actors.

An interesting and relatively early example comes from the auto industry. In the early 1900s, the car industry was small, and fragmented. Cars were expensive, bespoke, and mostly made for rich hobbyists. Even small industries can go through boom and bust cycles, and one early cycle was driven by the classic bugaboo of small businesses: supplier concentration and working capital risk. Since every car company used unique designs, they ordered unique parts. But cars are full of essential parts: if you have everything you need to build a car, except the axles, or the spark plug, or a particular screw, you have lots of expensive inventory on the books and no cars to sell.

So, in 1905 a group of car company representatives formed the Society of Automotive Engineers. Their goal was to agree on the dimensions and quality of the basics: tires, screw threads, lock washers, etc. In many cases, they didn’t narrow things down to one model, they just picked the most popular 20% or so of designs and made those the acceptable ones. This improved the economics for both the carmakers themselves and their suppliers: car companies could expect backup suppliers if their main provider failed, and suppliers had backup customers, too. Even though the largest car companies didn’t participate in early standardization efforts, overall adoption of the standards exceeded 90% within a decade.

By 1921, a trade journal estimated that standardized parts were saving the industry $750m a year, or 15% of the entire value of car sales. The industry didn’t need a top-down mandate to enforce this—even within the auto industry, the change was driven by smaller companies, not top-down. All it really needed was an industrial cluster and a shift in inputs—parts shortages led to less work for car workers, which meant a surplus of free time that they could use to hash out disagreements over carburetor lever arms and the like.

Standardization is not always an unambiguous win. In the early days of the credit card industry, Visa, Mastercard, and American Express all figured out that they’d be better off with a default set of card dimensions, point of sale specs, and the like; better to make manufacturing a commodity business so they could focus on the pure-play network-effects payment product. But this meant that when Sears launched the Discover Card in 1985, there were compatible card readers and card manufacturers, and consumer habits were already ingrained. Within two years of launch, they were the third largest credit card brand in the US, with 22m accounts. IBM famously built its PCs with commodity components, and let Microsoft license a variant of the most popular IBM PC operating system to other manufacturers. In short order, competitors who didn’t have IBM’s R&D or administrative overhead were stamping out much more affordable clones.

Creating a standard seems to be most effective as a way to commoditize the complement. A company can make one part of its product an open standard, because it has a competitive advantage in benefiting from more supply of whatever product it commoditizes. A fun example of this comes from the world of tabletop role-playing games. In 2000, Dungeons & Dragons released a generic role-playing system that could be grafted on to other themes, like superheroes, spies, and Star Wars. One paper shows that the number of new tabletop gaming companies rose from 20 in 1998-99 (before the introduction of D20) to 68 from 2000-1, after, while employees per game company dropped by half.

So, Dungeons & Dragons' publisher had many new competitors. But that also meant that many more people had 20-sided dice and knew the basic mechanics of games. Since playing a game means matching a group of players who are all interested in the same game, an increase in the number of games published, and the number of unique players, can be a net benefit for the biggest incumbent: among any random set of d20 players, the system they’re all most likely to be familiar with is D&D; there isn’t data available, but it’s possible that this open system both reduced D&D’s share of total players and increased its share of total tabletop games played, by adding so many nodes to the network. (This same dynamic played out in credit cards: Discover was able to grow fast, because Sears was a major merchant at the time, bootstrapping the network effect. But eventually, the much larger Visa and Mastercard networks won out, and Discover’s importance waned.)

Some standards persist even though they’re imperfect, because they give every better product something to measure itself against. FICO scores are imperfect, but various attempts to replace them haven’t caught on. Since it’s hard to have a full-time job that consists of 1) checking FICO scores and then 2) approving every loan above a threshold, the FICO standard is sticky because every credit analyst can benchmark default rates to them. A similar dynamic holds in bond ratings: BBB persists as a rating because credit managers want to show that they earned BB-level returns while taking single-A risk.

A standard doesn’t have to be perfect, or even consciously designed, to keep working. In fact, they’re more effective when they canonize the market share leader as the only way to do things. This means they often echo much older technological decisions: programmers generally use up to 80 characters per line, because early computer displays were exactly 80 characters wide, and that was because IBM’s punched cards were 80 columns.

A standard is suspiciously similar to a software product: it consists entirely of information, and can be produced at nearly zero marginal cost. In fact, the category of software it’s most similar to is the API. There are open standards for building web pages that any browser can read (HTML, Javascript) and there are closed standards for turning that page into a page that can send a text message (Twilio), collect a payment (Stripe), force users to prove that they’re human (reCAPTCHA), match free-form text input to specific entities (Bing Entity Search), etc.

In one sense, an API is just a privately-held standard: it has the same net economic benefits, but the standard creator earns a royalty on their use, rather than giving them away. This has some complicated effects:

  1. The first-order impact is that it leads to more standards for more products. Public goods tend to get underproduced, but if there’s an entity that captures the upside, it has an incentive to create them. That’s especially true for anything that has a low-but-nonzero cost; running it as a public service is prohibitively expensive, but when it’s charged for it’s a trivial cost.
  2. Over the long term, though, it can discourage certain kinds of users.

When car companies standardized, they deliberately avoided standards for the parts of cars that were changing the fastest. If there’s a default way to build an engine, that’s the end of innovation in engines; better for each unit produced to be a bit more expensive and for the state of the art to keep on advancing. But revenue-generating APIs are also information-generating APIs; when Twilio sees a great service built on Twilio, they know they’re getting revenue, but they also know that they could capture more revenue by turning that service into a feature.

An API company is a standards-producing body with a scheming brain attached. Standards bodies, of course, are run by human beings, but they have a definite lifespan, defined by the job tenure of the people running them. One of the most interesting tidbits from The Box (also a story about standardizing!) was that shipping companies were able to get unions on board with labor-saving technologies because the shipping company is immortal, while union members don’t have a strong economic interest in what happens to their job after they retire. So a company with a low discount rate can buy cooperation from a person with a higher discount rate.

Some companies that rely on APIs worry that they’re just doing R&D for their biggest supplier—this is a common fear about AWS, for example. So it’s in the interest of API providers to define what businesses they don’t want to be in.

There has been immense value in standardization for a very long time, and it’s easy for even smart people to miss: Charles Kindleberger once pointed out that in The Wealth of Nations, Adam Smith doesn’t understand why a merchant shipping between two countries in Europe would have both shipments stop by the port of Amsterdam on the way to their destination. Smith attributed this to nervous merchants who didn’t want goods out of their sight for long, but Kindleberger notes that one of Amsterdam’s major industries was the measuring and grading of commodities. 17th and 18th century Amsterdam was the ancient ancestor to modern pay-per-API-call businesses: a network effects-driven economic actor that specialized in taking lumpy and complicated inputs and outputs, and converting them into a form that everyone could understand and use.

Further reading: Standards as Public, Collective and Private Goods and Intercompany Technical Standardization in the Early American Automobile Industry were both sources for many examples above. Paying With Plastic is broader, but has many interesting examples.

Subscribe now

A Word From Our Sponsors

As economic activity comes online, new data trails are left behind. Over 200 innovative investment firms leverage this data for alpha-generating insights. Thinknum’s alternative datasets power investment decisions across more than 500,000 public and private companies around the world.

For example, track job listings to identify which companies are expanding in real time. Or track the individual tenants by commercial mall to analyze churn on a daily basis.

Interested? Come learn about all of our 35 datasets here.

Reader Request

I’ve noticed a spate of companies in the logistics and trucking sector that have raised rounds. Any Diff readers who work in the space and know what’s changing—why now?—are invited to reach out. Just reply to this email.

Elsewhere

I went on the Conservative Curious podcast to talk about legitimacy and institutions. And I’m in Palladium with a piece on why the US dollar will remain a reserve currency for a long time.

More Institutional Bitcoin

A few months ago, when Paul Tudor Jones announced that he owned Bitcoin, I wrote that the real meaning of this was not that a relatively fast-money investor was buying, but that institutions with longer-term outlooks were more likely to buy in the future. MicroStrategy and Square followed up with Bitcoin purchases of their own, both buying with the intent to hold, and now MassMutual owns Bitcoin, too ($, WSJ). This position is a tiny piece of their portfolio, just 4 basis points out of $235bn in assets. But it’s another step towards Bitcoin being treated as an alternative reserve asset.

(Disclosure: I am long Bitcoin.)

The Latest Asset to Trade at Negative Prices

Ubisoft is paying players $10 to try one of its free-to-play games.(It’s a $10 credit for any other Ubisoft game, not $10 outright.) Usually, bundling means getting X+Y for less than the price of X plus the price of Y, not getting X + Y for less than the cost of X. But in this context, it makes sense: free-to-play economics are driven by people hanging out in-game and buying cosmetic goods, so Ubisoft is encouraging people to migrate to the game so that once they’ve homesteaded a bit, they can be taxed.

Facial Recognition

The governor of Massachusetts is refusing to sign a bill that would ban facial recognition, arguing that it’s helped catch murderers and sex offenders. The technology is extremely controversial, but at this point inevitable; the relentless drop in hardware prices, and the high conversion rate from publishing interesting ML papers to fielding multiple mid-six figure offers from major tech companies, ensures that the technology will be widely available no matter what. It’s already being implemented in other parts of the world, so policymakers outside of China can’t determine whether or not it will be used, just whose interests it will be used in.

Recursive Product Development

DoNotPay, which provides easy access to legal services like disputing parking tickets, getting refunds, and collecting security deposits, now offers freedom of information act requests as a service. FOIA requests are a popular tool for investigative journalism, and they also have entertainment value. But what’s especially interesting about this launch is the backstory:

Browder said that DoNotPay “would not exist” without FOIA laws. “When we got started appealing parking tickets, we used previous requests to see the top reasons why parking tickets were dismissed,” he said. Browder said he’s hoping the feature will help consumers uncover more injustices — just like with parking tickets — to feed his product with more features. “The overall strategy is to use any interesting FOIA data to build great new DoNotPay products,” he said.

It’s commoditizing the complement all the way down.

Every Tech Company Becomes a Bank

Back in June, I noted that the threshold for venture-fundable niche startups was lower than I’d thought, because Squire, a company that sells software for barbershops, raised $34m. They’ve now raised another $45m, at triple the valuation. One reason is that they’ve moved from pure software to financial services:

The coronavirus has threatened the livelihoods of small and medium-sized business owners, making it harder for them to secure loans or financing to undergo tricky times. Salvant says that Squire took on $15 million in debt financing to create a banking-as-a-service feature for these business owners.

This is an interesting variant on the general idea that more debt-based financing is coming to software companies. These companies are also a good conduit for credit: they get real-time data on how businesses are performing, and they have direct relationships with clients. Adding a layer of banking on top of their product is a good way to lock in customers, and to arbitrage the gap between cheap capital for startups and expensive capital for low-growth small businesses.

Can Anyone Afford for Robinhood Not to Underwrite IPOs?

The function of an IPO underwriter is to work with a company, and with future investors, to figure out what the company will be worth when it’s publicly traded. Normally, that involves talking to major institutional investors, figuring out what the demand curve looks like, and baking in a little caution (either to ensure that the IPO goes off without a hitch or to do a very lucrative favor for clients). This is something banks can do when there’s a finite number of buyers, but it’s much harder with retail investors. Retail buyers are less valuation-sensitive than institutions, and they use different platforms.

Which is why the two biggest IPO pops recently have been fueled by retail investor buying. Underwriters say that retail investor demand is hard to predict, but I suspect that someone at Robinhood has already found the correlation between a) the size of an IPO price pop, and b) the number of Robinhood users who look for a stock quote for the to-be-public company. Because Robinhood users are doing more of the research on the app itself, Robinhood may know more about their demand than a bank knows about institutions' plans.

Retail investors have always been a swing factor in IPOs. When Netscape went public, Charles Schwab had to change their 800 number’s recording (“Welcome to Charles Schwab. If you’re interested in the Netscape IPO, press one…”). But now they’re a more measurable one.