Abraham Thomas introduces his new newsletter, Pivotal, with a great piece on the economics of data businesses. This is an area I've worked in as a buyer and seller, and everything he says rings true. Especially the point that data businesses start slow but accelerate over time. For some businesses the sell is easiest with the first few customers, for whom it's life-changing, and then slows down. But a data business has an accelerated route through the middle of the S-curve, when prospective customers realize they're falling behind current customers, while the product itself gets more useful as its hit rate—things it can do divided by things it sounds like it might be able to do—rises.
John Luttig reminds us that Microsoft is a very impressive company, especially its ability to scale in cloud with a model different from Amazon's. (Disclosure: I own shares of both.) Among other notable points: "Is Azure the fastest B2B product of all time to reach $10 billion of revenue?" At one level that's a function of the company's economies of scale, but at another level it's a testament to Microsoft's ability to survive diseconomies of scale.
Eric Seufert of Mobile Dev Memo asks If personalized advertising is banned, who bears the cost? Targeted ads subsidize free products, and given the scale of the ad-supported business, those products are clearly very popular. The political bet may be that if Instagram, Twitter, and TikTok all become worse products because making them better doesn't pay as well as it used to, the companies will be blamed. Which is its own kind of deadweight loss.
Was the Android tablet a missed opportunity to rethink computing? Building new interfaces people will love is hard. There are transitions, but they're slow, in part because they require a change in the mental models of users and developers. This piece looks at the "Bumptop" interface, which was more skeuomorphic than other tablet interaction models, but which never quite took off—though that may still happen.
A very good interview by VC Matt Turck with Richard Craib of Numerai. Numerai runs a distributed systematic fund; data scientists create signals from data Numerai provides, but the data is stripped of all identifying information. They don't know if they're predicting the correlation between sequential changes in the relative prices of two stocks over a period of years or measuring the lag between rainfall changes and cocoa prices by the minute. There are some systematic investors who do this reality-agnostic pattern-matching view, which can help them find patterns that either don't make sense or make sense for an obscure reason. Other systematic investors will try to use strategies if there's a reason they ought to work, which restricts the range of strategies they can employ but also keeps them safe from betting that a trend will persist even when the reason for it changes. Impressively, Numerai only lost 1.5% in March of 2020, a great performance given how many trendlines and market relationships broke during that time.
Books
Xbox: The Making of a Bad-Ass Machine: A quick read on the history of Microsoft's games business, which has gotten a lot more complicated with the pending Activision acquisition.
The Power Law: Sebastian Mallaby previously wrote More Money Than God, an excellent history of the hedge fund industry, and he's back with The Power Law, which covers venture capital. This book does a great job of starting with a theoretical framework for why VC is worth understanding (it's a third way of organizing large-scale efforts, besides firms and markets), and then delivers many, many interesting stories. Especially good: the book has a look at the early days of Tiger Global and DST's moves into late-stage growth capital, including interviews with people who don't usually talk to the media.
Open Thread
Drop in any links of thoughts of interest to Diff readers.
In advance of next week's series on the video game industry, and what other industries can learn from it, are there any games with unique new business models that are still under-the-radar?