The Value of Time Rounds to Either Zero or Infinity
"Tech-Media-Telecom" has always been a somewhat arbitrary investing sector; it seems to have gotten popular in the early 2000s, when "TMT Analyst" was a polite term for "Person hired as an Internet analyst, who turned out to be too talented to lay off after the crash." TMT arose because the Internet sector was briefly huge—289 Internet companies went public in 1999 alone—but then collapsed into 1) a collection of mostly walking-dead micro-cap stocks, and 2) a handful of dot-com survivors, which were interesting companies but were too short a list to take up 100% of a full-time analyst's attention.
The sector has since grown into a more accurate and useful definition: it covers any company that mostly manipulates bits, or helps other companies do the same. And one of the interesting things about the bit business is that its biggest complement at the point of consumption, across all domains, is time. Some media companies literally monetize time—Nielsen ratings for TV ads, subscriber-years for affiliate fees. At the other end of the spectrum, some of them make money by saving time, particularly by saving the very valuable time of software engineers. Twilio's enterprise value approximates the net present value of the cost of all the engineer-hours it would take to rebuild the relevant subset of their features in-house for every single user.
Across the information supply chain, the value of time either rounds up to approximately infinite or rounds down to roughly zero. And sometimes, this variance happens within a single app. When Facebook sends a push notification informing you of a new interaction, they're treating your time as priceless—don't be even a second late for this notification! But once you've surmounted the barrier of actually opening the app, future interactions implicitly assume that your goal is to fill as much time as possible. Infinite scroll and continuous queuing of new content are hallmarks of a zero-value-time environment.
This model is not completely sinister. Some chunks of time are, in fact, worthless. Twitter seems purpose-built to fill the increment of time from waiting for an elevator; long enough to be a noticeable delay, short enough and variable enough that 280-character chunks are the right increment of entertainment. They do get dangerous, though, because apps consistently optimize for higher engagement; the implicit goal of a Twitter PM is that you open your phone when you hit the button to summon an elevator, and you're too engrossed to notice when it arrives.
The generalized approach that most zero-time-value products choose is to take the zero value of time very seriously indeed: have an unlimited supply of content, and constantly update the primary user interface to provide either a series of easy decisions—a lot of data and brainpower goes into Netflix's presentation of shows you might like—or a completely passive experience. One innovation in the "stories" form factor was ephemerality, another was that you can keep consuming content without even using your thumbs.
At the other end of the spectrum are the products that assume the user's time is incredibly valuable. Bloomberg may be the canonical example, one I've written about at length before. Bloomberg is fractally good at reducing friction and increasing response time:
- It relies on terse shortcuts (and financial assets already have abbreviated names, dating back to when telegraphs were bandwidth-constrained). As a result, Bloomberg commands tend to have an information density only approached by Perl one-liners.
- It has APIs, so any repeat action can be automated.
- Bloomberg's help desk is extraordinarily helpful. If there is a way to use the terminal to do what you're trying to do, they'll tell you; if there isn't, your request is very likely to end up on an internal list of upcoming features.
- Bloomberg extracts bond quotes from messages, converting a mostly illegible medium—text chat is descended from phone calls, not letters—into a structured one.
- The company is always adding adjacent features to make the core product more useful. I don't have access to a terminal today, but my guess is that the SPLC function, which maps out supply chains on a company (example here) created a great deal of alpha throughout the pandemic, the semiconductor shortage, and the more recent plastics shortage.
Bloomberg is constantly trying to save users small increments of time. When you consider the fact that Bloomberg's customers are banks and hedge funds, and its users are, by far, the biggest operating cost for those companies, the economics really start to work. Every year, the cost of not-Bloomberg rises for its customers, which does wonders for Bloomberg's retention.
Superhuman tries to do something similar with email. It sounds overpriced, unless you compare it to the alternative: consider someone who earns mid-six figures, has variable compensation tied to their job performance, and spends half of their on email. $300/year is the wrong price point to look at; if it's worth learning the Superhuman workflow and keyboard shortcuts, the product needs to make users visibly more efficient. Say, 5% faster at email. And to their target market, that's worth thousands of dollars a year.1
The relationship between the mind-machine synthesis of an experienced Bloomberg user, or an Excel expert, and the revenues from selling these products is strong, but it's indirect. It's something that companies optimize for if they're somewhat ideological about it—if they insist that saving users time will produce benefits later on. Other companies are not quite so on the ball; since enterprise software is more likely to get chosen top-down, it doesn't prioritize rapid interaction the way consumer-facing products do. And since it's often adopted company-wide, users just have to live with the idiosyncrasies. This creates an interesting dynamic; it's actually okay for business software to have an opaque, hard-to-discover user interface upfront, especially if it trades that off against much faster use once the user is experienced. In the long run, this kind of software prospers by having low gross churn, and by increasing spend per customer over time, both of which are maximized when the product gets more indispensable as it gets used more, and when it moves roughly as fast as users can think.
There is one notable exception to the general rule of treating time as either priceless or worthless: Y Combinator. As an investor, the company implicitly treats time as insanely valuable, by giving companies capital and then starting the countdown to demo day. But at the top of the YC funnel is Hacker News, a site that is so good for procrastination that they literally implemented a procrastination-proofing feature, years before screen time fears and social media addiction were salient topics.2 HN is good for quick breaks, but it can consume lots of time. Noprocrast is a nod towards the idea that someone who likes Hacker News has better things to do than spend all of their time on Hacker News, but it's still a hard balance to strike.
I've speculated before that a good generic investment strategy in software is to always back the company whose products have the lowest latency, broadly defined. There's a lot of clunky software that people use because they haven't gotten around to a better version, but there are some products that are basically impossible to stop using because the gap between thinking something and getting the program to do it is so short for experienced users. If someone has used Vim, Emacs, or Excel for long enough, it would be worse for their productivity to amputate a pinky than to switch to a product with different keyboard shortcuts.
This is a potential opportunity for anyone who is looking for lazy legacy companies to go after. Try their demo, and if the product takes a long time to load, if tasks take five clicks when they could take two, if there aren't keyboard shortcuts for every possible action—there's blood in the water.
A Word From Our Sponsors
Here's a dirty secret: part of equity research consists of being one of the world's best-paid data-entry professionals. It's a pain—and a rite of passage—to build a financial model by painstakingly transcribing information from 10-Qs, 10-Ks, presentations, and transcripts. Or, at least, it was: Daloopa uses machine learning and human validation to automatically parse financial statements and other disclosures, creating a continuously-updated, detailed, and accurate model.
If you've ever fired up Excel at 8pm and realized you'll be doing ctrl-c alt-tab alt-e-es-v until well past midnight, you owe it to yourself to check this out.
Limits to Hardware-as-a-Service
Last year I wrote about the model of bundling hardware with software to escape the poor economics of pure hardware businesses. Cricut, which is in the process of going public, sells machines that help create custom crafts. (This video has some examples.) A few days ago, the company decided to go all-in on the as-a-service model, by requiring device users to sign up for a subscription service if they used the product more than a minimal amount. This led to calls for boycotts, a petition and many, many angry customers. Comments per day on the Cricut subreddit went from under a hundred to over a thousand.
And, in response, the company walked it back. They may try again. A company in mid-IPO can't afford a boycott, whereas a company that has gone public (and has made it past the lockup date) might be more willing to take the risk. But this illustrates why the transition from hardware to services is difficult: it's hard to figure out what to charge for, and if the company can't justify their pricing model, consumers will notice. Better to start with the hardware/software bundle in mind than to try to create a valuable bundle after making an appealing hardware product. For Cricut, it's also a bit of a reach: the company has a high-margin supplies business in addition to its devices, so it's capturing usage-based revenue already. There are limits to how much more they can get before customers rebel.
See also, my subscribers-only writeup of the Cricut S-1 ($).
Shareholder Activism in Japan
The Economist highlights a surprising trend: Japanese companies have increasingly independent boards, and are more willing to consider hostile takeovers ($):
Over 95% of firms listed in the first section of the Tokyo Stock Exchange, a grouping of mostly large firms, now have two or more independent directors, up from 22% in 2014.
This is especially notable for two reasons: first, many companies in Japan, scarred by the country’s long recession and slow recovery, have vastly more conservative financing than they really need. This slows down economic growth from the supply side (because they're not investing in more production) and the demand side (they're not paying dividends that turn into consumption, either). It's also notable because the value stock recovery has happened more slowly in Japan than in other places. M&A is, historically, a big driver of value stock outperformance; the cheap stocks of the 50s got snapped up by conglomerates in the 60s, and survivors of 70s economic volatility got swallowed by private equity in the 1980s.
Flipping (pt 1)
The NYT has a piece on the trend of companies buying out Amazon merchants. It's really two separate phenomena, driven by two forces:
- Growing sellers are working capital-constrained, and have to reinvest in inventory to capture upside. It's entirely possible to start a business that naturally generates, say, $100k in annual profits, and to go broke because you accidentally scale it on the assumption that the natural run-rate is $200k. A better-capitalized buyer can handle this risk.
- The acquirers who buy multiple Amazon sellers are benefiting from a financial free lunch in the form of diversification.
Thrasio might be a nice example of applying financial theory in a practical way: first, it may be the case that the price at which Amazon sellers sell their business is determined by the perception that Amazon will copy them, while the actual risk is lower. Second, and more interesting: copying risk raises the variance of any given seller’s returns: if they don’t get copied, they do fine; if they do get copied, profits go away. So Thrasio might be structured around benefiting from the one free lunch in finance: if the market is efficient, switching from a concentrated portfolio to a diversified portfolio in the same asset class produces the same expected return, with less volatility.
With multiple companies raising sizable funds to invest in this kind of venture (most recently Keith Rabois' OpenStore), some of the free lunch is disappearing; a competitive bidding process in which buyers all have similar advantages leads to a market where prices reflect those economics. And that creates an interesting situation for people starting an Amazon-focused brand. Their job isn't to build and run a business; it's to build and sell one.
(Disclosure: I'm long AMZN.)
Flipping (pt 2)
In other flipping news, independent fix-and-flip homebuyers are having an easier time getting loans as profits rise. The growth of iBuyers is slowly changing the house-flipping niche: over time, their job is going to shift from "spotting cheap properties" to "spotting hard-to-evaluate properties" as iBuyers' models get better at understanding the typical home. One risk in this business: while loans are available, they're not cheap (per Bloomberg: "The 7.9% average annual rate on a fix-and-flip loan is more than twice the 3.09% rate that a bank can earn on a 30-year mortgage"). Supply shocks in building materials are very expensive to someone who's paying a lot for money, and at least in plastics the supply shocks are already here ($, WSJ).
Liberty's Highlights shows a very informative chart from Sanderson Farms (ctrl-F "Can we make this chart"). Accounting is always somewhere along the efficient frontier between an informative summary and an excess of detail; the chart shows lumpy cash-based operating details alongside a smoothed accounting accrual. This is a very helpful way to understand the cadence of a business' cash needs.
The first group to really hype up Superhuman was VCs. They are 1) people who have highly variable comp, 2) people who spend a lot of time on email, and 3) people who, if they get to the right email a week late and miss out on something important, will literally never forget it. The real growth hack wasn't selling to them; it was convincing them not to keep it a secret. ↩
Was this caused by someone at YC looking at server logs and finding a floundering founder who was spending all their time on Hacker News instead of working on their startup? We may never know. It is probably true that, if Noprocrast has saved just a handful of founders from terminal time-wasting, it’s one of the most wealth-creating software features of all time. ↩