Newsletter Economics in 2022

Plus! Dogfooding the Metaverse; Distribution; Rebundling; Negative Yields; The Labor Shortage Continues

Newsletter Economics in 2022

Way back in the basically prehistoric year 2019, I decided that I should find a career where my success wasn't so closely tied to whether the market was going up or down. At the time, I worked at a company that offered research to hedge funds and other investors. In theory, we should have been hedged, too: the funds were mostly long/short discretionary funds which had a mandate to put up good numbers regardless of what the overall market did. In practice, there was a notable chill in the business every time stocks dropped, and that was particularly irksome if it happened right around when bonus season hit, like in 2018: paying a good bonus after a generally good year is just the normal course of business, but paying a bonus based on three wonderful quarters in the middle of a quarter where the Nasdaq has lost 20% of its value is a trade, a bet that the market's decline didn't represent a secular shift.

Anyway, I didn't like it. So I started a newsletter business, where revenue would be steady as long as I could produce content people wanted to read and where growth would show up if the writing was what people wanted to share.

But, as it turns out, paid subscription growth now rises and falls with the market—the r-squared between recurring revenue growth and one month lagged market performance was roughly zero for the first couple years of publication, but has been about .25 this year. A month after the market peaked at the end of last year, growth stalled; a month after the market started rallying in June, growth came back. If you're not going to be in the business of running dollar stores or repossessing stuff, your job will eventually have positive beta. Oh well!

Now's a good time to take stock of the newsletter industry: how does it work right now, why does it work, is the deceleration from the Covid boom meaningful, and what's next?


There are basically four revenue models for newsletters. Most successful ones use more than one of these (The Diff uses three):

As is true in almost every business, newsletters usually have one dominant revenue line and then a few other things that are worth doing on the margin but won't support the entire business.1 For this newsletter, that business is subscriptions, at least for now. There are viable ad-supported newsletters, which can sometimes spin off a subscription product as a niche upsell. There are some very lucrative newsletters that run sponsored posts. Hiring, investing, software services (for business-focused newsletters) and health products (for more entertainment-based newsletters) are natural ways to turn a newsletter into lead generation—the ad-supported model itself is really an outsourced lead-generation model.

Basically any media business is a balance between monetization and growth. Conveniently for newsletter writers, the same raw material can be used as either a way to advertise the newsletter or as a way to harvest that demand. Usually, the right growth pattern is to slowly reduce the share of free content as the newsletter gets closer to saturating its target market. Depending on the pricing, the optimal choice can be to eliminate free posts and market through ads instead.

There's an endless series of games publishers can play with free trials, discounts, cross-posting, etc. They're helpful, but not a panacea. Some people will sign up on a free trial and instantly cancel. Or they’ll use a single-use card so they can’t be re-billed. All this is basically fair: some fraction of free trial users have full intent to pay and are getting an extra week of free content as a signup bonus. Some are signing up to read a single article, or have very low intent to purchase (e.g. they signed up because of one post that was exactly what they wanted, and if they’d gotten another perfect-fit post in the week they signed up, they would have stuck around, but otherwise they'll churn). The upside to giving someone who probably has purchase intent a free preview is high enough that it’s worth the occasional friction. In a world of perfect price discrimination, some people would willingly pay the $20 monthly cost, or even the $220 annual cost, to read a specific story, but the transaction costs of capturing those readers are too high to charge by the article.

Sponsored posts can be a great business—I could probably cut my workload and increase my income if I stopped doing subscriptions entirely and switched to weekly sponsored posts instead. But it’s also a challenging business, because it’s hard to balance conflicts of interest even if they’re fully disclosed. For more than one post, I've started with one thesis in mind and ended up deciding the exact opposite was true; it's not fun to navigate that kind of situation when there's a direct financial incentive in the way. Which is not to say it can't be handled: a writer can judiciously vet every opportunity before agreeing to do a sponsored post. Or a natural optimist can be confident that they'll probably end up saying positive things because they're naturally positive people.

And the fact that sponsored posts entail tricky questions about conflicts and disclosures don't mean that other models avoid this problem. Any kind of writing that involves scoops and revelations, for example, benefits from a close relationship between writers and sources. And that relationship often means that each side is useful to the other. A completely fictional scoop is rare, but a scoop that's a well-tailored subset of the truth is not. (This showed up a lot when Uber and Lyft were running dueling negative PR campaigns: both companies used many of the same strategies, so they competed to see which of them could get a publication to write about the other one's adherence to industry norms first.)

A good reason to be careful about conflicts of interest is that either in moral terms it's not good to let financial incentives distort what readers get, and in practical terms it's bad business to allow even the appearance of that. And in the long run, the upside from lead generation, broadly defined, ends up being bigger than the upside from other business models. More on that in a bit.

Process and Volume

A question I get just often for it to be flattering is: how do you write this much? The Diff usually adds up to around half a million words a year. That may be excessive; the target audience is and always has been people who like to read a lot about a variety of topics, so it's not a good target to aim for in a more specialized or focused publication, or one that's aiming for a different audience.

Publication frequency and content volume trade off between three main criteria:

  1. A higher ratio of research and editing to writing means better quality. The highest-effort Diff post this year was this Jane Street profile, which was also the most successful in terms of total reads, new free subscriptions, and revenue.
  2. Value perception is partly a function of how much gets produced, but also a function of how much people end up reading. I've talked to many people about why they stopped reading The Diff and other newsletters, and a very common complaint is that they had a backlog of unread posts and felt guilty. (True elsewhere: The Onion's "Stack Of Unread New Yorkers Celebrates One-Year Anniversary" headline celebrated its twentieth anniversary this year.)
  3. Writers have a terrible sense of which posts will be popular with readers. Writing about how Canada's economy looks like that of poorer countries, just with a higher GDP ($) and the economics of gas stations ($) wasn't supposed to be a way to play to audience demand, but got lots of reads. Other posts seemed destined for virality and fell a bit flat. Because that's hard to predict, and because the impact of a single hit is so extreme that it's worthwhile to up the odds, and doing that means writing more.

So if you're building a business around content, it's a good idea to aim for volume, at least during the growth phase. Building a publication is partly an exploratory process of building an audience, figuring out what they like, and iterating on that. And volume means sample size. It's also a good way to get practice, and a good way to get mistakes out of the way early.

Producing content on a regular schedule is a matter of manufacturing good ideas, researching them, and writing them up. The research and writing are usually a pretty straightforward process, but idea generation is tricky. What's interesting is that over time, there's been a close to 1:1 correlation between the number of article ideas created and the number of new articles written; the to-write list grows over time, but the corpus of written articles grows faster. When the newsletter started, there was a list of about half a dozen future posts to write. Now, the to-write list has 152 entries, everything from how investment memos reflect Feyerabend's epistemological anarchism to notes on a writeup of Capital One. But that's after publishing 4-5x a week for two and a half years, so additions to the backlog have only slightly outpaced articles removed from the backlog over that time.

To-write lists are basically a centrifuge, where the easiest posts get removed from the top and the weightiest, most effort-intensive and uncertain ones naturally drift down. This can be demoralizing after a while, but it's possible to work around it. The rough breakdown of Diff articles looks like this:

If you're trying to manage output and morale, which is what any solo creator has to be good at, the last bit is an important one. One drawback to writing lots of quick pieces is the worry that they're going to skew towards the insubstantial. It's not impossible for something produced on a deadline to turn out to be good; Robert Graves cranked out one of the best novels of all time to pay off a debt. But it's a good bet that treating quality rather than time as the key constraint is good for quality.

Ideas get onto the to-write list in lots of ways: sometimes there's a news story, sometimes there's a pattern in stories, sometimes there's a question from a reader, or a conversation. Some posts are born as extended footnotes to other posts. Writing one post doesn't automatically lead to at least one good idea, and there are months where the to-write list is getting steadily drawn down. But on average, between stray ideas, updates, news, and conversations, the idea flow is reasonably stable.

And there are decent ways to frack the idea-well and get it producing again. Plutarch got a lot of mileage out of taking pairs of similar cases and determining what made them different. Doing a breadth-first search for the most notable company in an industry by reading every 10-K also works. Investor days are a reliable source of insights into where a company is and where it would like to be heading. And while reading one book per published post doesn't work for a 5x/week schedule, most books do lead to at least one good idea, and some end up inspiring lots of them.

Most posts start as a single cryptic thought in a big Notion page with lots of similar thoughts. (It used to be Google Docs, and before that it was Emacs, but what it’s always been is unstructured text.) It is very easy to over-optimize the process of tracking ideas, making links between them, etc. It's generally easier to have a short enough list that scrolling through them is feasible, and that ctrl-F is all the navigation necessary. Over time, some of these notes develop into something resembling a draft, or at least something with lots of fragmentary thoughts and links relevant to the topic. And sometimes the extent of the notes is just something like "Why is that bowling alley SPAC deal outperforming?"

There are not a lot of tricks that make writing easier. The process changes all the time, and overthinking process is sometimes a substitute for just publishing and getting feedback. But there are two decent tricks that help a lot:

  1. If one of the notes in the to-write list seems like a good opening sentence, it's time to move that one to the front of the queue; if you consciously figure out how to introduce a topic, you've probably subconsciously figured out a good conclusion.
  2. For emergency use only: if you're running out of energy and coming up on a deadline, and you're not finished, there is a wonderful trick: go back to the title and add "Part I" to it.

Perks and Opportunities

The main perk of writing about topics you care about is that it starts a lot of great conversations with people you'd like to know. Every issue of a newsletter has the potential to start a dialogue, and some issues of The Diff get enough responses to eventually break Gmail's threading.

I've taken to calling newsletters a brand and customer list in search of a product. Selling content is a reasonable model, and it's put a few people on the Forbes global billionaires list (though most of them are owners rather than content creators—Oprah and J.K. Rowling are probably the only ones whose wealth can be described as a direct result of creating and selling content. But John Malone made a lot more money in media by letting other people make the content while he controlled the distribution and played fun games with tax deferral and financial engineering.

There are many successful companies that start out as some kind of media enterprise and turn into something different:

Notably, all the successful businesses launched from newsletters benefit from a network effect: asset management is really just making connections between people who have money and need investment opportunities and people who have opportunities and need money to take advantage of them. A distinct source of total dollar alpha, as opposed to percentage returns, is LP-opportunity fit—lots of generalists figured out the subprime short, but the big determinant of total profit was being able to raise money to make that specific bet. A mailing list is essentially a way to get a cluster of people who are all on roughly the same page about whatever the topic is.

The value of a newsletter as a media product scales roughly with the subscriber base, but the value of potential connections within that subscriber base will scale superlinearly.

Since the cost of newsletters is lower than it used to be, and it's much easier to build one to the point that it's sustainable for a solo operator, we should expect the world to have many more vertical-focused investment banks, funds, and marketplaces in the near future. Current newsletter economics mean that it's possible for these companies, early in their lives, to essentially have negative-cost marketing for whatever they end up launching, and as we've seen with consumer Internet in general, lowering the cost to start new companies wildly increases the diversity of those companies.

It's still easy to be skeptical of the newsletter economy, in part because it did have the one-time benefit of Covid—eliminating lots of commutes for people with email jobs was great news for sellers of email-based media—but a more comprehensive bear case rests on assumptions about the fungibility of content. If newsletter writers are competing with mass-market entertainment, they'll probably lose over time. But what they're really doing is providing human curation on top of an infinite flow of commoditized news, and identifying trends and narratives that have signal worth tracking for their audience. That's a job that will be in higher demand as the quantity of content rises. And it's a job that can naturally extend into more lucrative models than subscriptions, sponsored posts, and ads.

Further reading: The Diff goes meta on the newsletter business roughly once a year. 2020's was Newsletters and New Media Economics, and last year was Newsletters Can Move Markets.

A Word From Our Sponsors

Stream by AlphaSense is a library of investor-led expert interviews from one-on-one calls with former executives, customers, competitors, and channel participants across a breadth of industries. Interviews are conducted by buy-side analysts and searchable using AlphaSense proprietary AI technology, empowering institutional investors to get up to speed on a company in seconds with Stream by AlphaSense.

With Stream, you’ll have instant access to the fastest-growing expert call transcript library in the industry:

Sign up for a free 14-day trial with Stream today


Dogfooding the Metaverse

The team building Meta's metaverse app has been asked to start spending more time actually using the app. They're also delaying some growth in order to improve product quality first ($, FT). It's not impossible to build something you, personally, don't use; in The Facebook Effect, David Kirkpatrick remarked that the Facebook team didn't spend all that much time on the site. Part of what made early Facebook great was that it was so minimalist, and that the company was so data-driven, so there were fewer rough edges for them to notice and there was plenty of indirect user feedback on what needed to be improved. A company that hits product-market fit early may not need to spend so much time using their product, especially if they're going for mass appeal, because they may get the wrong idea about the best way to use it. (Some failed competitors to Facebook over-optimized around theories of how people would prefer to use social networks, like Google+'s well-crafted "circles" concept—as it turned out, most users just didn't want to spend time carefully categorizing their friends, and ended up just using different apps for different social groups.)

One place that Meta's metaverse vision could use dogfooding is when it's used as a replacement for Zoom. There is a good case to be made that Zoom is unnatural, both because it's hard to see who is looking at you and because "making eye contact" means staring at a little green light, not a human face. Strapping on some goggles trades one kind of discomfort for another, but it could work.

The trouble with aiming for that use case is that the market for Zoom-but-with-VR-cartoon-avatars is smaller than the market for Zoom itself, and if Meta changed its name and refocused its strategy in order to build a small subset of a smaller company, that would be viewed internally and externally as a failure. So they may be stuck in a position where the right way to iterate on their product conflicts with the right strategy for which product to launch first.


FTX and Visa are partnering to distribute a debit card that will allow FTX customers to spend their balances. Crypto mostly started as an effort to build a new financial system, but as it turns out most financial systems have some level of integration—and since money has powerful network effects, the best way to expand a new kind of it is to plug into existing ones. Given that FTX makes money when people keep their assets in crypto and trade them around, it might look like a mistake to let users pull money out quickly, but one thing early PayPal found was that the easier it was to take money out, the more willing people were to put money in. A debit card is a way to cash out for individual transactions, but it's also a way not to cash out in large increments through an account transfer, but to do so one purchase at a time.


The WSJ has a good look at some of the bundling deals streamers are doing, including Disney's offers of Disney+, ESPN+, and Hulu, and the Walmart/Paramount partnership ($). The piece has some good details on the mechanics of the latter deal, and they're worth thinking about:

Paramount Global spent a year negotiating its partnership with Walmart, with each party agreeing to a marketing plan. The pact allows Walmart to sell ad slots on Paramount-owned services such as Paramount+ and Pluto as part of larger advertising deals, the people said. Walmart promotes Paramount+ programming—including shows such as “Paw Patrol” and movies such as “Top Gun: Maverick”—on the smart TVs in its stores as well as on its website. Walmart also promotes Paramount+ on self-checkout screens in stores and gives customers an option to subscribe to Walmart+ as well.

Under the terms of the deal, Walmart agreed to pay between $2 to $3 per subscriber per month for Paramount’s ad-supported plan, so long as the users activate their accounts, according to the people familiar with the deal.

A retailer has high surface area for converting attention into transactions, and retail is partly a game of treating the entire store as ad inventory that can be sold to the highest bidder. One function of this deal is to convert some of the visible surface area of the store, like the TVs, into a monetizable property. Product placement deals like this are messy, because attribution is so difficult, but since Walmart is getting a piece of the resulting revenue, it can be agnostic on which element of the partnership paid off as long as the entire transaction did.

Negative Yields

The share of sovereign debt with negative yields has dropped from a peak of $18tr during the pandemic to $2tr today ($, FT), with Japan as the only negative-yield market left. The best way to think about long-term interest rate trends is that there are many policy levers that can affect nominal rates in the short term, but demographics and technological change are the big drivers of real yields in the long term. If real yields are close to zero, governments still have more flexibility if rates and inflation are positive; they can cut rates without worrying that people will switch to cash or use tax overpayment as a savings vehicle. Governments didn't want to get into that position because of deglobalization and higher energy prices, but sometimes markets find a way to accomplish policy goals that policymakers couldn't quite get to.

The Labor Shortage Continues

Despite macro worries, companies in the US are increasingly hiring people with criminal records ($, WSJ). Hiring is always a game of asymmetric information and asymmetric risk/reward: most companies would prefer to leave a job unfilled than to face a material risk of crime committed against coworkers, customers, or the company itself. (Though in some industries, dealing with low-level criminal behavior from employees is just a cost of doing business. Retailers spend a lot of time and money managing "shrinkage," but they don't try to manage it down to zero.) Technology shifts might be a big driver of shifting the risk/reward for employers: when they're constantly supervising employees already, and managing their day-to-day work in a granular and mostly automated way, some of the recidivism risk dissipates. A low-trust environment is also an environment in which perceived untrustworthiness matters less.

Diff Jobs

Companies in the Diff network are hiring! Some key roles:

And if you're thinking about your next role but not sure when, please feel free to reach out; we're happy to chat so we know what to keep an eye out for.

Interested in hiring through the Diff network? Let's talk!

  1. This shows up a lot with tech companies, and the more successful they are the more likely they are to be reliant on just one outlier revenue source. The biggest exception was LinkedIn, which actually made meaningful amounts of money from company subscriptions, ads, and user subscriptions.