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 15,554 subscribers, up 254 week-over-week.

In this issue:

• Inside the House Report on Big Tech and Competition: Part II

• Yelp’s Crypto-Coinbase Moment

• Life at a Hypergrowth Startup

• Microsoft’s App Principles

• Square and Bitcoin

• Driverless

• Local Champions and Good Globalization

• China’s Domestic Consumption Push

Inside the House Report on Big Tech and Competition: Part II

This is part two of a series. Part I looked at the background of the report, its rhetoric, and its underlying antitrust theory. In part II, we’re reviewing the details of how Facebook and Google achieved and extended their market dominance, through product decisions, strategic relationships with other companies, and acquisitions. Part III will continue on Monday, reviewing Apple and Amazon—two companies that seem to have embedded some monopoly-looking economics in businesses that don’t have dominant market share.

A recurring theme in the report is that many of the traits that make big tech companies fantastic for shareholders and employees—their ability to continuously reinvest at scale, their power to shape markets so they capture a larger share of the upside and lock in users and advertisers—are much more worrisome from a regulatory perspective. There is not a strong legal case to be made against companies that are merely big and profitable, but there are precedents for going after companies that use their size to squash higher-quality offerings.

But a meta theme is that the lifecycle of most companies is that they grow by delivering much more value than they charge for, they raise funds based on their growth, and then they make those investments pay off by capturing more value. In its maturity, a company will capture a greater share of the value it creates. That’s part of the career cycle, too; early-career employees don’t get regular raises because their productivity is rising, but because they’re getting more indispensable over time. Your relationships with your coworkers and customers, and your knowledge of your company works, represents a form of vendor lock-in that lets you capture more of the economic value your employment creates. This is very much part of the deal; employees accept being proportionately underpaid early on in exchange for a shot at being proportionately overpaid later, and the company’s bet is that this works out over the full cycle.

So it’s important to look at these companies through the full cycle of their economic lives. Very few tech companies have been able to stay dynamic for decades, although the current crop of winners has had an unusually long run. Measuring their behavior now risks ignoring what they did to get to this point, and also ignores the statistically inevitable decline that at least some of them will experience.

With that in mind, let’s take a look at two of the report’s cases, against two companies that dominate their core businesses of social networking and search.

Facebook

It’s a bit revisionist to say that Facebook has retained “an unassailable position” for nearly a decade. It’s true that Facebook has been #1 for a long time, but only because it’s dealt with a constant stream of assailants. The pivots that kept Facebook ahead of the competition were all controversial at the time (the News Feed was essential for making the service real-time, but Facebook only knew it worked when 10% of their userbase joined a group devoted to hating the feed; Facebook’s revenue is mostly mobile now, but their mobile transition took years, and was controversial with investors—Facebook was actually sued by investors for not disclosing how quickly the service was transitioning to mobile).

The report quotes an insider saying that the “Only metric is getting another minute.” This is a quote from an unnamed Instagram employee. Based on the interview date, it’s probably the one who asked earlier in the report why Facebook is allowed to collude with Instagram. (Answer: the same reason Coke can collude with Diet Coke, instead of both products converging on a 50/50 corn syrup/aspartame-based sweetener mix.) So, not necessarily a reliable witness, albeit an informed one.

The House Subcommittee is right that Facebook dominates social networking. But Facebook is right that there are many other activities that compete for people’s time. The Big Three have retained high share of the network television business, even as that business has gotten far less relevant. But even ten years ago, Paul Graham could talk about why TV lost to Facebook. The year that article was written, Facebook’s annual revenue was $777m, or about what the broadcast TV industry brought in every three days.

This seems to be the common pattern in tech antitrust. There’s a broad human need that no one company satisfies—human connection, entertainment, shopping—but for any specific form-factor that solves it, there’s exactly one company that’s dominant. If you want to entertain people, there are many jobs you can pursue. If you want to entertain them by streaming long-form video, you have to work for or with Netflix; if your idea is short-form video entertainment, TikTok; if it’s aphorisms or witty comebacks, Twitter. (Tweets by Naval and Nassim have been collected in books. And Twitter is a powerful enough medium that before it existed, people who wanted to write tweets had to invent fictional personae.)

The report talks a lot about app-level “tipping points”—once Facebook is the dominant social app in a country, it never loses that status. Messaging apps are in some ways less sensitive to tipping points: while messaging apps theoretically have all-or-nothing characteristics, the recent crop of privacy-focused apps is different. Since privacy is the whole point, they grow more slowly, and in parallel, but they have high switching costs.[1]

The report also notes that Facebook has acquired major competitors, especially WhatsApp and Instagram. A counterpoint to acquisition as a way to cut down on competitors is that potential acquisitions are a subsidy to competitors. Since Instagram is worth more to Facebook than it is as an independent company, Facebook’s potential desire to acquire them is a source of cheaper capital. At the limit, every social media app gets to free-ride on Facebook’s mature monetization model, because if they succeed they can convert their business into shares of Facebook.

If it’s not legal to acquire a competitor, the nearest substitute is to copy. But it’s not helpful to consumers if Facebook gets a little more Pinteresty and Pinterest gets a lot more Facebooky, and ultimately as the products converge, the biggest one wins. Allowing big companies to acquire smaller ones that compete for a subset of their user-time might actually lead to a more diverse app ecosystem, albeit one without as much underpriced ad inventory.

One factor that adds some drag to Facebook’s growth is burnout. Sam Lessin is the author of some of the blunter internal Facebook memos cited in the report. Today, he’s a writer, who is broadly critical ($, The Information) of how lucrative the tech industry has gotten. This is an entropic force that slows down dominant companies. When they start, they’re underdogs, but if they win, everyone who worked there to be an underdog can quit. (And, after cashing in their options, they’re free to spend their time however they want, including criticizing their former industry. The “Mad as hell and not going to take it anymore” scene plays out in tech every time a big chunk of equity-based compensation vests.)

Google

The report’s coverage of Google actually does a good job of understanding why search is a winner-take-all market: the biggest search engine has more data on long-tail queries, so it tends to have fewer frustrating searches. And since the main reason to visit a new search engine is if the previous one didn’t deliver good results, the biggest search engine will tend to have the best results and lowest churn. Long ago, Google actually identified cases where Bing used Google’s own search results for long-tail queries (Bing argues that this was an unintended consequence of a useful feature of their toolbar.)

In one sense, search is a standard, like weights, measures, or the various grades of crude oil. But it’s a standard that has to be implemented at great expense, lends itself to language-level economies of scale. Google’s last remaining country-leading competitors, Baidu and Yandex, are both in different languages. And the search engines that most recently lost their market share dominance to Google, Seznam and Naver, were also in different languages.

The businesses that get cited most frequently as victims of Google are vertical search providers. Vertical search generally relies on Google for traffic, but is that because Google is dominant or because they’re arbitraging weaknesses in its algorithm? There are some vertical search products that get the first click—Amazon, Kayak, and Yelp are all good starting points for particular kinds of searches. But other kinds of vertical search get the second click, after someone searches Google for a term, clicks the #1 result, and is presented with a search page. Google argues that this is a bad user experience; the searcher looked for an answer, not a follow-up question. Vertical search companies argue that it’s anti-competitive, and that Google is deliberately refusing to do business with companies that compete with it. These are not mutually exclusive.

For repeat-use vertical search, it makes economic sense that a search provider would win over a search engine. Their economics support loyalty programs for users, which are a complement to deeper and more detailed search results. But for single-use searches (hiring a plumber, writing a will, tracking down a particular article) it makes sense for a general-use search engine to have dominant share. This is one reason such verticals are dominated by businesses that monetize after the search: there are many sites that can suggest a new name for a business, but Shopify is the one that makes the most money from people who have a business idea and just need a name before they get started.

In the report, Google admits to some bad behavior with respect to vertical search. The report quotes an internal memo about Google ranking its in-house vertical search products above competitors, noting that Google’s products would never rank well on their own merits. (How much of this is because of their merits as services or their merits as pages optimized to rank well on Google is unclear from the context.) What is clear is that Google gave its in-house products a boost in order to improve their traffic, at the expense of other products that were ahead of Google at the time. But Google’s vertical search products have continued to improve over time, and since they can monetize both by selling ads and by increasing Google usage in the aggregate, Google-owned vertical search tends to select against pages that exist solely to sell ads. CelebrityNetWorth, for example, is cited as a business that lost most of its revenue when Google started placing answers in the search page instead of directing searchers to their site.

What’s paradoxical about Google is that most of its aggressive competition is in other companies in the search business—either companies like CelebrityNetWorth that answer a single question, or companies like Yelp and vertical search sites that answer a category of questions. But the alternative to this competition is for Google to subsidize direct competitors, by giving them free traffic. Google can accurately say it’s in the business of changing text queries into answers, and that the companies it harms are in the business of changing search queries into ad clicks. Google’s harm to these businesses comes from making the Internet less commercial for some queries.

It’s notable that in at least one case, Google tried to promote its own vertical search, and failed. Google Video was ranked inline in search results, but YouTube still grew faster, forcing Google to acquire them. (The NYT article on the deal compared it to the dot-com bubble; one analyst estimates that YouTube is now worth 188x its acquisition price.) So Google’s actions are somewhat constrained. It can try to favor its own services, but if there’s a clearly better option, Google has to either buy it or live with the lost market share.

Google’s long-term incentive is not to have kill zones; it’s to have buffer zones. Google wants a world where search-dependent services can turn a profit, but not reap monopoly profits. Google also wants it to be relatively easier to start a service than to grow one; the more scalable something is, a) the more likely it is to be gaming Google in some way, and b) the more likely it is to eventually represent a threat to Google. If every content business eventually gets killed by algorithm changes, nobody will start a content business (or they’ll start one in a different medium that Google doesn’t control). If every e-commerce company’s margins get squeezed by Google, that means Amazon’s market share rises relentlessly; every time e-commerce consolidates, Amazon Prime looks like a better deal, and Prime users are more likely to start their product searches on Amazon.com than Google.

Consumer surveys reliably show that many Google users are unaware that the ads (which, in the search interface, are labeled “Ads”) are ads, and that they’d be less likely to click if they knew. This is evidence that Google’s design is not especially transparent, but also evidence that searchers' concerns are misguided: if the ads are so unobtrusive and well-targeted that many customers can’t distinguish between them and the best organic results, that’s evidence that ranking ads well doesn’t impede the quality of the product—it just means that more of the value of a good search engine accrues to the search engine, rather than the sites it indexes.

Many tech companies see a pattern where user growth plateaus long before revenue growth. The main way for this to happen is for them to create a vast consumer surplus upfront, and then capture a share of it later. While this pattern degrades third parties' economics over time (it’s great to depend on Google when they’re charging less than the value of the traffic they send), it also creates a strong incentive to produce all that consumer surplus in the first place. Google was founded to organize the world’s information, but it was funded based on its ability to charge for this, and many other companies raise funds on the basis that if they captured just 1% of what Google has—DuckDuckGo’s slice of the privacy-sensitive market, LinkedIn’s lock on searches for names, StackOverflow’s dominance in software Q&A—they’d be worth a lot. The pattern of value creation first and value capture later leads to ongoing disappointment, but it’s a fact of life beyond tech: plenty of other important relationships reach their peak fun level in the first few months but are worth maintaining long after the honeymoon is over.

Google’s use of Android is more dubious. Granted, it’s a fantastic business decision: lock handset companies into a Google-controlled operating system, gradually force them to install more Google-owned apps (to the point that a new phone with 16GB of storage came with 7-8GB taken up by Google’s products), and then track user data to assess competitive threats. This is hard to defend on its own merits. In the US, it’s less problematic, because Apple has high market share and pitches its phones as a more privacy-friendly alternative to Android (among other benefits). But in countries where Android is the dominant OS, Google’s lock on the market continues to grow. Here, the competitor harm is much more visible than the consumer harm: when Google tracks TikTok usage to refine its plans for a TikTok clone, that’s mostly a problem for ByteDance, not for users.

Chrome, like Android, is an example of Google extending its search dominance into an adjacent area and then reaping the benefits. Google promoted Chrome through its search page, probably the single most valuable display ad inventory in history. But Chrome also won by being a better browser. Leveraging distribution advantages is useful, but it’s not enough; when Windows 10 can’t resolve Start Menu queries by pulling up programs or documents, it defaults to Bing, and despite Microsoft’s dominant markeshare in desktop operating systems, this hasn’t made Bing a major search engine. Google’s aggressive marketing campaign for Chrome was also a bet that Chrome was a better product, and so far that bet has paid off nicely, both for Google’s business and for Chrome’s users.

The overall case against Google is weakest in search, where Google’s behaviors tend to be wealth-creating, but increasingly tilted towards creating more wealth for Google. They’re strongest for Android, which has allowed Google to run a Microsoft-in-the-90s strategy in reverse: use a dominant Internet application to create a popular operating system, then use that operating system to reinforce the original application’s dominance and extend it to other products. On the other hand, Android preserves a balance of power in mobile: precisely because Google monetizes it mostly indirectly through apps rather than directly by selling it, Android creates a lower-priced tier of smartphones that compete with Apple.

Conclusion

Tech antitrust is always backwards-looking, because technology moves fast and Washington is slow. For example, the majority of US TikTok users started using the app between when this report was started and when it was published. (The report concludes a 16-month investigation. TikTok’s US active users rose from 40m in October 2019 to over 100m by August 2020.) And as Benedict Evans points out, some of the data the report cites is nearly a decade old.

There is a good case that these companies have behaved aggressively, but it’s important to balance that against what consumers get. “Network effects” usually describe a business that gets more profitable as its user count grows, but that’s the second-order effect of the product being better as it grows. Scale effects are similar: every search engine needs to spider sites, construct and store an index, and rank results, which means that every new search engine starts by duplicating what already exists, at great cost, and providing a slightly worse product by default because it’s informed by less data. (DuckDuckGo, the most interesting direct challenger to Google, gets around this in two ways: it focuses on user privacy, and it offers some very useful shortcuts for certain use cases. Since DuckDuckGo doesn’t customize results based on individual users' data, the search results are less relevant to them—but also less likely to confirm their own biases.)

The economics of software reward scale, but generally because software products are better at scale. Google resolves long-tail queries well; two half-Googles would each be less than half as good at this. Facebook connects people, and two half-Facebooks would have fewer than half the connections (and would be able to support far smaller moderation staffs, and might feel more pressure to break end-to-end encryption, allow more dubious advertisers, etc.). Those wonderful scale-and-network-effect economics depend on fragmentation one level up and down the supply chain, so they’re always a tempting target for competitors, and the profitability of a mature tech company creates a public-market comp for funding incumbents.

This means that every successful tech company simultaneously exists in two states:

1. Enjoying cozy monopoly profits, and periodically tweaking its core product to shift slightly more surplus wealth to itself instead of the advertisers and users of the service; and

2. Desperately trying to stop getting disrupted, disintermediated, or rendered irrelevant by competitors, often competitors that don’t even look like they’re competing.

The biggest threat to tech companies is other tech companies, which means the best way to keep the market competitive is to ensure that there are rewards for building a company big enough to scare Google, Facebook, Amazon, and Apple as much as they scare one another.

Text-based search is mature as a product, but won’t matter forever. Voice search, for example, is intrinsically more vertical-specific: a voice-search product for finding local businesses will be very different from one that provides health advice or controls smart appliances. Training an algorithm is less feasible when results are presented in sequence rather than in parallel, and it’s more important to invest in advance in getting the answer exactly right.

Social is also a mature category. A generic friend graph is inconvenient to rebuild, so it’s hard to imagine a Facebook competitor that works by going head-to-head with Facebook. But some friend graphs are ad hoc and non-overlapping; office friends, poker buddies, a book club, and a church congregation all have very different social networking use cases, and a single large social graph won’t serve them all, and may actually degrade their experience. Facebook offers granular access-control tools, so anyone who wants to invest the time can choose exactly who sees which interactions, but spinning up group chat on Signal or Telegram is an easier way to create a single-purpose social network that’s unlinked to other identities by default.

Every company that benefits from repeat Google and Facebook traffic is scheming to keep users for itself; TikTok spent vast sums on user acquisition before it figured out how to get user retention up to sustainable levels, at which point paid traffic from social media started to be less important, and PR and organic sharing began to matter more. Now, if Twitter tried to suppress the popularity of TikTok videos on its platform, it would degrade the Twitter experience (they were on to something with Vine!), and Twitter’s best bet for providing bite-sized entertainment is to accept that some of that entertainment ends up advertising a faster-growing competitor.

A mature product does not imply a mature business. Modern planes and cars would be mostly recognizable to someone who had been in a coma for fifty years, but the businesses behind them have learned new tricks for increasing margins. The growth story for big tech companies isn’t over, and as that growth continues at a faster pace than the consumer surplus these companies generate, they’ll eventually have to either give up on growing or start seriously degrading the user experience in search of more revenue.

The current regulatory backlash against tech might be comparable to the Goldwater campaign in 1964: it got a few people very excited, but ended up attacking a system that was actually working pretty well. Railing against big government around the peak of US state capacity was an admirable act of bullet-biting, but ultimately led to a massive electoral blowout. And then, sixteen years later, a popular pro-Goldwater speaker ended up getting elected, and helped tear down some of the regulatory cruft that had accumulated since then. (Since politics is partly a game of taking credit for outside events, Reagan gets points for legitimizing some Carter-era technocrat-driven deregulation.)

Tactically, the right move for Facebook and Google is to draw out the antitrust process for as long as possible. Each company can hope that the other will face some kind of resounding economic loss from an out-of-left-field competitor, which will demonstrate that the consumer Internet field remains dynamic. Ironically, the biggest regulatory risk for every major tech company is that all of the other ones remain uniformly excellent at staving off competition and irrelevance, but over time that gets less and less likely. After all, if they’re all invincible monopolies, then soon enough nobody else will be left; all of Facebook and Google’s ad dollars will come from Amazon, all the products Amazon sells will be ordered on Apple phones, and there won’t be anyone in the middle left to squeeze. And at that point, the anti-tech case will be rendered paradoxical: they’re all unstoppable until one of them stops another.

[1] On my phone, I have WhatsApp, Signal, Telegram, Keybase, ChatSecure, and Wickr, because everyone paranoid enough to prefer an encrypted app is also paranoid enough to distrust some other encrypted apps. In fact, one way the shape of the market is changing is that more interesting conversations migrate to private chat apps. Since their churn and virality occur at the level of the conversation itself, not the platform, they lead to more fragmentation. Any messaging platform tied to a social platform loses its winner-take-all characteristics as users get worried that their private comments will be tied to their public identity and get them in trouble.

Share

Classifieds

Are you always trying to understand how trends become trends? One of my favorite startups, which surfaces hidden trends and explains why they are taking off, is hiring a curious thinker who also enjoys writing. The ideal candidate should be comfortable researching a range of topics and generating insights through a range of lenses: behavioral economics, consumer behavior and business strategy. Professional writing experience isn’t required, though the ability to write well and quickly is key.

If you’re interested, or know someone who is, please get in touch.

Elsewhere

Yelp’s Crypto-Coinbase Moment

Yelp has made a very interesting and widely-misinterpreted announcement: they will be flagging businesses whose employees have been accused of racist behavior. This sounds like taking a side in a culture war issue, but it ignores some of the copy. The notice appears when users check reviews of a business, and the notice ends with “We have temporarily disabled the ability to post here as we work to investigate the content. If you’re here to leave a review based on a first-hand experience with the business, please check back at a later date.” Leaving one-star Yelp reviews is a classic way to punish a company for its behavior, and reviews stick around whether or not a viral story turned out to be true, and regardless of how the company responds. Yelp’s decision to block reviews is in part a way to protect the businesses themselves; once a local business gets tied to a national issue, Yelp doesn’t want to be the default venue for discussing it.

Life at a Hypergrowth Startup

Patrick McKenzie has a retrospective on spending four years at Stripe. It’s a knowledge-dense and especially wisdom-dense document, worth reading both for the Stripe-specific insights and for broader lessons. On Stripe itself, I liked this point:

I don’t think Stripe is uniformly fast. I think teams at Stripe are just faster than most companies, blocked a bit less by peer teams, constrained a tiny bit less by internal tools, etc etc. There are particular projects which have been agonizingly long to ship; literally years after I would have hoped them done. But across the portfolio, with now hundreds of teams working, we just get more done than we “should” be able to.

A stupendous portion of that advantage is just consistently choosing to get more done. That sounds vacuous but hasn’t been in my experience. I have seen truly silly improvements occasioned by someone just consistently asking in meetings “Could we do that faster? What is the minimum increment required to ship? Could that be done faster?” It’s the Charge More of management strategy; the upside is so dramatic, the cost so low, and the hit rate so high that you should just invoke it ritualistically.

It may not be a coincidence that patrickcollison.com/fast exists.

On general business wisdom:

(In a way, every scaling startup is an experiment in empirical microeconomics research on “What parts of the typical corporate form are necessary and which are pageantry which we only keep around due to anchoring, the sunk cost fallacy, and tradition?” Every time a startup bites the bullet and hires a VP of Sales, a lifecycle email copywriter, a retirements benefits administrator, or a cook, count that as a published result saying “Yep, we found this to be necessary.”)

Microsoft’s App Principles

In light of how Google and Apple use their app store and OS dominance, Microsoft has articulated a set of principles for how it wouldn’t abuse that market power, if they had it. Microsoft has certainly gotten gentler over the years, and is in the fortunate position that its market share is highest in products that remain lucrative but aren’t as attention-getting as they once were, while its growth is coming from more competitive fields where Microsoft has just enough scale to see increasingly compelling economics. Microsoft’s perceptual transition from brutal competitor to trusted neutral party is one of the most impressive parts of their turnaround under Satya Nadella, and this declaration—timed to coincide with discussions of other companies violating these principles—is a good way for the company to reinforce it.

Meanwhile, in an example of big tech companies disrupting one another for long-term strategic benefit, Microsoft is adding price comparison features to its Edge browser. This mostly harms Google, by a) creating a way to search for products without Googling, and perhaps more importantly b) creating an ad unit that appears on the page users click through to when they Google a product and click one of the ads. So Microsoft is depriving Google of some click volume and some click value.

Square and Bitcoin

Square has spent $50m, or about 1% of its assets, on Bitcoin.

My comment from last time this happened still stands:

The event path for Bitcoin has always been gradual adoption by people who are less and less crazy. MicroStrategy is an interesting signal, because the CEO owns most of the voting stock. So while Bitcoin is technically an asset that a public company can use as a financial reserve, in practice it’s something a CEO can speculate on if he has largely unfettered control over other people’s money. Still, this means the next public company CEO who decides to take a flyer on Bitcoin has a precedent to point to.

Driverless

Waymo, after many years of approximations, has launched a driverless taxi service in suburban Phoenix. This product was delayed, but turned out to have great timing: a driver-free taxi service is just the product for a pandemic. Right now, the unit economics are still not great. As the article points out:

Eliminating the safety driver is an important step toward making Waymo’s service profitable. But it may not be enough on its own because Waymo says the cars still have remote overseers. These Waymo staffers never steer the vehicles directly, but they do send high-level instructions to help vehicles get out of tricky situations.

But optimal redundancy is easier to provide at scale, where the law of large numbers is favorable. Pushing back against this is the issue that self-driving is a problem where the first 99.9% gets solved by building the right general equipment and algorithms, and the last 99.9% by adapting to a given city. Waymo’s incentive is to grow fast once their economics are remotely viable, and expect those economics to improve, but the drawback is that every new city presents the PR risk of an accident, which would hurt their growth everywhere else.

Local Champions and Good Globalization

Nikkei has a lengthy writeup of KKR’s Asia Ex-China investment strategy ($), casting PE investment in the region as a way to recapitalize family-owned businesses and local champions facing a cash shortage. It’s interesting for how surprisingly normal it is: in theory, countries with low growth and deep financial markets should be net investors in countries that are growing faster but don’t have complex financial systems. In practice, that flow is often reversed, as export-driven economies accumulate dollar-denominated assets to keep their currencies cheap. KKR’s investments represent a bet that globalization will look more like it did in the 19th century, driven by local companies that raise funding overseas.

China’s Domestic Consumption Push

In the long run, China’s policymakers want the country’s economy to tilt more to consumption rather than exports and investment. In the short term, investment is generally an easier way to drive growth. Two recent stories show some evidence that this rebalancing is working: a combination of worse relations and higher transaction taxes have discouraged Chinese investors from buying properties in Australia, and duty-free shopping on the Chinese island of Hainan is taking share from South Korea and Hong Kong. (As Jordan Schneider of Chinatalk pointed out to me last time Hainan came up, some observers think it’s easier for the Chinese government to consider large-scale reforms on islands, where the spillover effects to other provinces are more contained. So Hainan may be a leading indicator for policies that China may consider elsewhere.)