8. Take Rates

Your margin is my opportunity.1

—Jeff Bezos

If you were an executive at an established business and you heard some hotshot dot-com founder utter the above threat in the mid-1990s, you might have laughed at the hubris. Later you would regret doing so.

Jeff Bezos, the founder of Amazon, was referring, without embellishment, to his strategy for seizing market share. The plan: minimize overhead, slash prices, eat rivals’ profits. Be lean, be mean. Be relentless.

The cost structures of the physical retailers that were Amazon’s competitors at the time prevented them from matching Amazon’s price cuts. Brick-and-mortar expenses, like rent, utilities, and shopkeepers’ wages, set hard limits to incumbents’ pricing. With no physical stores to support, Amazon could keep prices low. So Amazon pressed its advantage, undercut many of its competitors, and put them out of business.

Amazon’s lower cost structure lent itself to a deflationary business model, one that maintains or increases the value of a service while decreasing the cost to consumers over time. This combination of tactics has been popular since the early days of the commercial internet. It explains how Craigslist absorbed newspaper classifieds businesses,2 how Google and Facebook swallowed advertising-based media,3 and how Tripadvisor and Airbnb tackled the travel industry.4 In each case, the disrupters slashed costs and upended incumbents that were attuned to the cost structures of an earlier era.

Blockchains are the natural successors of this strategy. Just as internet startups undercut the high prices of traditional businesses, blockchain networks expose the soft underbelly of corporate networks: high take rates.

Network Effects Drive Take Rates

Networks make money by charging fees on network activities like commerce or advertising. The percentage of revenue passing through a network that the network owner takes for itself, as opposed to passing on to network participants, is, as you’ll recall, the network’s take rate. Absent other checks on a system, strong network effects usually mean high take rates because they lock in network participants who have few, if any, alternatives to turn to.

In the pre-internet era, scale was the main driver of pricing leverage. On the internet, it’s network effects that drive pricing leverage. Today’s biggest social media companies have very high take rates, demonstrating the strength of corporate network lock-in.

Of the large social networks, YouTube is the most generous with creators, taking 45 percent of revenue for itself and passing 55 percent to creators. In its early days, YouTube faced stiff competition from other up-and-coming video platforms that were offering to share half their ad revenue with creators. Feeling threatened, YouTube set up its revenue-splitting “partner program” at the end of 2007, and it has stuck to it ever since.5

Such generosity is uncommon, though. Facebook, Instagram, TikTok, and Twitter extract about 99 percent of their networks’ primary revenue source, advertising. These networks have all recently created cash-based programs6 to give creators kickbacks. Most of these programs take the form of time-bound “creator funds”7 and static pools of money rather than YouTube-style revenue shares. Creators receive mere fractions of these networks’ take-rate revenues, usually less than 1 percent, and the companies are under no obligation to continue supporting these funds over the long term. Worse, the fixed-pot model can make the relationship between platform and creator zero-sum since it forces people to fight over limited resources.8 As Hank Green, a longtime YouTuber points out, “When TikTok becomes more successful, creators make less per view.”

Even after accounting for their creator funds, the biggest social networks share almost nothing with network participants. This is great for the networks, but not for creators, who provide content without receiving their fair share of revenue in return. On the other side of the networks, these companies use their leverage to extract personal user data instead of money, which helps them earn more through better ad targeting. Network effects with lock-in amplify pricing power.

Apple has extraordinary pricing power9 thanks to its captive audience of iPhone users combined with the network effect derived from the iOS developer ecosystem. Apple exercises this power10 through strict rules around payments, which the companies subject to them loathe. Ever try to subscribe to Spotify11 or buy an Amazon Kindle book12 via an iOS app? You can’t. These businesses don’t want to pay Apple’s take rate, which can run as high as 30 percent. A common work-around app developers use to circumvent Apple is to accept payments only in mobile web browsers, not in apps. (The web and email are the last free havens on mobile phones.) On a technical level, Apple could override this bypass and force all transactions to route through the App Store, but Apple hasn’t dared to clamp down here. No doubt there would be a strong backlash, and probably legal and regulatory ramifications.

Some companies would rather go to war13 than fork over so much of their revenue to Apple. In fact, app developers are so fed up with Apple’s take rates that they’ve banded together to sue Apple14 over its dominant market position. But unless courts and regulators say otherwise (and barring some other unexpected business comeuppance), Apple can—and will—keep charging remarkably high fees. It has that power because it has a captive network.

If monopoly exacerbates rate taking, competition keeps it in check. Fees on payment networks remain relatively low thanks to the prevalence of interchangeable payment options. Multiple payment networks offer similar services, including Visa, Mastercard, and PayPal. The abundance of choice reduces businesses’ pricing power, to consumers’ benefit. As a result, credit card networks charge 2 to 3 percent on every transaction, a relatively low take rate, and much of that goes back to consumers in the form of points and other incentives. (One could argue these rates are still too high, a point I discuss later in “Making Financial Infrastructure a Public Good.”)

Physical goods marketplaces tend to have take rates in the midrange, higher than payment networks but much lower than social networks. For example, eBay (mostly secondhand goods),15 Etsy (handmade items),16 and StockX (sneakers)17 have take rates between 6 and 13 percent. Users can choose where to sell items and can cross-post listings on multiple sites. The take rates are lower partly because sellers earn lower margins from these goods, but also because of weaker network effects. Buyers discover items mostly through search results rather than social feeds, which lowers the cost for sellers to switch networks. Sellers can take their wares to any network they want because they own the physical goods they’re selling. When network participants own what’s valuable to them, switching costs drop and bring take rates down.

Protocol networks have no companies in the middle taking a cut of revenue, so they have no take rates. You own your domain name, and you can take it to any hosting provider you wish, no questions asked. Some access points, such as email and web hosting providers, charge for certain services; however, because protocol networks don’t have network effects that accrue to a central entity, as corporate networks do, hosting providers have little pricing power and must charge based on the costs of storage and networking, rather than as a percentage of revenue. As a result, even with these fees, the effective take rate—the actual price network participants end up paying to use the network—stays very low.

Effective take rates can be sneaky, like hidden fees that crop up at the checkout counter. Corporate networks often have effective take rates that exceed their apparent take rates. These networks raise rates by dialing down the organic reach of network participants in algorithmic social feeds and search results. Once creators, developers, sellers, and others attain a certain scale, corporate networks force them to buy ads to maintain or grow their audience.

You may notice, for instance, how a search on Google or Amazon yields an ever-increasing number of sponsored results18 (look for the “sponsored” label). Big companies use this technique to raise the effective rates on the supply side of the network, which for Google is websites and for Amazon is sellers. On Google, a website doesn’t pay for organic links, but it has to bid in an auction for sponsored links. On Amazon, sellers are charged a fee, but if they want sponsored placement, they’re charged additional fees.19 Google and Amazon know that users tend to click on the top links in search result rankings so, as they push organic links down, they are effectively forcing websites and sellers to pay more for the same exposure. As if that weren’t bad enough, these companies also use valuable screen real estate to promote their own products, which compete with those of their suppliers.

Google, Amazon, and other big companies were disrupters in their early days, when they were in the attract phase. Today, in their extract phase, they’re focused on squeezing as much revenue as they can out of the networks they own. Thus, corporate network owners not only suck up almost all network revenue, but they find ways to extract additional fees on top of that. Network participants get left high and dry. They spend years cultivating followings; then the rules change, and they are forced to pay even more to reach the audiences they built.

Big Tech’s high take rates are bad for network participants, but they’re great for their own profit margins. Meta has gross margins of over 70 percent,20 meaning that for every dollar in sales, it keeps more than 70 cents for itself (with the remainder paying costs related directly to revenue generation, like running data centers). Big Tech companies that own networks spend some of this windfall on fixed costs such as head count and software development. They realize the rest as profit. Inside these companies, thousands of employees work in management and sales, and some work on new R&D projects. But they also have layers of middle managers and wasteful bureaucracy enabled by the excess.

Where bean counters see fat margins, entrepreneurs should see blood. Your take rate is my opportunity, as Bezos might say.

Your Take Rate Is My Opportunity

Blockchain networks disrupt rent-seeking intermediaries, allowing them to take market share from extortionary corporations by lowering prices. Networks with a greater ability to lock in consumers have more pricing power. More pricing power translates into higher take rates. The higher the take rate of an incumbent network, the more opportunity there is for disruption.

Popular blockchain networks have very low take rates, ranging from below 1 percent to 2.5 percent. That means the rest of the money flowing through the networks goes to network participants, including users, developers, and creators. Compare the take rates21 of popular corporate networks with Ethereum and Uniswap, popular blockchain networks, and OpenSea, a marketplace built on top of blockchain networks:

^*Most popular fee tier.^** Calculated as total gas fees paid by users divided by total transfer value of ETH and top ERC20 tokens in 2022 (source: Coin Metrics)

Blockchain networks have low take rates because of rigid constraints set by their core design principles, namely:

• ▪ Code-enforced commitments. Blockchain networks commit to take rates up front at launch that cannot be changed except by the consent of the community. This forces networks to compete for network participants by offering commitments to lower take rates. In competitive markets, take rates will trend close to the cost of maintaining and developing the network.
• ▪ Community control. In well-designed blockchain networks, take rates can be increased only if the community votes to do so. This contrasts with corporate networks where the owner can raise take rates unilaterally, at the expense of the community.
• ▪ Open-source code. Because all blockchain code is open source, it is easy to “fork,” or create a copy of it. If a blockchain network raises take rates too high, a competitor can create a forked version with lower rates. The threat of forks helps keep take rates in check.
• ▪ Users own what they value. Well-designed blockchain networks interoperate with standard systems that guarantee that users own the things they care about. For example, many blockchain networks interoperate with the Ethereum Name Service (ENS), a popular naming system on the Ethereum blockchain. That means I can use my ENS name (cdixon.eth) across many different networks, and if the networks change the rules or raise the take rates, I can easily switch to a new network without losing my name or network connections. Lower switching costs mean reduced pricing power for networks, thus lower take rates.

One critique of blockchain networks22 is that their low take rates might be temporary: as blockchain networks proliferate, new intermediaries will pop up that raise the take rates, skeptics say. Moxie Marlinspike, a respected security researcher and founder of the Signal messaging app, wrote a widely read blog post arguing that because users are averse to even tiny user-interface frictions, they will end up clustering around easy-to-use front-end applications that siphon users away from blockchains. If these applications are run by companies, then we end up with the same problem we have today: a few companies with strong pricing power in control.

This is an insightful critique, sometimes known as the risk of recentralization. A similar dynamic undermined RSS, as discussed in “The Fall of RSS.” Twitter and other corporate networks siphoned users away from the protocol by offering lower-friction user experiences. This dynamic is also a risk for poorly designed blockchain networks.

Blockchain networks can avoid this fate if they can guarantee that users retain the credible threat of switching front-end clients even if users cluster around a few popular ones. To ensure this, the network must be designed to include the following:

• ▪ Low-friction user experiences that match those of modern corporate networks. This is why blockchain networks need a mechanism for funding much of what corporate networks fund, including ongoing software development and user subsidies like free hosting and name registration. Protocol networks never had a sufficient funding mechanism, a key reason RSS failed. (See “Building Networks with Token Incentives” for more on blockchain funding mechanisms.)
• ▪ Network effects that accrue to community-controlled blockchains rather than to company-controlled front-end applications. This means the things users care about—their names, social relationships, and digital goods—need to be blockchain-based and user owned. Applications are unable to acquire pricing leverage if users can easily switch from one application to another. When users own what matters, lock-in is much less likely.

Marlinspike cited the NFT marketplace OpenSea as an example of a corporate-owned application that could wrench control away from blockchain networks. But the blockchain networks that OpenSea interoperates with are well designed. When you sign up for OpenSea, you do so using a name that you own, tied to a blockchain like Ethereum. All the NFTs you own are also stored on a blockchain, not company servers. This makes it easy to switch to another marketplace while taking all the things you care about with you.

Marlinspike wrote his post in early 2022. Since then, new marketplaces like Blur have exploited the low switching costs of NFT platforms23 to take market share from OpenSea. In response, OpenSea lowered its take rates, demonstrating that blockchain-based ownership does, in practice, compel lower prices. In contrast, downward price competition is something you almost never see between corporate networks.

The low take rates of blockchain networks create strong incentives for developers and creators to build on top of them. For example, third-party startups add features and applications to DeFi networks without fear that they’ll regret having done so later. These startups know they can invest in and grow their businesses without the risk that the DeFi networks will change the rules, undermine them, and extract their profits later. Very few software developers are willing to become dependent on corporate financial networks like Square or PayPal. They may offer these services as one of multiple payment options, but they know better than to become reliant on them.

Blockchain networks should be designed to have take rates that are high enough to fund essential network activities but low enough to undercut corporate competitors. Blockchain networks offer a new model where far more of the economic surplus goes toward network participants and far less of it goes to bottom lines and bureaucratic bloat.

Squeezing the Balloon

To understand the tech industry, it is essential to understand that when one layer in a “tech stack” becomes commoditized, another layer becomes more profitable. A tech stack is, in this context, a set of technologies that work together to generate revenue. Think of the combination of a computer, an operating system, and software applications as a tech stack of layers built one on top of another.

When layers get commoditized, that means they lose their pricing leverage. In the physical world, this usually means competition is so fierce, and the resulting products so undifferentiated, that profits trend toward zero. Such is the case among actual commodities, like wheat or corn. In a tech stack, it’s more common for a layer to become commoditized when products and services are (1) given away for free, like the calculator app on an iPhone; (2) made open source, like the Linux operating system; or (3) controlled by a community, like the email protocol SMTP.

Clayton Christensen, whom we last heard from in the discussion of disruptive innovation in “Tokens,” generalized these ideas in his “law of conservation of attractive profits.”24 The theory is that commoditizing a layer in a tech stack is like squeezing a balloon. The volume of air stays constant but shifts to other areas. The same is true for profits in a tech stack (roughly, at least, since business isn’t as deterministic as physics). The overall profits are conserved but shift from layer to layer.

Let’s look at a concrete example. Google search makes money when a user clicks on a search ad. In between the advertiser paying and the user clicking, a stack of technologies intervenes: a device like a phone or PC, an operating system, a web browser, a telecom carrier, a search engine, an ad network. All these layers compete to capture a portion of each dollar that passes through the stack. The overall market can grow or shrink, but at any given time the competition between layers is zero-sum.

Google’s strategy with respect to search is to either own or commoditize the layers in the stack so that it can maximize its own revenue. Otherwise, a competitor that controls another layer could take away its profits. This is one reason why Google created products in each layer of the stack: devices (Pixel), operating systems (Android, mostly open source), browsers (Chrome plus the open-source Chromium project), and even carrier services (Google Fi). When a company like Google contributes to open-source projects or releases lower-priced versions of competing platforms’ products, it does not do so out of charity. It does so out of self-interest.

Here’s how this competition plays out on phones today. Because Apple controls the iPhone operating system and its default web browser, Safari, the company can charge Google25 a reported $12 billion per year for Google to remain the iPhone’s default search engine, and Google accepts this as the cost of doing business. Apple is using the popularity of the iPhone26 to squeeze Google’s search balloon. The payment would have been a lot higher if Google hadn’t had the foresight to build Android, giving it a big chunk of mobile market share. Google doesn’t even need to make money on Android. It just needs that part of the mobile market to be commoditized so it isn’t controlled by a competitor, like Apple, which could limit people’s access to Google’s search product. Thus, the fight for operating systems spills over to the fight for search profits.

By making Android open source (and bundled for free on many hardware makers’ phones), Google pursued a classic tech strategy known as “commoditize your complement.”27 Joel Spolsky, co-founder of Stack Overflow and Trello, coined the phrase in 2002,28 drawing on the work of economists such as Carl Shapiro and Google’s Hal Varian. Google commoditized a large share of the mobile operating system market, thus ensuring its search engine—its real moneymaker—could flourish, unimpeded, on a new computing platform. The move lessened Google’s platform risk in the industry-wide shift from PCs to mobile and improved its negotiating power, removing threats to its search profits.

Intel pursued a similar strategy by becoming the biggest code contributor to the open-source operating system Linux. Operating systems are complements to the processors Intel makes. When someone buys a Windows machine, Microsoft captures a share of profits that would otherwise have gone to Intel. When someone buys a Linux machine, more of that money goes instead to Intel. Intel supports Linux to commoditize operating systems, which complement its moneymaking processors.

Applying Christensen’s theory to social networks, one can think of the path money takes from users to creators, software developers, and other network participants as a tech stack. High take-rate corporate networks squeeze the balloon at both ends. They capture value at the center of the network, on behalf of the network owner, at the expense of complementary layers that build on top of the network, like creators and software developers. Network effects that result in lock-in force creators to work for free and developers to behave as they are told.

In the case of ad-supported media, the advertiser is the customer and source of the money flow, and users are a complementary layer to be squeezed. People give up their attention and personal data in exchange for network access. Protocol and blockchain networks, by contrast, have low take rates and therefore allow value to flow to users, creators, developers, and other network participants. They squeeze the balloon in the middle, to the benefit of the network edges.

In this sense, you can think of corporate networks as thick and protocol and blockchain networks as thin. Thick networks claim more profits for the center of the network and create thin complementary layers, with lower profits, for creators and software developers. Thin networks do the opposite, generating less profit for the network core and more profit for complements.

Let’s imagine you’re designing a social networking stack from scratch. Your goals might include some concept of fairness, such as that people deserve to earn money proportionate to how much value they create. You might also have societal goals in mind, like more uniform wealth distribution. But let’s just suppose, putting other concerns aside for a moment, that you merely want a network that encourages innovation and creativity. That means you will want social networks to be thin, the opposite of what we have today.

Think about it in terms of the infrastructure of a city, an analogy to which I’ll continually return. Roads should perform basic functions, but you don’t need them to be hotbeds of innovation. There isn’t that much creativity required; they just need to convey cars. On the other hand, you do want lots of creative entrepreneurs building around the roads: creating new shops and restaurants, constructing new buildings, expanding neighborhoods, and so forth. Roads should be thin, and their surroundings should be thick.

Social networks should be thin utilities, like roads. They need to support basic features and be reliable, performant, and interoperable. That’s about it. The rest of the features can be built around the network. The layers on top should be innovative, diverse, and thick. There should be boundless room for creativity in the media and software that complements social networks. (We’ll cover this at length in part 5, “What’s Next.”)

The web developed as a thin network—and look at the results. The network itself is a simple protocol (HTTP), and all the innovation happens on top, at the level of websites. This structure has led to a thirty-year run of explosive innovation across the internet.

Today’s corporate social networks are designed the opposite way, as thick networks. Almost all the value flows to the networks themselves—Facebook, TikTok, Twitter, and others. To the extent there is innovation, it involves startups trying to build competing social networks, rather than building businesses on top. In other words, startups have to build entirely new, proprietary roads in order to support new cities on top, instead of simply building on top of preexisting public roads. Social networks squeezed the balloon in a way that stifles innovation.

The same is true for modern financial networks. Payments should be an easy and inexpensive commodity, a basic utility, like sending email. We have the technology to do it, as we’ll cover in “Making Financial Infrastructure a Public Good.” This would make payments a thin layer in the finance and commerce stack. Today, it’s inverted: there are some very profitable payments companies, and the field remains an active area of entrepreneurship where startups and venture capital are lured by the industry’s persistent take rates. Again, the balloon got squeezed in the wrong places.

Blockchain networks are like a rubber band. They reshape the balloon, making the thick belly thin. DeFi makes payments, lending, and trading thin. The same is true for blockchain networks in areas like social networking, gaming, and media. The broader societal goal should be to build new tech stacks where users, creators, and entrepreneurs are not squeezed but rewarded.

Take rates are only half the economic equation for blockchain networks, though. The other half is token incentives that fund software development and other constructive activities. Tokens are a powerful tool that, like all tools, can be put to good or bad uses. Designed properly, they can make a network an attractive place to build a career or business. Achieving these design goals requires careful planning.

If take rates are the stick, token incentives are the carrot.