Original Article: Thin Applications —— Placeholder
Translator: Evelyn｜W3.Hitchhiker
When I started thinking about the architecture of blockchain applications in 2014, I described it as a layered "stack" of functionality. The first iteration described a blockchain-based "overlay network" with specific decentralized services on top, forming a "shared" protocol and data layer. On top of them, independent applications would consume these protocols and redistribute their services to users.
Blockchain Application Stack (2014)
I predicted that this architecture would dominate new online services within a decade as cryptocurrencies took over the web. So it’s interesting to look back at how this idea has evolved over the past five years. The most obvious mistake was the idea that we would build everything on top of Bitcoin (Ethereum didn’t exist yet). Now we have a plethora of blockchains to choose from, which is much better than when we started. We also called “overlay networks” Layer 2. Today, “Web3 application stack” would be a better name. The overall framework seems to hold true.
Two years later, this model brought me to the controversial "fat protocols". My suggestion is that in the future the "protocol layer" will capture most of the value of the crypto market, while on the network it will be captured by the "application layer".
Fat Protocols (2016)
This observation is informed by the evolution of the application stack. Most of the "work" and data exists at the protocol layer, while applications tend to provide more limited interface services. But in 2014, the "business model" of the protocol layer was not obvious. But as we at USV invested in the early crypto space, the potential of tokens became clearer.
In 2014 and 2016, there weren’t a lot of examples to observe in the real world. And in the grand scheme of IT history, it was still “early days”. But now we can observe hundreds of cryptographic protocols and applications in many markets. Going into 2020, after Placeholder has tested different aspects of these ideas, it’s a good time to refine old ideas and consolidate what we’ve learned so far.
Encryption service architecture
Large web companies tend to scale their platforms and monopolize information by locking users into their own user interfaces. Crypto networks, on the other hand, tend to offer a single service and cannot "own" the UI because they don't control the data. Specialization does help in this regard, because the more decentralized the web is, the harder it is to coordinate a whole suite of services under a single UI like Google, Facebook, or Amazon can. So instead, crypto/Web3 consumer applications are independently built on top of multiple "composable" protocols in what we can call a crypto-services architecture (it's a lot like microservices, but it's made up of sovereign components).
An encryption service architecture
In DeFi, people call it “money legos.” Consider Zerion (a Placeholder investment), Instadapp, and Multis, who are using many of the same protocols to build similar crypto-financial applications like Ethereum, Compound, Maker, and Uniswap. This allows them to offer a full suite of financial services (transactions, borrowing, lending, trading, investing, etc.) without having to build all of these functions, infrastructure, and liquidity in-house. These protocols offer specific services over many interfaces, and applications built on top of them share resources and data without the risk of a centralized platform, while shared infrastructure reduces overall costs. And these same dynamics are also showing up in corners of crypto, like DAOs and games.
Crypto-as-a-service architecture is great for startups. Entrepreneurs can launch new applications quickly and cheaply by outsourcing large chunks of functionality to various networks. And every application is equal in terms of protocol costs and resources (unlike network infrastructure like AWS, where the smaller you are, the more expensive it is). The above companies stand out because they brought fully functional products to market before their first real round of funding. This is an initial glimpse into the level of capital efficiency that "thin" applications using this new model can achieve, compared to the increasing amounts of money that network companies need to raise to compete with existing companies.
Bring your own data
Non-custodial is another way cryptocurrencies cut costs for adoption. The business model of large networks depends on creating data monopolies, because locking users into proprietary interfaces is the most advantageous way to extract value from their information. They also compete more than they cooperate, so as a user you have to visit different platforms to get your pieces of information. And, increasing security costs and new regulations turn data into a liability. This benefits companies that can afford these costs. But startups with limited resources have to find other models to compete.
As a cryptocurrency user, you bring your own data. No one has a monopoly on control. When you log into a crypto app by connecting your wallet, you are sharing the "keys" needed to access this data in order to find your information in the relevant network. You can share them with any application, so when you move from one interface to another, your data moves with you. Moreover, you control the "private key" (basically a password) required to be able to act on that data, like signing messages or authorizing transactions. Therefore, you have effective custody of your data and no one can manipulate it without consent (unless you entrust your keys to a custodian).
For example, in the crypto-collectibles space, artists can tokenize their work using Rare Art and sell it on OpenSea. Then someone (like Jake) can buy it and display it in a virtual gallery on Cryptovoxels. Similarly, in DeFi, if you use Maker through Zerion, and then log into Instadapp with the same key, you can instantly interact with your Maker loan there, too. By building on the same standards and networks, these applications are interoperable by default, and users can move freely between interfaces without losing information or functionality.
Crypto-service architectures combined with non-custodial data models allow startups to compete more effectively with centralized incumbents. This echoes the way open standards drive IT market cycles. Putting ownership and control in the hands of users unloads a lot of costs while meeting many of today’s consumer needs. It does require companies to give up a lot of what makes traditional online services work. But what you lose in control, you regain in potential efficiency and scale. Adopting these models allows businesses to run at very low costs, and applications benefit from each other’s success because they contribute to a shared pool of resources at the protocol level. As a network, thin applications can scale more efficiently across markets. Every digital artwork minted on Rare Art indirectly increases the utility value of OpenSea, activity on Instadapp benefits Zerion (and vice versa), and so on. But what seems unclear to many is how exactly they can create long-term business value and defensibility when everything is open.
Value Capture vs. Return on Investment
Fat Protocols suggests that crypto protocols “capture” more value than application interfaces. However, a common mistake is to conflate the idea of value capture with investment returns (similarly, it is a mistake to think that a token on Ethereum or other smart contract blockchains is itself an “application” — more often than not, they represent the value of other protocols with their own application layer, rather than Ethereum’s application interface). Many people conclude that investing in the application layer of cryptocurrencies is a non-return, even though the original article qualifies the success of the application layer as a requirement for the value of the protocol to grow. To be clear, less overall value ending up in the application layer does not mean that there are fewer opportunities for outsized returns that can be earned in application-type businesses. It also does not mean that there are always returns in protocols. Value capture is more about TAM and other macro elements, while returns vary based on factors such as cost base, growth rate, and ownership concentration. The difference between protocols and applications is how these elements fit together.
Looking at value from a cost perspective is a more precise way to think about the distribution of value. The basic principle is that in markets, cost is a strong determinant of future value. Therefore, we can estimate a market’s value structure by studying its cost structure. In crypto, the protocol layer networks bear most of the production costs, so they require more investment — which means more value must be accrued to that layer to maintain equilibrium (otherwise the investment wouldn’t happen). Applications have lower operating costs and require less investment, so they naturally have lower claims on market value. However, the ownership and cost structure of networks is far more distributed than that of private businesses. Generally speaking, investing in tokens will usually give you a smaller piece of a very large pie to cover your cost of capital.
For example, buying $10 million of ETH when the Ethereum network is valued at ~$15 billion would give you only ~0.06% of the network’s market cap (based on current supply), and would require a net increase of $60 billion in network value to provide a 5x return (where your ETH holdings would need to be worth $50 million). Meanwhile, a $1 million seed investment in a successful application business for a 10% stake in the company would give you $50 million in the same stake if the business grew to a market cap of $490 million — which would be less than $10 million if subsequent investments were taken into account.
But networks and businesses handle value very differently. The forces that drive a public network from $20 to $90 billion are very different than the forces that drive a business from $10 to $500 million. Token prices are chaotically determined every time a transaction takes place in public markets, where networks gain and lose value per dollar invested faster than private businesses. This complexity adds a lot of leverage to every investment dollar that flows in and out. At the same time, business value may be a well-known function, but private, early-stage investments can be riskier in unpredictable ways.
Finally, we must also consider the combined value of all protocols underlying an application to assess relative value. For example, Zerion relies on Ethereum, Maker, Compound, Uniswap, and other protocols to operate. The combined value of these networks is far greater than the individual value of Zerion or its peers. But again, this has little to do with the return on investment of applications and businesses using these protocols. Crypto networks may scale to store trillions of value, but, eventually, growth will flatten out. Then, the majority of the market's value may be stored in the protocol layer, and excess investment returns will move to where there is more growth. But today we are far from this equilibrium state, even though we find high-return opportunities in both layers.
P2B2C
Thin applications are cheaper to run because they push many costs onto protocols and users. But competitors have access to the same production and data resources, so they can substitute for each other in ways that are not possible on the traditional network. In a way, it is similar to the retail model, where storefronts serve as "interfaces" for various goods, differentiating themselves through branding, curation, and customer experience. However, rather than "B2B2C", it is more "P2B2C", from protocol to enterprise to consumer.
Protocols provide specific services, which are bundled and distributed to consumers at the application layer. As in retail, prices are set by the crypto networks that produce the services (a la MSRP), and fair competition at the application layer makes it difficult for anyone to unfairly jack up their own prices. This setup is great for users and solves many of our complaints about the web. But it raises new questions about the defensibility of the application layer. How do you create long-term business value and defensibility when everything is open and competitors can easily substitute for each other?
In crypto, application businesses must create value beyond the functionality of the protocol. In many cases, known business models like subscriptions or transaction fees make sense. However, as infrastructure matures and applications become thinner, new business models are needed. There are a lot of interesting experiments here. For example, Blockstack is innovating with “application mining” and NEAR offers a royalty structure to its developers (both are reminiscent of Amy James’s salutary protocols idea). I’m curious to see how they develop, although I’m not yet fully convinced that protocols should dictate the economics of their applications. It would take more words to talk about all the experiments, so here, I’ll focus on three general strategies: building cost moats, vertical integration, and user-staker models.
Building a cost moat means concentrating unaccounted-for costs and externalities in the protocol. The scale of these costs is inherently defensible because it makes it too expensive for competitors to catch up. For example, Coinbase creates a lot of business value by capturing two very expensive externalities of crypto that users are willing to outsource (i.e., fiat trading and custody) and through classic economies of scale. But there is nothing new. The market will not let Coinbase raise trading fees on cryptocurrencies, but they will use trading fees to recoup the large investments they make to provide these services. In contrast, “thinner” applications like Zerion do not internalize these costs, so they do not charge additional fees - but as a result, they cannot use Coinbase’s business model and cannot justify charging the same fees. This approach is effective, but it is also expensive.
Vertical integration in crypto explores the possibility that successful applications might amass enough users to “become their own supply(side).” They could do this by turning themselves into the “supply side” (like miners) in the protocols they integrate into, and directly servicing their users. We saw this happen in the old retail model, and it’s happening again as Amazon promotes its own products over those of its competitors. Amazon takes little to no profit on items sold on its storefront, and then uses the platform and its perfect demand data to create its own supply with unprecedented efficiency. Could a cryptocurrency application make a similar move? What if the application later forked the network? Would the market allow that? It’s undesirable for the application layer to have too much control over the protocol. That’s what happened on the Web. But it’s a possible outcome.
Finally, the idea of user staking is to use tokens to distribute value and benefits to users. Generally, the way it works is to have users stake a certain amount of the application’s own tokens (not the protocol’s tokens) to unlock benefits like discounts or rewards (but there are many variations). At first glance, they are similar to the loyalty/rewards systems used to retain customers in other hyper-commoditized markets like airlines and credit cards. Except these projects don’t offer any benefits at all. The innovation is in designing token models that allow users to profit from the growth of the application. It goes beyond marginal benefits like discounts and includes users in the upside of the business.
For example, Nexo and Celsius, which offer crypto-collateralized loans, use tokens in this way. Nexo offers discounted interest rates when you repay your loan with NEXO tokens. At Celsius, the more CEL you hold, the better interest rates you unlock, and if you choose to earn your deposit in CEL, you can get better interest rates based on the size of your deposit. Since the supply of NEXO and CEL tokens is limited, as these applications grow and more people buy and use these tokens, then they are likely to appreciate in value. So there is rising value beyond a simple discount. We've even seen this model with SAAS businesses like Blox, who will give you a discount on monthly fees if you stake their tokens. So how do we take this concept further into the mainstream?
The user staking models are what I’m most attracted to because they represent true business model innovation. The examples above are more like traditional web apps. They are more centralized and regulated than “thinner” apps like Zerion. But what I like about their staking models is that they change the relationship users have with services. Web users are forcibly locked in through the centralization of their data. Crypto apps, even if they are built in a more traditional way, don’t have the same ability to lock you in. However, user staking creates an “opt-in” economic lock-in that benefits users by turning them into stakeholders in the success of the service. It creates defensibility through user ownership rather than user lock-in. This leads to a range of fascinating consequences that will be explored in future work.
