Using the bus factor to spot single points of failure in crypto

One of my favorite mental models for evaluating cryptocurrency is the bus factor.

The bus factor is a macabre—but catchy—way of evaluating how many of a project’s contributors have to be hit by a bus until that project starts to stall. Of course, the traffic accident is just a metaphor for all the ways contributors can disappear in practice.

Take, for example, one of the most famous trade secrets of all: the secret formula of Coca-Cola. If only three employees knew it, the company’s bus factor would be at most three. It could be lower still if another part of their business hinged on even fewer people. But if these three secret-keepers ever went on vacation together and their plane crashed, there would be big trouble in Coca-Cola land.

This is pretty close to how it works in practice, by the way:

“While several recipes, each purporting to be the authentic formula, have been published, the company maintains that the actual formula remains a secret, known only to a very few select (and anonymous) employees.”

Projects with a low bus number are vulnerable to an unlikely but devastating event, or—more likely—a chain of events. When the number is high, projects can survive through very adverse conditions, as disappearing members do not leave irreparable holes in the organization.

The optimal bus factor

According to Wiki’s definition, the bus factor is primarily used to evaluate software projects. Their explanation is that key technical experts can be harder to replace than the more generalist roles of higher management. In my opinion, you can use the method in all areas of an organization to identify single points of failure.

The first thing to note about single points of failure is they are often a double-edged sword. If they fail, you’re in trouble. But while they don’t, they often represent valuable assets that cannot be replicated by the competition, leading to a valuable moat in business. So the optimal bus factor is not immediately apparent.

If you’re a Coca-Cola manager and were spooked by my previous description, you might be compelled to share the trade secret with more than a few employees. But by doing that you increase the risk that the recipe eventually leaks – thereby reducing its asset value as well.

Luminaries like Elon musk can also bring a bus factor of one to any project they work on, but that doesn’t mean other companies wouldn’t love to have them. As a result, many of the most —and least—valuable projects will have a low bus factor.

Still, projects should strive to increase their bus factor wherever they can. Kailash Awati explains that the key to “increasing the bus factor amounts to spreading the knowledge around so that there is a degree of redundancy in skills and knowledge within a team.” His suggestions are to (1) keep the complexity of processes low, (2) maintain excellent documentation, and (3) actively encourage cross-training.

The bus factor in cryptocurrency

There’s hardly a better application for the bus factor than cryptocurrency because crypto is all about building systems without single points of failure. And superficially, they seem to succeed at that. The average blue-chip cryptocurrency probably has a higher bus factor than a non-crypto system of similar sizes.

These benefits mainly come from the open-source culture that by default encourages Awati’s best practices: there are good introductions to contributing to bitcoin (like here, here, or here); strong social covenant around communicating in public channels (mainly IRC and the developer mailing lists these days); finally, there is plenty of documentation to study the system and its history scattered across different sources like GitHub, the Bitcoin Developer Guide, or StackExchange.

But despite the open-source culture, as an investor, you should be aware of at least five areas of key man risk that frequently appear in crypto projects. There might be more, these are just the ones that I’ve noticed.

Vision

Many crypto projects have a visionary leader or prophet-like figure at the helm.
Bitcoin (and arguably crypto as an asset class) has Satoshi Nakamoto, Ethereum has Vitalik Buterin, Litecoin has Charlie Lee, Bitcoin Cash has Roger Ver, Tron has Justin Sun, and so on.

Some of them are technical leaders as well, others are more responsible for marketing and/or funding (two roles that we will dissect a bit later.) In this section, I want to focus on their role as Schelling points for the community.

Every (crypto)currency has value because a group of users believes in a shared story—though not necessarily in Harari’s sense of the word. That doesn’t mean money is not real and different monies can’t have properties that make them more or less suited for the job.

But a cryptocurrency can be forked an infinite number of times, and the desirable properties are an emergent property of a large number of people converging on the same fork. Having a shared story helps tremendously with that.

The visionary or prophet is the main storyteller of that shared fiction. The goal for all cryptocurrencies should be to eventually emancipate from their leader, but if that exit is too early or too sudden, there can be a strong market reaction.

In June 2017, a post went viral on social media claiming Vitalik Buterin had died in a car accident. While the post turned out to be a hoax, the value of ether had dropped almost 15% before Vitalik stepped in with a “proof of life”, demonstrating how crucial investors thought he was for the project’s success at the time.

Image Source: Twitter

In contrast, the biggest risk from Bitcoin’s prophetic figure is not Satoshi’s disappearance, but ironically his return (either with conflicting messaging and/or by selling his presumed stash of early bitcoin.)

While Ethereum has arguably grown beyond the stage where Vitalik was seen as irreplaceable for the success of the project, many other coins still have significant key man risk in this area.

Evangelism

To understand the special kind of role evangelism plays in the success of a cryptocurrency, it is important to understand how investors price them. The pricing of every money is effectively a Keynesian beauty contest. Fundamental value plays some role in the overall valuation, but bitcoin isn’t worth $140b based on its current adoption as a payments (or savings) network.

Investors ascribe it a massive premium because they expect adoption, and hence fundamental value, to grow and become much larger in the future. These expectations—and the expectations of everyone else’s expectation—can fluctuate dramatically, giving crypto its famous volatility.

For most projects, adoption does not mainly depend on technical progress but on awareness and education through evangelism. Most attributes of Good Money are emergent (e.g. a large network of people to pay; liquid exchanges; a strong social commitment to transaction finality; a strong social commitment to a certain monetary policy). As a result, projects improve as more people are convinced to join the community and are dedicated to the project’s core values.

If the visionary sets the overall trajectory of a project, the evangelists are the ones who carry the word out into the world and actively convince new people to join. Roger Ver and Justin Sun are examples of this “chief evangelist” type. Vitalik Buterin, too, spent the first years of Ethereum traveling and educating developers about the project.

In some cases, the value of a coin depends primarily on the evangelism effort on a small number of people or entities. When that is the case, you also have a very low bus factor. (Good catch if that reminds you of the Howey Test. Having a single dominant evangelist is also a risk when it comes to U.S. securities laws.)

Funding

A topic that receives a lot of attention recently is that of developer funding (e.g. here or our own analysis here). The gist is that OSS infrastructure (and that includes cryptocurrency protocols) must be actively maintained in the same way we maintain physical infrastructures like buildings, roads, or bridges. But OSS development is challenging to fund because the software is, well, free. The number of independent entities that contribute to developer funding is another good application for the bus factor.

If that number gets too small, the remaining entities increasingly gain leverage over who gets to work on the protocol full-time and who doesn’t; if they are employed by for-profit companies, developers can weigh the interests of their employer over the long-term benefit of the protocol (this is the default for any coin controlled by a single company, e.g. Ripple); finally, if the number of code reviewers drops too low, an attacker could manage to inject malicious code changes, or genuine bugs are simply not found before they become a problem. If the number of funding sources drops to zero, the project effectively returns to hobbyist status.

R&D

A project’s key man risk increases the more it relies on arcane technical knowledge, e.g. in the field of cryptography. Only a handful of people can understand, let alone maintain, complex cryptographic proof systems like the Bulletproofs of Monero or the zero-knowledge proofs of Zcash.

Cryptocurrency projects wildly diverge when it comes to technical risk. While bitcoin has low technical risk for the use case as “digital gold” today, at the other end of the spectrum projects like Filecoin or Dfinity are effectively still at the drawing board. Ethereum represents a weird blend between the two extremes, with ETH1 working quite well— apart from problems like state bloat—and ETH2 (and previously its transition to proof-of-stake) sitting in the research stage for years.

I think investors should strongly discount projects in “research limbo”. For problems that don’t provably have a solution, every day spent in limbo decreases—rather than increases—the chance that a solution even exists, and the project can eventually realize its promises.

Another aspect of the bus factor is not just the leverage of “fate” over key personnel, but also the leverage of key personnel over users or investors of a given protocol. Although the founder’s reward of the Zcash protocol was supposed to expire later this year, few industry insiders were surprised when the Electric Coin Co. changed the monetary policy to extend their salary for at least another four years.

This happened with consent from the community, but one has to ask how much consent is even possible when the alternative is for the protocol to lose its developer team. The Zcash team has this leverage because their research is so cutting-edge that the space of independent developers to replace them is effectively empty.

Block production

Finally, we can talk about the consensus process (though this is less a case of “someone can get hit by a bus” and more “someone can seriously harm investors”.) Cryptocurrencies propose a protocol for participants in the network to follow. Some of the rules can be enforced with cryptography but for others—like the time ordering of transactions—there is no way to force participants to follow the desired protocol. They can deviate if they want.

Cryptocurrencies tackle that problem in two different ways. In a permissioned network, a gatekeeper checks at the door that participants are good guys who will always follow the protocol. But the existence of the door and the gatekeeper can become a single point of failure.

Instead, networks like Bitcoin or Ethereum use a permissionless approach. (1) Permissionless networks don’t have a door, and anyone can participate in block production. (2) Hashpower distribution can get pretty centralized as well, but that’s not the important part.

The main difference between permissioned and permissionless systems when it comes to being a single point of failure is that a permissionless system includes a built-in, gatekeeper-free mechanism for replacing hostile actors, while a permissioned system does not.

Therefore, in a permissioned system, a community must defer to some other Schelling point for guidance to replace a hostile actor (this could be any of the Schelling points described above, a visionary, a developer, a funder, etc.) Meanwhile in a permissionless system, there are mechanisms in place to (potentially!) resolve the issue without deferring to someone else.

As a result, power is typically more concentrated in a permissioned setting and can lead to a lower bus factor.

Takeaways for investors

There are some important takeaways from applying the bus factor to cryptocurrency systems. The bus factor is a tool to identify single points of failure. Cryptosystems have value when they have none.

As an investor, you should be aware that, despite the smokescreen of decentralization, single points of failure can, and do, exist in various parts of these projects. These single points of failure represent tail risk that should be considered for a holistic valuation.

In some ways, the risks are more distinguished in crypto than in traditional companies, because the conditions are more adverse. The average company has fewer enemies than the average cryptocurrency project, and there are fewer ways to profit financially from its demise. (For example, Ford and Böhme argue that the existence of sufficiently deep derivatives markets can break the incentives of any permissionless system.) As a result, single points of failure can be triggered not just by strokes of bad luck, but also actively by hostile parties.

(1) Some people would argue that permissioned networks are by definition not cryptocurrencies, and I tend to agree. But there is still a fair number of permissioned networks in the crypto space, so it makes sense to talk about them.
(2) Now, you need to solve two brand-new problems: (1) who gets to propose the next block, (2) how to get that guy to follow the protocol. The first is solved with proofs of cost (e.g. PoW or PoS), the second is done by rewarding good behavior with tokens.