Aggregation Theory, Thin Protocols, and Recentralization: Augur Edition

Joel Monegro of PlaceholderVC wrote Fat Protocols in 2016. Based on the data at the time – specifically that Union Square Ventures would have generated higher returns by simply buying and holding Bitcoin rather than investing in each of Coinbase’s funding rounds – the fat protocols theory made sense.

Using the thought experiment presented below, I’m not going to make a general argument against the fat protocol thesis. I believe with extremely high conviction that a global digital reserve asset will be worth tens of trillions of dollars. That asset is going to be the native token of some adversarial cryptographically bound network.

Instead, in this essay I’ll propose a thought experiment in which a dapp steals value from the protocol that its built on. This is categorically separate from Joel’s fat protocols thesis: it doesn’t make sense to argue that Coinbase somehow steals value from Bitcoin or Ethereum.

I’ll assume the reader has a practical understanding of what Augur is, how its native REP tokens are used to report outcomes, that REP tokens should be valued as a function of cash flows using a DCF, and that there are separate fees for market creation and payouts to REP reporters. For more background on Augur, please see the Augur research paper that Multicoin Capital published in August 2017 (note: the Augur team has changed some of the REP payout mechanics since we published that paper; most everything else is still accurate).

The Opportunity

We can generally say that Augur is comprised of two parts: the protocol (the smart contracts on the Ethereum blockchain) and the end-user application. Anyone can hook up any user experience (UX) to the Augur protocol. There is no way for the Augur Foundation or any other group to stop a third party from doing this. This is by design.

Augur launched on July 9th. It is the most technically complex project to be deployed on top of Ethereum. Although it works as advertised, it’s clunky and expensive. It doesn’t run as a seamless web application, it loads slowly, the UX is janky, there are no push or email notifications, settlements take a long time, there is no mobile support, etc.

Why doesn’t the Augur Foundation fix these issues? While they can address some of these, the Augur Foundation will never fix others, most notably anything that involves push or email notifications, or anything that censors content (e.g., assassination markets). Some of these functions require a degree of centralization.

But a company that builds on the Augur protocol can solve all of these problems. For the rest of this thought experiment, I’ll refer to any such hypothetical company as AugurCo.

If there aren’t already people building AugurCo, I’m sure there will be very soon. The opportunity set is clear. I suspect that within a few months, AugurCo will offer a generally passable UX, and that within 18 months, AugurCo will deliver a UX that rivals modern consumer experiences across desktop and mobile.

Fat App, Thin Protocol

For the purposes of this thought experiment, aggregation theory can be simplified thus: Owning the end-customer relationship in an internet-based value chain affords the company that owns the customer relationship power over its modularized and commoditized suppliers.

The Augur protocol is not neutral. While market creators set market creation fees, the protocol itself sets the reporting fee so as to target a REP market cap that is 7.5 times the value of the open interest across all of Augur’s markets. For simplicity, let’s say that the reporting fee is 75% of total fees generated, and that market creation fees account for the remaining 25%.

AugurCo controls the UX. Therefore, AugurCo can choose which markets to show in its UI. Naturally, AugurCo will create markets itself, and not show markets created by third parties. Therefore, AugurCo will receive 25% of the total revenues of the markets it creates. Reporting REP holders will receive the other 75%.

Absent highly concentrated REP ownership, AugurCo is going to generate much more income than any individual REP holder.

AugurCo will have material fixed costs. But AugurCo’s gross margins are likely to be on the order of 70-80%, in line with most software companies. At scale, a number of AugurCos can collectively generate upwards of 20% of the income produced by the Augur protocol.

Or can AugurCos generate even more profits?

Let’s fast forward a few years and make a few assumptions. Let’s say there are 7 major AugurCos. Each has carved out its own niche in the prediction market ecosystem. Perhaps the niches are segmented based on market (sports betting, financial derivatives, politics), or based on geography. For the purposes of this thought experiment, it doesn’t matter.

So there are 7 AugurCos, each of whom controls ~14% of global Augur volumes, global volumes flowing through the Augur protocol are $20B, and the market is relatively stable.

Let’s consider some options that each of the AugurCos has:

1. Fork the Augur protocol, ignore the existing REP distribution entirely, hand out newREP tokens to your best customers, and keep some percentage of newREP tokens for yourself.
2. Fork the Augur protocol, hand out newREP tokens honoring the existing REP distribution, and adjust the fee model such that AugurCo collects 50% of revenues, instead of 25%.

In either case, so long as AugurCo has sufficient scale and ability to resist censorship, the only rational thing to do is to fork the Augur protocol.

How can we quantify sufficient scale and ability to resist censorship?

Assuming $20B of global annual volume, 1% aggregate fees, and 14% market share per AugurCo, each AugurCo will generate $7.1M of revenue. All AugurCos together will generate $50M of revenue. REP holders will generate $21.3M of revenue per AugurCo, and $150M across all AugurCos.

As an individual AugurCo, the question of sufficient scale boils down to this: is $21.3M of payouts to outcome reports enough to keep the system safe? Or will the system need $150M on an absolute basis? Only time will tell.

Going Full Circle: Centralized Prediction Markets

If an AugurCo is going to fork Augur, why even bother with decentralization at all? Why leave 25% or 50% of revenues for a decentralized network of reporters? Why doesn’t AugurCo just recentralize the whole thing and collect 100% of the fees for itself?

I suspect some AugurCos will try this at some point. The opportunity to grow revenues 100-300% overnight will be too tempting for some to resist. If they pull it off, then it also implies that decentralized prediction markets largely don’t matter, so long as fees are reasonable (Las Vegas bookies take on the order of 10%; Predictit.org takes between 5-10%. These fees are not reasonable).

There may be a new class of predictions that require a decentralized prediction market – such as the one discussed here – but it seems likely that these P2P prediction markets will pale in volume relative to derivatives, sports betting, or politics.

0x Relayers And The ZRX Token

Given the high level analog of Augur:AugurCo and 0x:relayer, let’s quickly consider the fat app, thin protocol hypothesis for 0x relayers and the ZRX token.

The Augur system must have some fees because reporters must be compensated. There is simply no way around this. Because the incentive model of Augur depends on reporters being a large group of unaffiliated, rational, profit-seeking individuals and companies, reporters will not run reporting operations as loss leaders for some other business line. Reporters are by design not businesses.

On the other hand, the 0x protocol is neutral. It does not – and should not – impose any fees for trades facilitated by the 0x protocol. Relayers compete to own the UX of trading using the 0x protocol. They may generate revenue by taking fees using the native ZRX token as it was intended, or their core trading operations may serve as a loss-leading function for some other revenue line. The bases of competition that will ultimately dictate which relayer(s) will win is still unclear.

Unlike the Augur protocol – which must pay fees to reporters in order for the protocol to function – the 0x protocol doesn’t generate any fees, and therefore even a very successful 0x relayer cannot double or triple revenue overnight by forking the 0x protocol. Therefore, there is not a material incentive for a 0x relayer to fork the 0x protocol in the same way there is for an AugurCo to fork the Augur protocol to double or triple revenues overnight.

Conclusion

While it’s been technically feasible to modularize the front end and back of an app for a long time, there’s never been a strong economic incentive to do so. Crypto changes that.

On one hand, this division will be mutually reinforcing and net value-accretive: companies will rely on protocols for certain pieces of back end logic, and the protocols will need the companies to make the protocol usable and bring the protocol to market. On the other hand, the companies building on top of the protocol will perpetually consider how they can steal value from the underlying protocol itself. In most instances, the protocol will be unable to defend itself.

Thanks to Denis Nazarov for providing feedback on this post.