Coinbase's exchange made $1B in 2017. It might have to make less to succeed.

reach out on twitter @ethanruhe

TL;DR:

• Collecting fees on trading volume is Coinbase’s source of revenue.
• Coinbase’s YoY trading volume grew >50x in 2017, to $80B.
• Reports suggest Coinbase did $1B in revenue last year, implying a 1.25% effective mean fee rate.
• They starkly segment fees: ~2.5% on the app, ~0.25% on “Pro” trading platform.
• Retail competition at 0 fee is emerging.
• Other pro trading platforms generally charge significantly less than Coinbase.
• Large financial exchange comparison: NASDAQ exchanged $14.8T in H1 2018 for $0.5B in fees (Q1, Q2 data), an effective fee rate of 0.0035% (_i.e.,_Coinbase’s rate is ~350x higher).
• Volume is positively correlated with prices; both have been falling since the start of 2018.
• In the first half of 2018, Coinbase did $77B in volume. But, they only did $4.7B in June, 18.6% of January’s volume.

Conclusions:

• In 2017, volume was split about 50/50 between retail (app +website) trades and pro platform trades.
• If Coinbase maintained the same retail/pro split, they booked about another $1B in revenue in the first half of 2018.
• Current fee rates are unsustainable in the long run, and competition will drive them down.
• Exchanges simply don’t support big-tech-scale businesses; put another way, Apple makes more setting the default search engine in Safari than NASDAQ does with its entire exchange business.
• Success of Coinbase’s ambition will depend on building an ecosystem of complimentary products that achieve scale, not exchange fees (CEO Brian Armstrong lays out their broad 4 step strategy here; their product news is here).
• Coinbase seems to be in a unique position: 10x growth may depend on their current core business shrinking (i.e., exchange with less friction and lower fees).
• The gravity of Coinbase’s talent pool + sober focus on long-term problems + a mountain of cash all mean that Coinbase is well positioned, but its breakout path still isn’t clear (like Facebook before mobile, Instagram, and WhatsApp).

Code for pulling Coinbase data and replicating my analysis is available here. If this is interesting to you, let me know, and I can do a deeper dive on what the data tell us.

A longer version follows below…

Why is it interesting?

• The strong case for cryptocurrencies as the biggest platform/protocol shift in a long time.
• Coinbase could become a juggernaut, but their current revenue model isn’t what will get them there.
• The company is unusually long-term focused (reflected in thoughts like this and the fact that they give employees a 7 year exercise window on options).

Coinbase’s revenue growth is rumored to be almost historically fast. The widely-reported rumor is that they did $1 billion in revenue last year.(Allegedly the bulk of that came during the price surge in the waning months of 2017.)

With this data point in mind, we can estimate what proportion of Coinbase’s revenue comes from casual consumer speculators and how much is from their more sophisticated professional trading platform. We can then estimate how 2018 revenue is shaping up for the company.

This analysis is possible because data are unusually accessible for a private company: running an exchange, Coinbase’s near-exclusive revenue source to-date, requires publishing price and volume data. These data are key drivers of revenue and therefore give a unique view into the growth of the private company.

Coinbase has never aimed to be only an exchange. Brian Armstrong, the CEO, lays out their broad strategy here. They’ve been aggressively launching and buying products, especially in the past year. In order to be a ≥$100 billion company, Coinbase will need both successful new product bets and for institutional cryptocurrency adoption to happen sooner rather than later. The closest analogy to what Coinbase has to execute over the next few years is Facebook circa 2012–2014: a huge platform shift at a favorable time (consumer adoption of mobile in Facebook’s case, institutional adoption of cryptocurrencies in Coinbase’s), one or more successful bet-the-company moves to capitalize on the opportunity, and possibly homerun product acquisitions (_e.g.,_Instagram and WhatsApp).

How does Coinbase make money?

On an exchange like Coinbase’s, trading involves matching buyers and sellers. Coinbase takes a cut of each transaction.

Limit orders involve a trader offering to buy or sell as much as a set amount of an asset at a specific unit price (e.g., “I’ll buy as many as 10 bitcoins at a price of $5,000 per bitcoin,” or “I’ll sell as many as 2.5 bitcoins at a price of $5,010 per bitcoin”). Limit orders underlie the market and “give liquidity.” Those who place limit orders are called price makers since they literally name their price. To date, limit orders have been free on Coinbase in order to encourage liquidity growth.

A market order involves specifying a buy or sell against outstanding limit orders. In a market order, a trader specifies an amount of one asset to be exchanged, on the best possible terms, with outstanding limit orders on the other side. For example, “sell 100 bitcoins” results in selling 100 bitcoins to limit buy orders in descending order of price. Depending on the size of the market order, it may fulfil as little as part of one limit order or as much as multiple limit orders.

The fees Coinbase charges for market orders are starkly segmented across what they refer to as retail traders (those who use the mobile app and coinbase.com) and professional traders (those who use pro.coinbase.com, which was formerly called GDAX). Anyone can perform retail trades with the app or on Coinbase.com. To date, the pro platform has also been easy to join. It just has a slightly more complicated interface. (Note: in May 2018 Coinbase launched an institutional trading platform called Prime, but this analysis will not attempt to separate that from Pro as there is no benchmark for how to do that cleanly.)

Coinbase Fees

Retail trades (mobile app and coinbase.com) are subject to a fairly complex fee schedule that can vary with Coinbase’s discretion depending on market conditions when executed (documentation is here). For use in the model, we’ll simplify it to the following:

• Conversion fee of 1.49%: For every $100 in retail cryptocurrency transaction, Coinbase takes $1.49 for converting between fiat and cryptocurrency.
• Exchange fee of 0–2%: Coinbase says “Your exchange rate for buying or selling digital currency through our Conversion Service is calculated as the market rate of the digital currency on Coinbase Markets platform, plus a spread between 0 to 200 basis points determined by the size of your transaction, market volatility and length of time using Coinbase.” So, in addition to the conversion fee, they take a 0–2% exchange spread relative to their underlying exchange price. This is analogous to the “taker fee” we’ll see below. At the time of this writing, I initiated a $3,865 bitcoin purchase when the price was $7,730 and was quoted a purchase volume that implied a 50 basis point, or 0.50%, exchange fee. I have a relatively long track record on the platform and short-term volatility was relatively low, so I’d guess this is on the lower end of the spectrum. For the sake of simplicity, we’ll assume a 100 basis point, or 1.00% mean exchange fee for retail orders.
• Increased conversion fees if using a credit or debit card. In general, this is a 3.99% conversion fee instead of 1.49%. For the sake of simplicity, we’ll assume that additional 2.5% just covers the cost of payment processing and is therefore revenue neutral.

So, under the simplification above, we’ll assume that for every Z dollars in retail trading volume, Coinbase generates: Z * (1.49% + 1.00%) = 0.0249*Z dollars in revenue.

Coinbase Pro (previously GDAX) Fees

Coinbase Pro charges 30 basis points, or 0.30% in exchange fees for price takers. Accounts that trade $10–100m or >$100m over a 30-day period receive a 10 or 20 basis point refund, respectively. (Documentation is here.)

The market data we’ll use doesn’t attribute volume by-trader, so we have to assume the relative proportion of Pro traders’ volumes in order to calculate a mean exchange fee. Because it is very easy to create a Pro account and even the discounted taker rates are prohibitively high for high-volume traders, it seems reasonable to assume that most volume is subject to the full 30 basis point fee. For the sake of easy comparison, we’ll assume the mean Pro fee is 24.9 basis points, or a tenth as much as the Coinbase retail fee.

As the market matures, both the absolute rates and the degree of retail vs. pro price discrimination seems unsustainable. As is, any savvy trader should never use retail Coinbase. To-date, the segmentation has been quite successful, however. Many consumers either don’t realize the difference, or put a big premium on the convenience of the mobile app (as evident by the fact that it reached #1 in the app store for a while in late 2017).

Expect serious downward pressure on Coinbase’s fee rates in the future.

Competitors generally charge significantly lower fees for “Pro” style exchanges. Ceteris paribus, expect pressure towards zero fees for investor-facing exchanges as firms seek to capture market liquidity. Of course, everything else is not equal and being a leader in governmental and institutional regulatory compliance has been and may remain a competitive advantage for Coinbase.

Coinbase will face similar retail fee headwinds. Robinhood already offers “free” exchanges though custody isn’t completely transparent. Other entrants are sure to pop up as well.

Exchange Numbers

Proportion of Revenue from Retail and Pro

We’re going to make another simplifying assumption that the relative proportion of retail and pro traders has stayed constant because we don’t have better data to suggest otherwise.

Given the assumptions:

• that proportions of retail and pro are stable
• a 2.49% revenue rate on retail volume
• a 0.249% revenue rate on pro volume;
• and, $1 billion in 2017 revenue

we can calibrate the proportion of volume from retail and pro traders as follows: $1 billion = (X * V * 0.0249) + ((1-X) * V * 0.00249) where X is the proportion of volume from retail traders, (1-X) is the proportion of volume from pro traders, and V is the total volume traded in 2017.

Based on Coinbase’s API data, they did about $80 billion in trading volume in 2017. Plugging that into the function as V, we find that X is approximately 50.2%. So, Coinbase appears to have done about half of their volume at retail and half at pro rates. $1 billion revenue on $80 billion in volume.

These Numbers in Perspective

NASDAQ, the second-largest stock exchange in the world, represents about $12 trillion in market cap. In the first half of 2018, NASDAQ traded $14.777 trillion. They claim to do about 1 in 10 security share sales in the world. Their exchange generated $517 million in trading fees based on this volume (Q1and Q2 financial results), which includes not just cash equities but more expensive products like derivatives. That’s a mean fee rate of about 0.0035%, orders of magnitude less than Coinbase. And, NASDAQ even has the advantage of near-exclusivity for many of its traded assets.

An exchange is a good business, but margins go down as liquidity goes up. In the long run, Coinbase doesn’t want a world in which their current fee rates are sustainable. That is indicative of an inefficient and speculative market, not the frictionless economic protocol they hope cryptocurrencies become.

An exchange simply doesn’t support big tech scale businesses. Put another way, Apple makes more setting the default search engine in Safari than NASDAQ does with its entire exchange business.

Coinbase’s Volume Trend

This is the cumulative volume traded on Coinbase (note the first years of the company’s existence aren’t shown because you wouldn’t be able to see it anyway):

Annual totals are as follows:

Finally, monthly totals are as follows:

Appendix: Potential Material Fallacies

These are some of the reasons my analysis above could be wrong:

• Fundamental misunderstanding of how trade volume is counted. It’s possible Coinbase obfuscates or does not include all trading in the data available on their API. For example, they may subtract retail trades. I don’t have any insider information, so I can’t rule this out.
• Historical fee rates may have differed significantly. This analysis applies current fee rates to historical data. Coinbase has certainly played with these rates, but I wasn’t able to find a detailed historical fee schedule, so I’m applying the current rates. To the degree these differed, this analysis will be inaccurate.
• We don’t know the % of pro trading that qualified for high-volume discounts. This would affect the retail vs. pro volume proportions calculation.
• $1 billion in 2017 revenue may not be accurate.
• The retail vs. pro mix has shifted in 2018. That would affect my estimate of Coinbase’s H1 2018 revenue.