Episode 103: A Wall Street We Can All Trust with Brad Katsuyama

Jon Prial: Trust. Trust is something we inherently understand most likely when we're thinking about working with people. We know what it means when one says," I trust that person. I think I'm being treated fairly." Here at Georgian Partners, we've been writing and talking a lot about how to get a customer to say, " I trust this company," not that easy, but doable. Now, there are times you never think about where trust might be broken between you and a company. It's sad to say it's happening everywhere. I personally no longer trust I'm getting the best price when I buy an airline ticket or I purchase something online. Supply and demand can be subverted a bit when technology, data and algorithms are involved. Today, we're going to be talking about trust in the securities market. Think of those traders on the stock floor, buying and selling orange futures in the movie Trading Places. Supply and demand. Now that might be a bit extreme, but you get the point. Prices fluctuate in some reasonable fashion based on supply and demand, but there is something most of us hadn't thought about until a 2014 Michael Lewis book Flash Boys was published about high frequency trading. One can have a faster connection to an exchange mainly by having a shorter distance between two points. Now, some traders are not buying on supply and demand. Traders are buying and selling to make fractions of pennies on transactions because of your trading plans and because they can do it faster than you can. That's affecting your trade prices. A bit like printing money and it really isn't fair. Trust, what can we do about it? Today we'll be talking with Brad Katsuyama the CEO of IEX. IEX is one of Georgian's latest portfolio companies and Brad is one of the heroes of Flash Boys. IEX, the Investor's Exchange, is a stock exchange for US equities that's built for investors and companies. Approved as a national stock exchange in June of 2016, they are already doing more notion of volume than the TSX, London Stock Exchange and the Deutsche Boerse. I'm Jon Prial and welcome to the Georgian Impact Podcast. Brad, we're glad to have you with us.

Brad Katsuyama: Thank you.

Jon Prial: I've always been a big fan of Michael Lewis, loving his new podcast Against the Rules where he really digs into fairness and what's happening to referee, so to speak. Of course, I encourage everybody to read the book, Flash Boys, as your story is such an important starting point in really thinking about this topic. Without getting too deep, just explain to me how things became, not fair. You were at RBC, you were executing large trades, which because they were large, you couldn't execute them in one batch, so what happened then? What did you discover?

Brad Katsuyama: Sure. At the time of discovery, I was running US trading for the Rural Bank of Canada, and my clients were pension funds and mutual funds and hedge funds so looking to trade in larger size. It just turned out that in 2006, let's say I had a hundred thousand shares of GE to buy. I'd be able to pull up a quote. I would see there's a hundred thousand shares. It would be offered at a particular price and I would say, buy a hundred thousand GE$ 10 go, and I'd be able to buy a hundred thousand shares. In 2007, same situation would happen. I's see a hundred thousand shares at, you know,$ 10, I'd try to buy it, but I'd only be able to buy 80,000 shares and the other 20 would seemingly disappear and the price would tick higher, which would mean that I would have to buy 20,000 shares at a higher price. By 2009, I'd see a hundred thousand shares offered at$ 10, I'd try to buy it and I would only get like 40,000 shares. 60 would evaporate and then I'd have to pay a higher price for the remainder.

Jon Prial: When you said you could only buy 40, 000, let's say that means somebody else went and scooped up the 60, 000 before you got there?

Brad Katsuyama: Yeah. The interesting part was, and again, life has a lot to do with luck and I think RBC offered me, in 2009, a job to be the head of global electronic sales and trading and essentially what that meant was I went from managing a group of human traders to a group of computer programmers, network engineers, people that were building algorithms that the trading team would use, including the team that I had previously run. I took that job because I wanted to start to learn more about technology and how it was affecting the stock market and these people, whether they knew it or not, knew more about how the stock market actually worked, then me and my trading team did because they understood the interlinkage of the stock market. They understood that the hundred thousand shares that I saw on my screen were essentially orders that were represented at multiple different exchanges, physically located in different data centers, across New Jersey. The New York stock exchange trades, no shares, zero shares down at wall and broad street. What you see on CNBC is a television studio. Any order sent to the New York stock exchange gets beamed out to a data center that they built in Mahwah, New Jersey. Essentially what they said is, your message is to buy 5, 000 at this exchange, 20, 000 at that exchange, et cetera, adding up to a hundred thousand shares all at$ 10 are being sent out from RBC downtown Manhattan. They're going up the West side highway of New York where the fiber routes run. They're going out the Lincoln tunnel, where the connectivity is to New Jersey and they're getting to these different exchanges at different times, simply because they're physically located in different places. This is going to sound completely berserk, but the difference in arrival time between getting to the first exchange, which was BATS would get their first and our last order would arrive at the New York stock exchange up at Mahwah, New Jersey, 30 miles north of where BATS was two milliseconds later, so two one thousands of a second was the difference in arrival times. one of the people I hired at RBC to help me, who's now a co- founder of IEX, he was building high- speed trading infrastructures and when I hired him at RBC, he said,"You know what? I just built an infrastructure that can get me from BATS to the New York stock exchange and 476 microseconds." I said," What?" Basically four times faster than RBC and basically when I asked," How is that possible?" and he said," Well, exchanges are willing to sell me things that allow me to take my servers, my client's servers, put them in the same building, they're angling to be as close to the matching engine as possible. We have our own fiber optic networks. They're erecting microwave towers." Basically, it's just a race and so essentially my order would arrive at BATS. I would buy all the GE that was there, but I would create a signal that there's a big buyer of GE coming and then people would race me to the other exchanges to do one of two things. One is that they want to cancel all their sell orders because they don't want to sell GE anymore because here comes a big buyer. Second, they would want to buy shares ahead of me to sell back to me at a higher price, which many people would think, well, that's front running. Maybe it's electronic front running, but it's not their client. It's my client that they're trading ahead of and the exchanges are the ones that enabled them to do that. In many ways, the interesting part is when people read Flash Boys, they get really upset at high- speed traders, but in reality, high- speed trading in the form it exists today, would not exist in the same way if the exchanges hadn't sold them the ability to do it, knowingly sold them the ability to do it. In many ways, the referee in our game is actually making more money selling things to one of the teams than they are actually reffing the game. The exchanges make more money selling high- speed data and technology than they actually do for matching buyers and sellers, which is a crazy concept for, I think, people to grasp.

Jon Prial: I'm going to come back to the exchanges and how they make money a little bit just to kind of close on this piece. I think a point that Michael Lewis made in the book was that high frequency trading was akin to card counters in blackjacks. They only made a trade when they saw an edge, when they saw these hundred thousand share purchase of GE and a day never went by, he said, where they lost money. Here you are working hard trying to get your job done at RBC and make the most money you can for your clients, so you ended up creating an algorithm to sort of adjust this. You didn't go to the Mahwah place first. You went to the Mahwah second, is that correct? Well, how did that play out?

Brad Katsuyama: What we actually ended up doing was we ended up creating an algorithm that got called THOR that was essentially... Instead of sending all the orders out at the same time and arriving just based on physical proximity at different exchanges at different times, what we did is we staggered when we sent out the orders with the goal of arriving at all exchanges as close to the same time as possible and we got that inter exchange arrival variance from two milliseconds down to 290 microseconds. Right when we did that, because basically, it was, they would have to break the speed of light to get from one place to the next fast enough, our fill rates went from 40-50% to a 100%. Basically we ran this, we build this algorithm and it solved the problem. Now, the interesting thing is it's not that simple. Because networks speed up or slow down based on the amount of messages, the latency between RBC and the New York stock exchange was not static. It was changing. The algorithm essentially sent probing orders that would populate a table, and then when we got a client order to send out to the exchanges, we would take the last snapshot of the table and use that as the reference for the inter exchange latencies. It was a bit harder than it sounds, but we got it done and basically RBC in the US went from number 19 to number 1 in US electronic trading products as ranked by the client in one year. Basically it put us on the map in a huge way and that was kind of the launching pad for starting our own company.

Jon Prial: You got back to where you were when you were trying to purchase a hundred thousand shares. You got back to where it was, where you got the price you wanted for your a hundred thousand shares. You figure your way around those, I use the term scalpers first, I guess, right?

Brad Katsuyama: Yeah. That's what it is. The only reason they wanted to buy is because I want it to buy. It's not like there was some independent decision to say," Yes, I'd like to own this stock."

Jon Prial: Right, no supply and demand per se at all.

Brad Katsuyama: You want to buy it so I want to buy it too, and I don't want to sell it anymore. I think the way I view it is that our clients at RBC were paying a tax, an unnecessary tax, and we found a way, in a novel way to eliminate that tax if you traded through RBC. the issue though, is that clients could only send so much business to RBC. We can do a hundred percent of the trading for some of these clients so one client put it in a really great way, which kind of sparked the idea for the exchange was to say," Well, you've only really helped me trade better with RBC. How do you help me trade overall?" and that was the idea around starting an exchange, because at the heart of this, the exchanges are responsible for what's happened.

Jon Prial: So let me step back. I think it's important for our audience to learn a little more about exchanges a little bit and the business model that's there, how they make money. I'll give you an example. You listen to podcasts in your community, listen to podcasts all the time. I learned about squirrels and how many squirrels take power outages out and I guess it was 1985, I believe where a squirrel in Trumbull Connecticut, 1987 squirrel took out the power in Trumbull, Connecticut and shut down NASDAQ for over an hour and the New York stock exchange folks were cheering because they had all the business. I always think that a stock sits on an exchange, an NYSE stock, it's a NASDAQ. I guess trading is happening everywhere, so if you don't mind, take me through just the basics of what's within an exchange, these different marketplaces.

Brad Katsuyama: Sure. At one point, NASDAQ predominantly traded NASDAQ listed stocks and New York predominantly traded New York listed stocks, but really since the mid 2000's, regulation came in place that allowed a stock, let's say for example, Microsoft lists on NASDAQ, IBM lists on the New York stock exchange, but those two stocks, Microsoft or IBM, can trade on any exchange. Any exchange that gets a license is now free to trade. Microsoft trades can trade on IEX, IBM can trade on IEX and et cetera. The primary exchange, where you're listed, they are a regulator, your listing standards, they're the ones who kind of enforce those and then also they're responsible for your open and your closing auction. Those end up happening at your primary exchange, but during the day, stocks trade everywhere. Exchanges used to be, as I think they really should be, member owned utility like functions, where they were there to provide a service and that service primarily was matching buyers and sellers, giving people a central place to meet, a fairly transparent set of rules. The issue though is that that position is also a position that can get abused. In the early 2000's, what happened is the New York stock exchange went through a pretty big specialists scandal where the specialists on the floor of the New York stock exchange, many of them were found guilty of front running orders and using their own books to trade ahead of different clients. There was a lot of uproar around that. Essentially they wanted to force the markets to become more electronic and that led to a set of regulation that essentially electronified stock exchanges in the US. At the same time, though, the exchanges went from basically nonprofit utility like member owned models to for- profit companies. Now you have a regulator which is an exchange. Exchanges regulate themselves. They're self regulatory organizations. They're also regulated by the SCC, but now they're for profit entities with a huge incentive to make money, but at what costs? I think now we have the outgrowth of that conflict.

Jon Prial: Where's the collection of money, obviously, if I use a broker and purchase stock, the broker collects a fee. What are the other fees along this, I guess I'll use the term supply chain? What are the other fees along the way here?

Brad Katsuyama: You have fund holders, so people that have pensions is managed by a pension fund. A pension fund may outsource part of that managed to an asset manager. The asset manager may charge a fee to that pension fund. An asset manager, when it wants to trade on an exchange by regulation, they have to go through a broker so the asset manager will pay a commission per share typically to a broker to execute an order for them. Traditionally stock exchanges would then charge a fee to the broker, which would be some fraction of what the broker charged the asset manager. It's kind of like a line of fiduciaries. Now where things got really out of whack, and it's one of the biggest issues in the stock market today, is that exchanges actually to incentivize brokers to send them orders, started to also pay the broker what's known as a rebate, except the broker keeps the rebate and does not pass it back to their end client so a lot of people refer to it as a kickback.

Jon Prial: That's because you said earlier that a broker could go to any exchange. A broker doesn't need to go to the New York stock exchange to trade IBM or NASDAQ, to trade Microsoft. There is incentivizing going on. I get it. But is that it? Why do they have to pay a broker to send them an order? Is there some way to foster competition? You tell me. Better price? Better tech? Can there be something else?

Brad Katsuyama: Well, because exchanges are selling such distinct advantages to high- speed traders, the only way to get someone to trade against such a advantaged customer is to pay the broker to deliver someone to trade against. When someone has such a distinct advantage, why would anyone on their own want to trade against that counterparty. They're paying the brokers and it's billions of dollars a year.$ 2. 8 billion were paid last year in the form of quote unquote rebates.

Jon Prial: There are people in the know, and obviously I'm not. Now, I trust my financial advisor and through her, the mechanism where trades are made and yeah, my pennies are being gathered along the way so there are elements in the system that are not fair. Now, those that know don't trust the system 100%, and this is a great set up for our discussion in IEX. I have to ask you, what is your relationship today with high- frequency traders? Do you have any friends?

Brad Katsuyama: It's a great question. You know what, there are some high- speed traders that really like IEX and the reason why is that there are good high- speed trading models, and there are bad ones because high- speed traders really are just arbitraging like securities for a profit. If you think of a traditional arbitrage example would be, if an ETF is created based on 30 underlying companies, there will be times where the ETF price may get inflated relative to the underlying companies. A high- speed trader will short the ETF, will buy the underlying companies, and then will let them settle for fair value and make money from those pricing discrepancies so that arbitrage exists. High- speed traders are the ones that arbitrage and monetize that and what it does, it provides a service because it keeps the price of the ETF in line with the price of the underlying stock.

Jon Prial: As it should be. As it should be, exactly, the price it should actually be.

Brad Katsuyama: I have absolutely no issue with that form of high- speed trading. Not all high- speed trading or automated trading is bad, but I look at the arbitrage. Why does the arbitrage exist and is there any purpose of that arbitrage in the first place? The issue again, is exchanges are mechanically creating arbitrage opportunities so that they can sell the high- speed traders the ability to take advantage of those and who's left? Who's on the other side of that trade? Pension funds, mutual funds, real investors, real people's money, and again, it's a tax on them that's totally unnecessary.

Jon Prial: Excellent. You decided you need to do something different. Tell me a little more about who benefits and who you're driving into this exchange and the role you play.

Brad Katsuyama: Our prime beneficiaries end up being brokers and the investors that they trade on behalf of, and certain high- speed traders who are also disadvantaged by the most aggressive... We call them predatory traders.

Jon Prial: I don't want to go another round into milliseconds and microseconds and stuff, but give me the root net summary of how it is that you pull this off.

Brad Katsuyama: One of the big advantages of the exchange sell it's something called co- location. Taking your servers and co- locating them in the same data center next to the exchanges servers.

Jon Prial: Sure. Shorter paths. Got it. Absolutely.

Brad Katsuyama: Absolutely. One of the first things we thought about when we were building our own exchange is to say," Well, what's the opposite of co- location?" and the opposite of co- location ends up becoming a speed bump and the best way to actually push people far away from our exchange is we coiled 38 and a half miles of cable in a box, stuck that in between us and anyone who connects to us. Slowing down a regular broker by 350 millionths of a second, which is the speed limit is 350 microseconds, they won't care. That's 1/ 1000th the speed of you blinking your eye, you don't care, but 350 microseconds is an eternity to some high speed trading strategies that are looking to pick up information and race ahead of others. We built a speed bump, essentially. It's like an$ 8, 000 piece of equipment that basically created like a massive, massive war on wall street, because a lot of people did not want IEX to become an exchange with a speed bump, especially the other exchanges who are selling speed. We did the opposite and it's free. It's free to connect to IEX.

Jon Prial: That's amazing.

Brad Katsuyama: Yeah. That's essentially what we did. We used technology to fight against the advantages of the other exchanges we're selling and that was the start of our journey.

Jon Prial: I'm going to have a whole different vision now when I think about when you originally talked about this going up the west side highway and out the Lincoln tunnel, now I got miles. I've just turned this Lincoln tunnel into a loop- de- loop going all the way around before I get out the other side.

Brad Katsuyama: That's right. That's right. That's essentially it's what it does but if you think about it applied to that example, we are in a cluster of data centers of which BATS is also there. Let's say your order arrived at IEX first, but you get information of a trade on IEX back through the speed bump, so by the time a high- speed trader realizes a trade has happened on IEX it's too late to do anything because the value of that information has shrunk because again, you're finding out late, you can't go and race ahead, you can't do all these things. It takes away the assymetry of people, learning of price changes at different variants of time.

Jon Prial: They get to sell 100 or buy 100, 000 shares of GE or that anybody figuring out that they could pick off, as you said, 40,000 or 60,000, they get their transaction to what they want the full 100, 000 shares. So as this evolves, how does your business model work in terms of transparency and trust? How do you really represent yourself out to your customers?

Brad Katsuyama: Well, I think one of the biggest assets IEX has, is trust. When we discovered what was happening at RBC, whether it was obvious to us or not, we had a choice. We could take this knowledge that the exchanges are willing to sell advantages and that proximity and speed, all these things are a core part of a scheme to make money. We could have gone off. A lot of high- speed trading firms were started by people who worked at exchanges. So we could have gone off and used, I guess, this knowledge for a different purpose. We decided to take this information to our clients and start to educate pension funds and asset managers about what was happening in the market. We've been doing that for nearly 10 years. These asset managers and pension funds in turn have supported us every step of the way. We got our funding from asset managers. They supported us in our launch when we applied to be an exchange and it was hugely controversial. They wrote letters to the SCC and supported us but I think because they trust us, the things that we're pushing for and what we're trying to do are exactly the things that they want in this marketplace. And so trust for us is something that has been earned over time, but it also affects the way that we operate, because the first thing we think about every time we're vetting an idea is, is this something our core clients would want us to be doing? If the answer is no, we're not going to do it. It's a guiding principle, but it's also a huge asset for us.

Jon Prial: Let me ask a question. Let me tell you what I trust. I trust in governance. I'm looking at a company, we look at companies, I trust in a great executive team and if there are issues I trust a board of directors is on the ball and pays attention. I know it's not the case all the time, [ inaudible 00:23:22] but I'll sleep at night, but now all of a sudden we have the [ inaudible 00:23:26] 500 not listing stocks that have these dual class shares because that impacts governance. There's no fairness and accountability if nobody's really watching the chickens in the coop, there's no hedging and what are your thoughts on that? How does that play in into your mindset here?

Brad Katsuyama: For me, it's a lesson I've learned over time. Incentives are what drives decision- making and I think if incentives are properly aligned, trust can be built. I think when I look at things around governance or Michael Lewis has podcast talks about the ratings agencies for bonds, the incentive was never to create a rating that provided the right guidelines to the investing public on the safety of, of certain debt securities. The incentive was rate as many securities as possible and deliver the rating because if you, if I don't give you the rating you want, you're going to go down the street and get a rating from somebody else.

Jon Prial: I love that. I think that's a great summary. The incentive piece. I think that's great. Let me try another angle with you. I love this. Moneyball, another Michael Lewis book, there's Billy Beane, the general manager of the Oakland athletics baseball team, he took a totally different approach and how to choose players for his teams and because of that, all sports are different. Data analysis is everywhere. In my mind, and I mean this in absolutely the most positive way, you are Billy Beane. Why do you see this technology going? What's going to happen next in terms of how you approach this? What are the next set of game changes we're going to see?

Brad Katsuyama: It's a funny analogy because probably the most influential book in my career was Moneyball because I had just moved from Toronto to New York, working for the Royal Bank of Canada. We were ranked number one in Canada and I think at the time I moved here, we were ranked 23rd in the US. We had a hard time hiring people. We were the underdog, the perennial underdog, and my boss at the time, a guy by the name of Bobby Gruber, bought the book Moneyball for everyone on the trading desk and I remember reading that and I'm sort of a baseball fan. I kind of grew up with the Blue Jays winning in the early nineties the World Series twice., but I read Moneyball as a business book and said," My God, this guy found a way as an underdog to just think differently," so what are the conventional acceptable norms that make no sense? That kind of philosophy has powered me throughout my career is to say, just because something's been done, doesn't mean that's the way it should be done. I think in many ways, the way we've learned and evolved is the idea to say," Well, why isn't it being done? And are those obstacles that we can overcome?" because again, acting irrationally could be the most rational thing to do if the incentive is to act rational. You know what I mean? I think it's just trying to understand circumstances and trying to understand what can change and the ramifications of doing that and I think that when you're willing to be transparent and your motives are clear and your incentives are in line, you build a model where even if someone likes you or doesn't like you, they will trust that your incentive is to build a market that is aligned with a particular vision and if that vision is their vision as well, it doesn't matter if they like us or not. We're we're both going in the same direction and we're completely transparent about the things that we want from this market. I'll give you a good example. There's two that come to mind from a business standpoint and this is just about thinking longer- term one is that we've built a machine learning algorithm that predicts price changes in the market, which is a similar I'm sure signal that high- speed traders have worked on where if I think Microsoft is going to tick down in the next two milliseconds, a high- speed trader is going to use that and they're going to try to sell Microsoft because they think the price is going to go lower. We at IEX use that same prediction, Microsoft's going to tick lower, to prevent a buyer from trading at what we think is about to be a bad price. We use it for defensive reasons. High- speed traders use it for offensive reasons. That signal is very powerful. We're on our fifth version. The sixth version just came out, which is actually an improvement on the fifth version. As we run the numbers, we will do one and a half million dollars less in revenue on an annualized basis by preventing a subset of trades on IEX from happening, because the signal is now better at predicting these price changes,

Jon Prial: Now it's a much more fair transaction than getting Microsoft at the right price, so as opposed to a generic speed bump, you actually have some, I don't know if this is a great where we're going with these analogies of Michael Lewis story, but there's a cop with a radar gun going," Wait, time out, time out, time out. I'm going to actually flip my lights on and slow you down because I don't like what's going to happen to the person that really wants to buy that Microsoft spot.

Brad Katsuyama: It's a prediction on top of the speed bump. The speed bump actually allows us to make the prediction but the important part here is that here's something that's going to cost us money in the short term, but in the long run, we will get more orders because the performance on IEX is of higher quality.

Jon Prial: Right because you're differentiating your company based on trust and fairness within these transactions. You're actually using this as a corporate differentiator.

Brad Katsuyama: Exactly, so if I said," A million and a half dollars, you know what, no, we can't do that, let's just..." no, absolutely not. This is about delivering the best price that we know how and if the signal is going to cost us$ 5 million a year, we would still do it because that$ 5 million is us collecting on someone else's bad execution. That's one version of what we've done. Another one that we did, which is again, along the same lines. The exchanges are overcharging the industry. They've been complaining for years, that market data costs, connectivity costs, the exchanges are gouging the industry and the SCC called the panel and brought a bunch of people down, brokers, investors, academics, exchanges, and said," We can't get our arms around how much exchanges are charging for data," The amazing thing about this is that exchanges charging, it's a subscription service where you have to pay me every month, but we're talking hundreds of thousands of dollars a month for data, but it's not their data. It's the broker's data. It's the investor's data. The exchanges actually don't generate any orders. They're just broadcasting other people's information. New York says," Well, look at what NASDAQ charges we charge a little bit more," and NASDAQ says," Well, look at what New York charges, we charge a little bit less." All the exchanges are doing the same thing. IEX said," You know what? We're an exchange. We're going to create a study that tells the industry how much it costs to produce market data. We don't care what you charge. How much does it actually cost to produce? And based on that, we show the industry a cent 2000 to 4000% markups is what they're charging when you take our costs as an exchange into account." When we did that, we sacrificed tens of millions of dollars of revenue, because we can never charge what the other exchanges charge. We give our data away right now for free, but this was about transparency. It was about shining the light on the exchanges so even you could say," Yeah, well you're sacrificing revenue," which we might be, but no, we're putting ourselves in a position to win by aligning with the industry and showing them that there's an exchange model. Yes, we don't pay rebates, but we're also not going to gouge you on market data and other things because again, it's a long- term play for us. I think transparency can't just be in our mission statement or something we put on our website, you got to live it.