Automated Trading Desk never wanted to be seen as a typical securities firm. That’s one reason founder David Whitcomb, a Rutgers University finance professor, started it up in Charleston, South Carolina, 750 miles south of Wall Street. He defined it not as a brokerage but as a “technology company with a trading focus” and hired chief technology officer Jonathan Butler as its first employee.
That was back in 1988, when ATD was Whitcomb’s laboratory: a proprietary trading firm designed to test his theory that computerized trading systems could outperform humans. Whitcomb turned to Butler, who had just earned his bachelor’s degree in computer science from the College of Charleston, to develop software code for automated limit-order trading.
Algorithms developed under the supervision of Whitcomb, Butler and Steve Swanson -- another College of Charleston graduate, who succeeded Whitcomb as CEO in 2001 -- became the core of ATD’s trading operation, both for the firm’s own account and, for the last three years, as an agency executing orders for others.
The algorithms, which predict extremely short-term price movements in stocks and execute split-second trades to take advantage of those projections, are closely guarded -- as are the names of the buy-side and brokerage house clients that send trades through ATD. But there is no hiding ATD’s impact on the equities industry: The firm says it accounts for some 6 percent of the Nasdaq Stock Market’s daily volume and 3 percent of the volume in listed stocks.
Of 300,000 trades on a typical day, 95 percent are untouched by human hands. “We have the technological tools, and we know how to use them,” says Butler, 40.
Most of the current order flow is retail, which puts ATD in competition with such volume-trading shops as Jersey City, New Jerseybased Knight Trading Group. CEO Swanson says that ATD is likely to take on more institutional business, particularly as its base of hedge fund customers grows.
The firm is making its mark with a staff of just 65 working in the $38 million, 68,000-square-foot headquarters and technology center that the company moved into two years ago in the Charleston suburb of Mount Pleasant (the address: 11 eWall Street). Extending its automated approach into market making, ATD in 2002 acquired Chicago Securities Group, a specialist firm at the Chicago Stock Exchange, which added 30 employees.
“One can’t help but be impressed,” says Damon Kovelsky, a capital markets analyst with Financial Insights in Framingham, Massachusetts. He warns, though, that ATD may be “too far out on the technology curve for its own good” -- its systems so arcane that potential clients are reluctant to trust them with critical order flows.
The most recent financial results published by privately held ATD have been trending downward: Revenues fell from $87 million in 2001 to $62 million in 2002, while net income dropped from $49 million to $15 million. The firm says it stayed in an investment and expansion mode even as markets were declining.
“We are a technology company, and we have to constantly invest in people, systems and tools for solving problems in a very customized way for our clients,” Butler asserts.
The CTO discussed ATD’s technology focus in a recent interview with Institutional Investor Assistant Managing Editor Jeffrey Kutler.
Institutional Investor: Did you have any idea that you would end up in finance?
Butler: I got bitten by the programming bug when I was 12. I’d convinced my parents to buy some Tandy Corp. stock, because shareholders got a 10 percent discount at Radio Shack, where I got my first computer. Being a shareholder got me interested in the stock market. By the time I was in college, I had a brokerage account, traded various stocks and even got into some options. But the technology -- the ability to create things -- was the most interesting and exciting part for me.
How conversant were you in the academic theories behind ATD?
I got pretty immersed in the development stage. Dave Whitcomb, our founder, would send a steady stream of faxes that would greet me each morning. We would talk them through over the phone, and that became the basis for our first trading system.
How does your technology today compare with what it was then?
It’s a lot more distributed -- and complicated. Instead of three central processors for trading 400 to 500 stocks, we have several hundred servers and trade a couple thousand stocks. The great advantage of today’s distributed computing is that everything has its place and its backup, and the capacity is much more scalable. In the old days the only way to increase capacity was to buy one more of those very expensive machines.
What volumes are you handling?
We do hundreds of thousands of trades a day, and we had our first 100 million-share day in 2002. The daily average is 40 million to 50 million shares, not including our Chicago floor operation. On some days order placements are much higher than those totals. Our system is very patient; it will place a limit order, and if it isn’t satisfied, it will cancel.
Under what circumstances are trades not automated?
There are triggers -- they may be regulatory rules, or the models themselves recognize a certain pattern that isn’t working right. In those cases, a trade might be kicked out to a person to minimize a loss or assure best execution. Our modelers, who are traders, will go back and examine auto trades that don’t perform well and think about what they would have done in that situation. That’s all part of optimization.
What impact do exchange rules and market structure changes have on your infrastructure?
Anytime an ECN changes the way it operates, or any market structure rule changes, we have to update and test our systems. We also have to serve the needs of our customers in how they want to communicate with us. These choices have changed many times over the years, but now, fortunately, many customers have adopted the Fix [financial information exchange] protocol, which has made that aspect of our connectivity much smoother. But the data flows have gone from a trickle 15 years ago to a flood. There is always some kind of new feed or piece of data coming out that somebody has to look at, and we deal with that.
How big is the technology group that works for you?
Probably half the people in Mount Pleasant, about 30, are into the technical area of operations. Another 15 to 20 are traders and modelers, who are researchers for us as well. They make suggestions to improve the way our systems handle orders. We started out being our own customer, and a very demanding one at that. If we didn’t like something, we got in there and changed it.
How do you manage such a sizable operation with so few people?
There isn’t a lot of hierarchical paper-pushing. Where we see a need, we act quickly and make changes as we see fit. We divide up into a lot of teams, which multitask. Six or seven teams might work together on a given project, but different teams take lead roles at different times.
Will you take the ATD model abroad?
We have been doing some trading on the Toronto Stock Exchange and have done some experimentation on Euro-next Paris. But we’ve backed off a bit to concentrate on our customer efforts, which we felt had to be really solid to maintain the quality we want as the business grows. It’s a dilemma we sometimes face: We can’t wait to use the technology in as many places as possible, but we have to be cautious about spreading ourselves too thin. We are, however, looking to get back to using our systems in other markets by the end of this year.
Are the systems easily compatible with overseas markets?
Just as we have been able to handle many different customers with a wide variety of needs, we certainly could trade with our systems in many different markets. There are always going to be particular details in a market, such as trading symbols, that we would have to customize for, but that is one of our competencies.
