Size matters.
That’s the mantra over at Citigroup Global Markets these days, where Dan Keegan, the recently promoted co-head of the electronic trading division, is steering his group to become bigger and bigger.
"Owning volume–and then optimizing that liquidity to provide customized solutions for those customers that matter to us–will separate the wheat from the chaff," Keegan said.
Formerly head of sales for the electronic execution group, Keegan moved into his current position in May, following the departure of Steve Swanson. Keegan tells Traders Magazine his "personal goal" is for Citi to expand its share of the domestic market from 13 to 20 percent and rank among the top five trading firms.
Being big gives Citi two advantages, Keegan explained. First, size means Citi’s brokerage division is able to give its customers better service. Second, size gives Citi’s investment banking division an advantage when it competes for a deal.
"This isn’t just about aggregating liquidity on behalf of our institutional customers," he said. "It’s for the broader enterprise and all the customers we are trying to service."
On the brokerage side, controlling a lot of volume means Citi can guarantee its institutional customers they will have a good chance of finding a match within Citi. It also allows Citi to tailor solutions to clients with different needs–some may focus on price while others focus on size, for instance.
Much of Citi’s inflow meets in Citi Match, the big broker’s dark pool. Not coincidentally, it was Keegan who developed the crossing system from a platform he built at ATD. Keegan joined ATD, a pioneer in the use of technology and sophisticated analytics in wholesaling, in 2006 to build up the company’s institutional business. He joined Citi when Citi bought ATD in 2007.
The Citi Match pitch to the buyside trader goes something like this: Imagine a product where, on average, you can be 20 percent of the volume in a stock over the life of the order without ever hitting a bid or lifting an offer. Plus, you can capture 70 percent of the spread at the same time.
Citi Match captures flow from the broker’s institutional customers, various Citi trading desks, ATD as well as its Lava Trading unit. The dark pool executed 37.3 million shares, single-counted, in April, placing it among the 10 largest dark pools tracked by Rosenblatt Securities.
Market Maker
On the investment banking side, size could give Citi the upper hand when vying for an underwriting mandate. If Citi consistently trades a good chunk of the issuer’s stock, it will be able to assure the company it will make a market in the stock regularly, Keegan explained.
Investment banks will often only trade their corporate customers’ stocks aggressively in the immediate aftermath of a deal or before the next one.
"We think it is important that a corporate customer understands you are not only there to help them stabilize the stock on day one," Keegan said, "but that you are an active market maker on their behalf for the next six months. Oftentimes, capital dries up after the deal heat is gone and only resurfaces as you get closer to a secondary."
Keegan predicts that five firms will control 80 percent of the volume of the U.S. market in the next three years. The exec wants Citi to be one of them, noting that it is already one of the five biggest providers with its top 200 accounts.
Citi has about 13 percent market share. That number has not changed substantially in two years. Keegan noted that its market share was on the rise until the firm ran into "Citi-specific problems" related to the 2008 financial crisis.
A good chunk of Citi’s inflow comes from its stake in retail giant Morgan Stanley Smith Barney. Citi has a 49 percent stake in that venture, created after the bank sold its Smith Barney unit to Morgan Stanley. But Citi topper Vikram Pandit has indicated he will eventually sell it all to Morgan Stanley. All flow would likely head to Morgan Stanley when that occurs.
Achieving critical mass is not easy, sources say, whether the aggregator is a broker or an exchange. Bill Harts, an operating partner with Bessemer Venture Partners and former head of strategy for Bank of America Securities’ equity trading group, notes that size is indeed helpful when it comes to servicing clients, but difficult to attain. Harts is also skeptical of a growth strategy that relies on the matching of institutional orders with retail flow. "Unless they are benchmarked against a passive metric like VWAP, institutional clients typically don’t have the patience to wait in a dark pool for the right retail order to come along," Harts said. "They want to get their trades done quickly. This accounts for the popularity of so-called arrival price algos."
While Citi is not alone among big brokers targeting market share and viewing themselves as liquidity aggregators, its philosophy is not universally shared among the bulge bracket. Morgan Stanley, for one, has a different idea of its place in the market. "The role of the exchange and the role of the broker have become blurred over time," Andy Silverman, Morgan Stanley’s global co-head of electronic trading, said. "We are not here to compete with the exchanges. We are here to service our clients."
At Citi, the electronic trading group handles upward of 70 percent of Citi’s total U.S. equities inflow. It counts as customers traders both inside and outside the firm.
To win an increased share of the American pie, Keegan is relying on all three of the divisions within the electronic trading to pull their weight. That’s algorithms, Lava Trading and ATD.
Shane Swanson, Steve’s brother and ATD’s former general counsel, is responsible for the Lava unit. Lava’s main business is routing orders to market centers on behalf of broker-dealers.
Young Kang, formerly a trader with the giant hedge fund SAC Capital Advisors, runs algorithmic trading. Kang joined Citi in 2007 and subsequently orchestrated a revamp of its algo product line, bringing in an entirely new crew of analysts and developers.
Jeff Martin, now president of ATD, took over the reins of the wholesaler when Steve Swanson became Citi’s co-head of global electronic trading in June 2009.
It’s no coincidence that former ATD executives dominate Citi’s electronic trading group. About two years ago, former ATD chief executive Steve Swanson was charged with integrating the various departments within the group. Previously, Lava and ATD had been allowed to exist as separate businesses. But Citi management believed that needed to change.
In June of 2009, when the integration was largely complete, Citi announced the group’s new management lineup. Steve Swanson was made co-head of the department along with Shakil Ahmed, who is also responsible for quantitative strategies.
Shake-up
Keegan became global head of institutional sales, with particular responsibility for Citi Match. Shane Swanson got Lava and Martin took over ATD. Other former ATD executives assigned to top roles include Jack Vensel, responsible for electronic trading in Europe, and Myrick Crampton, in charge of predictive technologies, an increasingly important activity across equity trading.
With the shakeup, some longtime employees departed. Notably, Richard Evans, Citi’s London-based head of global electronic trading, left in November 2008. He has since joined Morgan Stanley. Dave Weisberger, a veteran of Salomon Smith Barney who once ran Lava, also left in 2008 and now works at a hedge fund.
All three of Keegan’s lieutenants–Kang, Swanson and Martin–face the same dilemma as they try to build up their businesses: trying to satisfy ever-changing customer expectations while dealing with an ever-changing market structure.
And if there is a common denominator among the challenges each face, it is that of dealing with the onslaught of high-frequency trading. With HFT strategies accounting for around half of all shares traded, it is having a major impact on all brokers. In Citi’s case, the impact has been generally negative for ATD, positive for Lava, and a challenge for the algo desk.
New System
Kang’s algorithmic trading platform was built as a response to a stock market shattered into dozens of fragments in recent years, as well as the tremendous increase in proprietary (read: high-frequency) trading.
"Three years ago, when I came here and looked at the platform, I knew we had to make a choice: Continue to put bandages around it–duct tape it and make it serviceable–or get ready for how the market would evolve two or three years down the road," Kang explained.
The exec chose to completely scrap the platform and build something new. The new system would have to deal with market fragmentation that was "out of control," faster trading speeds and an increase in gaming by HFTs. "Could we have retrofit the existing system to handle of all that?" Kang asked. "No way. It simply did not have enough brainpower to make the decisions it needs to on the fly."
The new platform, in use by Citi’s traders since April 2009 and now available to the broker’s do-it-yourself buyside customers, is able to suck in a huge amount of both historical and real-time data to make its trading decisions. That’s something the old system couldn’t do.
Citi’s original platform, put together in haste in order to get the broker into the algo business as quickly as possible, could only take in a limited amount of data. "Instead of looking at a half dozen pieces of data, we can now look at hundreds of pieces of information," Kang said.
That capability is necessary as there are about two dozen exchanges, ECNs and dark pools where a significant amount of trading occurs. And because those pools of liquidity have differing rules, pricing strategies, degrees of information leakage and kinds of participants, the intelligence of an algo platform becomes even more critical.
Three years ago, Kang explained, the buyside expected its brokers to be able to access all the available pools. Today, they want much more. They want to know to which venues their brokers are routing. Plus they want the option to steer clear of certain market centers. "The buyside is starting to catch on about information leakage," he said.
A recent study by Tabb Group echoes Kang’s observations. The consultant surveyed 123 head traders at traditional asset managers and hedge funds and concluded that the buyside was focused on the performance of brokers’ algorithms. Traditional asset managers were especially keen on the venues to which their orders were shipped, noting they might want to "dictate the configuration of the routing criteria."
Still, having the latest and greatest algorithm does not necessarily translate into more business, sources said. Relationships count for a lot. Tabb noted "the broader relationship is important for a significant proportion of firms and rates as a factor in the selection process." Harts considers them more than just a factor. "If a buyside trader has the choice of 10 VWAP algorithms, for instance, which one is he going to choose?" he asked. "His decision will be dictated by his relationships. A good sales trader is critical to selling a good algorithm."
Out of the top 13 providers of algorithms, Tabb ranked Citi as No. 7, with a 28 percent share of all buyside traders surveyed. Credit Suisse and ITG rank at the top of the pack.
If routing has become a preoccupation in Kang’s department, it is an obsession in Shane Swanson’s. Citi entered the business of routing broker-dealer orders to market centers in August 2007, by leveraging Lava Trading’s prodigious ColorBook network. This "GOTO" routing business is run under the banner of Citi’s LavaFlow ECN, which launched in February 2007.
Under the GOTO program, ColorBook’s smart-order-router technology transports orders into the LavaFlow ECN, Citi Match and ATD, as well as all other public markets and 16 dark pools. Customers are all broker-dealers-high-frequency trading firms, retail brokers or proprietary trading houses.
On one day in July, for example, about 245 million shares entered the system with about 76 million executing on LavaFlow. The balance was routed away under LavaFlow’s GOTO market participant identifier.
The GOTO service competes with exchanges and other specialty brokers on the basis of price and service. Citi’s standard rate is $0.0025 per share–less than that of most exchanges. That means it loses money on nearly every share routed to an exchange or ECN. But because enough volume sticks to Citi or winds up in another dark pool, Lava "survives," Swanson explained.
In the early years, LavaFlow got the majority of its orders from Citi. Now, because of the push into the order-routing business, less than half of the ECN’s flow comes from Citi, according to Swanson.
The ECN’s market share hovers around 1 percent. That’s double from last year, but still meager. "We are still a tiny sliver," Swanson said. "But we are showing the right growth pattern. It’s just not the trajectory I would like to see it on. We flirt with 1 percent. We go up. We fall back. We are gunning for 1 percent. We hope to break that this year. Then it’s onward and upward."
The biggest challenge, Swanson noted, is getting traders to connect to the ECN. "People have to be incented to get the connection," he said. "They have to understand how to use it."
Liquidity Aggregator
Growing the ECN, as with growing the order-routing business, is not an end unto itself. The ultimate goal is to bring in flow that can paired off against Citi’s buyside algorithmic flow.
That could happen in the ECN, where some algorithmic orders may be resting, or it could happen in Citi Match or ATD. Before sending a GOTO order out to the public market, ColorBook may "ping" Citi Match or ATD.
The more orders the Lava unit brings in, "the greater the opportunity to wash that flow over other liquidity we have so as to increase the number of crosses we do internally," Swanson explained. "Citigroup is a liquidity aggregator. We want to bring in as much liquidity as we can."
For Lava, the good news is that a big chunk of its future inflow could come from high-frequency traders, according to one analyst. Aite Group did a survey last September that discovered that high-frequency traders are not big users of smart order routing services. They prefer to trade on a single exchange rather than trade across multiple data centers. That reduces latency. As a group, HFTs comprised only 1 percent of the business, according to Aite. "Once you start crossing data centers, you acquire latency," Adam Honore, the study’s author, explained, "and then you are not high-frequency trading."
Still, the future looks brighter, Honore said, as providers of co-location, or proximity hosting, and smart order routing adapt. Today, only one-third of all smart order routing providers co-locate. That number should increase "if the data centers, including the exchanges and third-party providers start aggregating multiple liquidity pools," Honore said. Lava does not co-locate its servers, Swanson said.
With high-frequency trading accounting for maybe half of all shares traded in the public markets, the impact of the firms that deploy these strategies is obviously significant. And while Swanson’s Lava unit is benefiting from their connectivity needs, Jeff Martin’s ATD is trying to cope with the stiff competition they bring.
Some firms using rapid-fire, co-location-dependent trading strategies have entered the wholesaling business, forcing traditional players such as ATD to rework their trading strategies, reduce their latency and spend more on price improvement.
"It’s a big change," Martin said of the new entrants in the business. "It makes it tougher for us obviously."
As a wholesaler, ATD solicits order flow from retail brokers, making a market in about half or two-thirds of the flow. The rest of the orders it routes away to the public markets or dark pools. Some is matched up against flow from Lava’s or Citi’s institutional customers.
To win that flow and appeal to brokers who might otherwise route to HFT players, Citi has had to spend more on price improvement. Assuming a quote of $10.00 bid to $10.01 offer, Citi might buy stock at $10.00 and sell it to a customer at $10.007. That improves the price the customer receives but reduces ATD’s spread.
"You can say that three-tenths of a cent is not that much, but if you multiply that by the billions of shares we trade every month, then those numbers become quite substantial," Martin explained. He added that Citi spends between $5 million and $6 million per month on price improvement.
In addition, if ATD can’t fill the order internally, it will route it away to the market, absorbing any routing charges. "We work as partners with our clients," Martin said.
High-frequency shops, on the other hand, don’t work this way, he noted. These players accept immediate-or-cancel orders only. If they can fill the order with their capital, they will. If not, the order is kicked back to the client.
Higher levels of price improvement are not the only costs that firms using high-frequency trading strategies have imposed on ATD. The drive by HFTs to wring every microsecond out of their trading processes has forced ATD to spend substantial sums to reduce its latency as well, Martin said.
"It’s a microsecond game," Martin said. "We have to continually improve and rework our systems to be able to respond to orders faster. If there are 500 shares available in the market, my job is to get that 500 shares before it disappears. You better be fast or you’re going to miss it."
In the best-case scenario, of course, ATD wouldn’t have to turn to the public markets to get its customer filled. It would find a trade inside Citi Match. That’s the goal, anyway.
"Our algorithms have been completely revamped," Keegan said. "And we have optimized all that flow such that pieces of every one of those orders sit inside Citi Match. At the end of the day, we want to prove that owning and optimizing order flow will provide the customer with a superior execution experience."
