Through their investment subsidiaries and custody divisions, banks - which plug into high-value payment systems for clearing and settlement of trades - are a pivotal hub in electronic trading. Last year, Banc of America Securities (BAS), a subsidiary of Bank of America (Charlotte, N.C.; $1.21 trillion in assets), created Electronic Trading Services (ETS) to develop products for the firm's institutional investor clients. BAS managing director Rob Flatley discusses how electronic trading developments, such as algorithmic trading, are increasing demands on banks' technology infrastructures with InformationWeek's Steven Marlin.
BS&T: What's the linkage between securities trading and payments systems?
Flatley: Trading and payments systems are inextricably linked by the processes of executing, clearing and settling trades. Trade settlements form a large component of high-value payments among banks.
BS&T: How do banks' corporate customers fit into the trading picture?
Flatley: Corporations are huge providers of trade order flow through their pension funds and their relationships with investment managers. They are focused on achieving the optimal way to fund those pension plans. They're comparing returns up to the fourth or fifth decimal place, which means they need to trade efficiently and at the lowest possible price.
BS&T: Has the Financial Information eXchange Protocol (FIX) reduced manual steps in trade execution?
Flatley: Not entirely. The industry group that provides that common language, FIX, hasn't devised standards for algorithms. The major sell-side firms - Bank of America, Goldman Sachs, JPMorgan - all employ their own specifications [for algorithms]. So a lot of translation has to take place. It causes a lot of delays in the ability to execute trades.
Another issue is a lack of interoperability among order management systems, order routing networks and FIX engine providers. The buy side ends up having to manage a three-pronged relationship with all sorts of interdependencies. That's created a lack of usability for what should be a straightforward protocol.
BS&T: How does your IT department work with the business side?
Flatley: In ETS, our business looks more like a software company than a traditional broker-dealer. IT is an integral part of the business. We run it like the development department of a software company rather than the IT department of a bank. IT works with us to define the strategy of the business, which is very different from the traditional model.
BS&T: What kinds of computational resources are needed to perform algorithmic trades?
Flatley: As a broker-dealer, you need to be efficient front to back for the entire securities processing cycle. Suppose you receive an algorithmic order to trade 100,000 shares of stock - the algorithm slices that order up into an average order quantity of 120 shares. That means you may get 1,300 different executions for that single order, which puts a lot of stress on your system and a lot of downstream systems. Multiply that by 10,000 other orders coming in at the same time, and you have an idea of the computational complexity.
In order to trade algorithmically, your systems have to be capable of dealing with 1,300 separate executions. In addition, executions may have to be filled on [multiple electronic exchanges]. You need to be efficient in managing costs across all these markets. Not only do you need to be efficient in the front office, but you need to be efficient in the back office as well. There's a whole process of affirmation, confirmation and reconciliation that occurs after a trade.
BS&T: What is the role of custodians?
Flatley: The custodian services [Banc of America Securities'] customers. Major banks have multibillion- or even multitrillion-dollar custody businesses. Those businesses are like retail banking - they're all about scale. They receive shares in and are asked to send shares out on a daily basis based on instructions they receive from Wall Street. If a buy-side firm, such as Putnam, uses Bank of New York as a custodian, and Putnam sells 50,000 shares of Dell, it sends BoNY a message that says, "Sell 50,000 shares of Dell and Banc of America Securities bought it." And BoNY gets a similar message from us. All of those processes now are performed electronically. The whole post-trade cycle has become more efficient.
BS&T: What economic forces have come into play with respect to electronic trading?
Flatley: There's been a lot of downward pressure on commissions over the last four or five years. Commissions went down rapidly from above 5 cents through the 3-cent barrier, and down to subpenny for certain types of transactions. Large financial institutions have had to figure out how to make money at less than 2 cents a share. So they've invested in technology, people and servers to make that happen.
BS&T: How fast is algorithmic trading taking off?
Flatley: The algorithmic channel accounts for only 5 percent of overall trading today, but it has been projected to double each year for the next three years. According to a survey by Financial Insights, 94 percent of hedge funds and 66 percent of institutions already have traded algorithmically, so there's been a broad base of early acceptance and experimentation.
BS&T: How will firms like yours meet that demand?
Flatley: They're scrambling to come up with strategies on both the processing and distribution sides. They've done two things really well - they have built front- and back-office systems that scale well and provide a rich degree of configurability, and they've also developed sophisticated distribution channels to enable plug-and-play access to their trading algorithms.
BS&T: What are the key technology considerations involved in algorithmic trading?
Flatley: From a distribution perspective, you need to make product as easy as possible to interface with - which largely is a matter of configuring third-party systems. Within your own infrastructure, you've got to design an algorithmic server and application to be flexible, because the market changes so rapidly. We've implemented a rules-based technology so we actually can modify the algorithms or create new algorithms without creating new source code. That means a financial engineer who's not a programmer can go in and create and test algorithms without a laborious cycle between IT and the business.
You're sending a lot of orders to different markets, and you have to be able to manage the status of those orders. That taxes computer resources. We learned how to create an efficient memory-based application in order to manage order status. From a telecommunications perspective, you've got to have fast lines down at the exchange. We try to place ourselves as advantageously as possible on the networks to eliminate any latency. So when we send an order, we can get the order down to the floor in 10 milliseconds to 20 milliseconds instead of 100 milliseconds. How you physically locate your servers and configure the network is very important.
BS&T: Describe your backup and recovery systems.
Flatley: In e-trading, you've got to have a fault-tolerant environment - cold backup, hot backup and failover. If any part of our execution cycle fails - router, switch, telecom provider - it automatically fails over to another component. So we're in a constant fault-tolerant mode. We've also built our own monitoring system, called The Eye, that watches all of the hardware and connectivity components, so we measure latency on every hop and line. Those tools automatically will switch us over to a less-clogged channel.