Goldman Sachs' high-net-worth private asset management business has recently gone live with a new rules-based filtering technology that aims to keep unwanted stocks out of individual portfolios. Almost all client portfolios have owner-imposed restrictions and parameters which place some stocks and, often, entire sectors on a "do not touch" list. With a wide network of financial advisors ranging from the U.S. to Europe to the Far East, Goldman was looking for a way to ensure that all such stipulations were respected so that errant purchases were not merely corrected at great expense, but avoided entirely.
It was important to Goldman, a global investment banking and securities firm, to make the business of managing portfolios more efficient through technology, says Goldman Sachs Project Manager Lokesh Gupta.
"Some may say, 'We don't want to buy a tobacco stock, ever.' Some may require that their technology exposure does not go above 30 percent of the portfolio value," says Gupta, "and our European clients are very sensitive about their currency and country exposure."
Gupta, who works in the private-wealth management technology area of Goldman, which supports that sector of the business, says that there are many levels of input-business managers, compliance officers, management teams and private-wealth managers-that contribute to deciding which stocks to purchase for a portfolio. For that reason, it was specifically important that any technology implemented be centralized, as well as fundamentally structured to catch mistaken trades before they go through, not after.
"If we have a system that can check for rules as things are happening and on a real-time basis, then a lot of these inefficiencies and inaccuracies can be eliminated and that's definitely going to result in a lot of savings," says Gupta.
He says that Goldman did have an internally developed programming language used to do portfolio filtering but it was esoteric and not easily used by IT personnel unfamiliar with its intricacies. The result of remaining under such a system would have meant relegating valuable staff to a single post, leaving managers with little human-resource flexibility. Goldman had an idea of what it needed, from a technology perspective, right from the start. Systems at the firm were largely based in Java, meaning that a Java-based rules engine could easily be integrated into existing Goldman architecture. Gupta says that the search process for a vendor involved talking with people in the industry and getting on the Web for a look at providers.
Though he names the chief players in the rules-based space as Blaze Software, Sandia Labs and ILOG, Gupta refused to name the specific technology firm Goldman has chosen.
Robert Cooper is vice president of the Industry Solution Division at ILOG, a software company that designs components for optimization, visualization and business rules. One of ILOG's technology solutions is a rules-based engine, JRules, targeted at firms that need to sift through mounds of data in a hurry. In financial services especially, speed is a major priority and the reason why ILOG's Java implementation, which typically runs slower than other languages such as C++, can be tweaked to increase performance.
Users of JRules can write their own rules in a Web browser that then creates the logic for keeping a particular stock from entering a particular portfolio. The system is client-server or Web based with development and deployment licensing fees beginning at $10,000 each.
"It's one hundred percent Java so you can slip the entire engine inside of a Web-based application and we do have a C++ version as well," says Cooper. He adds that C++ can process about 10,000 rules in a second to Java's 8,000.
For Gupta, the decision came down to more than dollars and cents. He says initial high costs put him off but the vendor's subsequent pricing flexibility saw the deal signed. "Ultimately, it was the vendors goodwill, how many production implementations they had-these were the factors we took into consideration before we made a decision."