The $bn struggle to replicate Goldman Sachs' special powers
There's a piece of Goldman Sachs that every other bank wants. Billions have been spent in its pursuit, but with the possible exception of J.P. Morgan, no other bank has achieved it.
The genesis of Goldman's powers is held to be SecDB (Securities Database), the risk analytics and pricing system evolved by the firm over the past 20 years. Ever since the financial crisis, SecDB has been revered as the Platonic ideal of the black box crunching billions of numbers to produce actionable data. While other banks have multiple risk systems, Goldman has one. One system to rule them all. Turns out everyone else wants one too.
The appeal is summed up by an ex-Goldman senior technologist who worked with SecDB for years. "When Lehman went bust, it took us about twelve hours to work out Goldman's credit exposure to Lehman across each of the entities we'd been trading with. Because every trade that GS had ever done with anybody was recorded in SecDB, we had the data by Saturday night. Other banks were still trying to work out their exposure months - yes *months* - after Lehman went under."
He, and other SecDB progenitors and acolytes, say the system's power comes from its uniformity. "SecDB is about having all your data (time series, trades, security information, emails...everything) in one place, with one way of accessing it," he tells us. "At Goldman, you have one place to calculate the Black Scholes price of an option, and one place to get the discount curve for the U.S. dollar. Everyone goes to that place and everyone uses the right formulation."
It's this that other banks have tried replicating - with varying degrees of success. In January 2006, J.P. Morgan hired Kirat Singh and Mark Higgins, two of the core SecDB programmers from Goldman. In February 2007, Morgan Stanley hired Jay Dweck from Goldman to run a new 'Strategic Analytics, Modeling and Systems Group.' That same month, UBS hired Tasos Kontogiorgos and Jack Yang from Goldman to work on its own iteration of the system. And in November 2014 Deutsche Bank hired Sam Wisnia, the former global co-head of Goldman's strats team, to implement a single risk pricing system for derivatives trades. “Wisnia was brought in by Ram Nayak and Colin Fan to sort out Deutsche’s risk and pricing systems,” a Deutsche insider told us last year. “Under Anshu Jain, Deutsche had a sort of start-up mentality. They were going for revenues and growth, and had a piecemeal desk-by-desk pricing system."
It's a measure of the desirability of SecDB expertise that Singh didn't stay long at J.P. Morgan and Kontogiorgos and Yang didn't stay long at UBS. Singh was poached to set up a similar system at Bank of America in 2010 and Yang and Kontogiorgos moved together to Credit Suisse in August 2012, where they remain to this day. "When I left Goldman I was seriously hunted by another banks to do a SecDB*," says the ex-GS senior technologist. SecDB replicators are hot.
It turns out, however, that recreating SecDB outside of Goldman Sachs isn't that easy. While other banks think they want a SecDB, the ex-Goldman technologist says they have no idea of the commitment required. "When I told the banks headhunting me, "You're going to have to spend $1bn and it's going to take a minimum of 5 years to get anywhere at all" they all got a lot less interested," he tells us."These banks just don't realize how much it cost Goldman to develop SecDB over 20 years."
Many have tried. Many have failed
This might be why many of the banks that have tried the one platform route, appear to have failed. Dweck, for example, left Morgan Stanley in 2011 and now sells violin-shaped swimming pools illuminated in fuschia by underwater cables. Morgan Stanley now claims to have a, 'state of the art cross-asset valuation risk computation platform', but it's not clear that this extends across the firm like SecDB. Neither UBS nor Credit Suisse appear to have firm-wide risk systems and are understood to have taken a more piecemeal approach instead.
J.P. Morgan has arguably come closest to replicating SecDB with Athena, the Python-based platform it began building under Singh and Higgins in 2006. J.P. Morgan declined to comment for this article, but Athena is understood to have been operative at the bank since 2009 and will reportedly be rolled out for access by clients from next year in the same way that SecDB is being repackaged to clients as Marquee.
Equally, Wisnia appears to be making headway at Deutsche Bank. Wisnia was indirectly praised by John Cryan during the German bank's fourth quarter results call. “The new derivatives platform enables us to value and control and manage collateral for a broad range of derivatives,” said Cryan, “It makes us incredibly more efficient and is a big cost saver. It standardizes the way we look at derivatives across asset classes and improves our controls and saves costs too.”
The real question, however, is Bank of America. It is said to have spent seven figures on Quartz, its own iteration of the SecDB system. The bank didn't respond to our request to comment for this article, but there is talk of "issues." Last August, three of the most senior technologists working on Quartz at BAML left the bank. Although Quartz has been rolled out across the bank, the word is that it's not really used. "You talk to people at BofA and they say Quartz is just there in the background," says one senior quant who worked for Merrill Lynch in the past. "It's not integral like it was at Goldman."
For some of the technologists who were at Goldman during the evolution of SecDB, this is no surprise. "SecDB isn't just about the technology, it's about the culture," says one. Goldman insiders say the SecDB platform combines the unusual virtues of extreme transparency and extreme centralization. Engineers across the firm have access to the code underpinning SecDB and anyone at any level is able to pull something out and change it - although the implementation is then approved centrally. "You can always change something and try it out, but you have to push it back to the main version before it becomes official," says one technologist. "It encourages a huge amount of collaboration and transparency. SecDB is an accident of Goldman's collaborative politics." Anecdotally, it's not like this elsewhere: even J.P. Morgan's Athena platform is said to be dogged by bureaucracy which stifles the changes that have allowed SecDB to evolve with the market.
This isn't to say that SecDB is perfect. As we reported last month, detractors argue that the platform is a victim of its own success. Because SecDB originated in the pre-Python era, the Goldman technologists who built it were obliged to develop their own language - Slang. Although Goldman is now encasing Slang in Python and Java wrappers, skeptics say the persistence of the language used only by Goldman will inhibit the bank's ability to hire top technologists from other firm and to leverage open-source code repositories developed by tech competitors.
"Slang and Secdb have lots of issues," says one Goldman technologist. "But when you've got 20 years of dealing with the issues you get very good at it."
While Goldman's 'one system' has had years to iron out its creases, rival banks are only just beginning. They may yet decide it's not worth it. "There's no need for this single position-keeping risk system" says one senior quant with experience at European and U.S. banks. "The speed at which products are traded in today's world makes these single systems less relevant and accurate. Banks want them because they think it will save money, but they don't realize how much they cost to implement."