"High frequency quotes and orders are difficult to manage, very unwieldy, and amount to a lot of redundant cost both for the exchange and for customers we provide it to," notes Mark Schaedel, vice president of data products at NYSE Euronext. For instance, the NYSE adds 750 million records a day to its repository. "1010data lets us focus on the licensing and distribution business, rather than the data management portion." 1010data manages the data repository and the analytics engine. NYSE's data services website communicates with the 1010data backend infrastructure through an application programming interface.

NYSE currently offers 30 products on this infrastructure. The core product is its TAQ database for North American trades and quotes. "That allows investors, researchers, academics, algorithmic traders and quantitative traders to have a comprehensive picture of everything that took place during the previous day," Schaedel says. The most popular use of the TAQ database these days is to optimize algorithmic trading strategies. "They can run through a number of what-if scenarios based on the previous day’s trading to see what would have happened if they’d taken a different strategy or approach, they can run through various P&L strategies and then optimize those strategies for the next day’s trading," Schaedel notes. Customers can enter the NYSE website and define a custom query, such as, show me all the trades and quotes for a certain sector during a range of dates.

NYSE Euronext also uses the analytic tools internally, for instance, its research staff is constantly measuring the performance and the competitive position of the exchange and executive staff receive frequent reports on performance and business metrics.

1010data has an interesting history. One of its founders in 2000 was Joel Kaplan, who had been in the first analytical proprietary trading group at Morgan Stanley, according to David Frankel, vice president at 1010data. "His was the first group that did real black-box trading where the computer decided what to trade and executed it hands-off," Frankel says. "We all know this is very demanding. It was all about sifting through the transactional histories, making sense of the data and coming up with strategies, and executing those strategies in real time. Joel Kaplan’s group was doing 2% of the daily volume of Stock Exchange -- one out of every 50 shares was his execution." The group even wrote its own implementation of the programming language APL as close to machine level as possible to make it ultra-efficient.

1010data is not a real time system, it’s designed for historical analysis. Customers export data in batches once a day or once an hour, whatever frequency is required. Frankel says the analytics tools can work across normally incompatible data sets.