Big Data For Execution

The term “big data” has become an ubiquitous cross-industry term for describing vast amounts of data, the implications of storing and analyzing such data for organizations, and the technology to do so. It has also generated an insatiable demand for data scientists, a profession predicted to grow enormously over the coming years. It is even described as “the best job in America” in this recent article.

Transaction Cost Analysis (TCA), a speciality of QB, is essentially “big data” for execution. When done well, TCA can give fascinating insights into trading, uncovering key characteristics not known before. A key purpose of TCA is to compare different execution methods to see which one performs the best.

In addition to comparing methods of execution, it can also provide insights into other aspects of the order flow. For example, what if you discovered that you generally execute better in the morning than the afternoon? Or if you have significant positive or negative alpha in your flow, for which you could either speed up or slow down / delay your execution?

Like any data set, sample size can be a challenge for those with less data to analyze – larger sample sets are needed to be statistically significant and provide conclusive evidence. However, we don’t believe those with smaller sample sets should be put off from pursuing TCA. Eventually, enough data will be collected, the value is in gaining insights along the way ¬†that when combined with intuition and logic, could help improve the execution process.

Machine learning is highly applicable to TCA – this is an area QB is actively working on. Developments in this space will reduce the need for human analysis, making comprehensive reviews of ever-expanding data sets more scalable. Instead of manually searching for patterns and insights, intelligent observations of value are produced and flagged automatically. Have no doubt – big data and machine learning have a huge future in our industry and the evolution of trade execution. Large, unstructured datasets will be better tackled without the time-consuming normalization process needed today to ensure that the data for TCA is reliable and comparable. With these advances, TCA will be a more accessible, and essential, tool for traders, portfolio managers and investors.

Quantitative Brokers
New York
October 5th, 2017