TCA Benchmarks

TCA provides the quantitative framework for Best Execution policies and procedures. When we say quantitative framework, what we mean is that TCA compares properties of the order with properties of the market. Many of these properties, such as performance versus a VWAP benchmark, are continuous numbers that we can do things with like compute value-weighted averages, inter-quartile ranges, standard deviations, and other statistics. In our example, we refer to a benchmark, in this case the VWAP benchmark. Benchmarks in TCA are computed from the market data and then by comparing those benchmarks with order properties such as the average price of the order, we can measure performance by computing the relative difference between the order price and the VWAP and then dividing by the VWAP.

A question that arises often is “what benchmark should I use?”. To get the most value out of TCA, it is important to begin with a business question and once the question has been defined, choose the benchmark that best addresses the question. If the instructions to execute were to match the VWAP over a period, then choosing a VWAP benchmark to measure performance is better than a benchmark such as the open price. A key point here is that we can ask more than one question of our TCA performance data. This means we may employ multiple benchmarks in our analysis. We can also improve benchmarks to address our specific questions.

To illustrate, I will relate a recent travel experience. In early Feb, I went to Montréal to visit my son. When I arrived in Montréal, I had a short walk to my hotel. When I stepped outside, I could feel the cold. A quick check showed the temperature to be -28 C. My benchmark was -28. But it was also windy making it feel even colder. On the weather app, it also showed a second benchmark, called wind-chill. The wind-chill combines the wind speed with the temperature to compute a new benchmark that considers environmental factors. (In the summer the humidex is often used to indicate how hot it feels.) The wind-chill was -42 C, and because of this, I had to take extra precautions to manage my risk (of frostbite) that I wouldn’t have had if the wind-chill was the same as the temperature. Both benchmarks tell us something: the temperature of -28 tells how cold it is, while the -42 tells us how cold it feels. If our concern is how much a bridge expands or contracts, we will use the temperature. If we want to know how to dress before going outside, we will use wind-chill or humidex.

TCA is completely analogous in that we have environmental factors such as volatility, depth and spread. We can create new benchmarks by combining existing benchmarks with things like prevailing liquidity to get a more precise answer to our question. For example, we can use the stocks prevailing liquidity characteristics to adjust our performance – we would expect our order filled price to be close to a price benchmark like arrival price when volatility is low and higher as volatility increases, so we can adjust our performance based on the volatility which then allows to do more of an apples-apples comparison.

Another example is to incorporate volume-time as is done in a pwp benchmark. While many performance metrics use a single benchmark, some use multiple benchmarks. A simple example is computing delay costs or looking at the time evolution of adverse selection.

We also have a concept of ‘business’ benchmarks that can be functions of multiple ‘market data’ benchmarks. For example, the ‘decision’ price for an order may be a function of several benchmarks – for example, previous close, open, or mid-point. This approach gives us flexibility to generate bespoke benchmarks that can more precisely address business questions. Used in this manner, TCA provides relevant analytics to derive insights. Begin by first defining the problem, then deciding which benchmarks to use, we will make the best progress. As Einstein said: “If I had an hour to solve a problem, I'd spend 55 minutes thinking about the problem and 5 minutes thinking about solutions.”

TCA is a framework which means it can be used in many ways. This also means it can answer many questions. Performance metrics are computed by comparing order attributes with benchmarks derived from market data. We are not limited to a single benchmark, but we also don’t need to use them all in our Best Execution analysis. Each benchmark provides a different perspective. A benchmark like VWAP may help you determine how well your broker matched the instructions for a VWAP order, but a benchmark like the open may help you decide whether to participate in the opening auction.?

What benchmark(s) do you use? What benchmarks would you like to see? We would love to know as we roll out our TCA platform as part of FactSet’s Best Execution Analytic for Smarter Trading (BEAST) service. We compute many benchmarks already but are always interested in other’s thoughts.

Please feel free to reach out directly with any thoughts or comments.

Chris Sparrow?[email protected]