Considering that FX algos burst onto the mainstream the better part of a decade after their more popular counterparts in Cash Equities, the relative rate of adoption for FX TCA has certainly been noteworthy. At an industry level, we are witnessing an increasing appreciation of the need for more comprehensive transaction performance data. Coupled with advances in data science, a regulatory backdrop focused on transparency and an increasing focus on best execution, it has led providers of TCA services to invest in increasingly better technology and innovative solutions.
The evolution of customer needs
FX TCA has come a long way and as it continues to evolve, so do the needs of buy-side traders and their overall roles in the execution process.
Static post-trade reports are giving way to data delivery via interactive portals, pre-trade models with market colour information supplement traders’ decision making processes and real-time TCA is adding a new dimension to analytics. At the same time, market practitioners analyse the factors that influence trading performance by delving deeper into the data at a micro level.
Clients that have fully embraced TCA are now looking beyond traditional high level metrics such as implementation shortfall, risk transfer comparisons and reversion statistics, to name but a few. They are increasingly focusing on bespoke data requests that allow them to normalise information across all their trading partners, ultimately helping to streamline the data analysis effort. Some institutions will at a minimum have their own quantitative analysts while others may partner with independent TCA data providers.
Regardless of the means utilised, the “big picture” goals remain broadly the same:
- Better understand outcomes of trading decisions
- Improve execution performance
- Introduce a data driven element in the flow allocation process
It is therefore evident that the role of buy-side traders is in part being shaped by their quest for data, in that they are becoming more and more proficient in quantitative data disciplines.
While we believe that traders need TCA tools to inform and empower them, we must also recognise that TCA is not a “one size fits all” discipline. Each trader will have their own set of priorities, desired execution outcomes, and their TCA needs may differ widely. In fact for some, a single well-chosen metric may be all that is required to effectively assess their algo execution performance. As such, the varying degrees of adoption make the provision of bespoke TCA solutions equally as important.
Real-Time enters the fray
An area of exciting innovation is that of Real-Time TCA. Solutions in this space offer clients the ability to evaluate trade performance “on the fly”, enabling them to observe and analyse execution progress vs. target benchmarks while their orders are trading.
The immediate feedback offered by real-time TCA tools, provides invaluable transparency and accountability to clients across the spectrum of TCA engagement. The wealth of available information will typically range from high-level order statistics to market-level fill data and from execution venue breakdowns to market data execution profiles, all fully interactive. These would be supplemented with information relating to the trading decisions, including algo strategy amendment sequences.
Finding patterns in your data
Under an exploration lens, TCA can be seen as edging more towards the creative. This is where the data at hand can transform into new and un-expected insights which can elevate trading behaviours.
An obvious but interesting point in the TCA debate is that no trade is like another. Each has its idiosyncrasies, ranging from execution goals to algo selection, market conditions and market reactions. An interpretation of performance in context is much needed in this case. A trader may have a wealth of metrics on past individual executions, but still find it challenging to confidently calibrate an expectation performance for their next trade. The fact of the matter is that there is a lot of noise in TCA data.
However, when looking at the trades at a portfolio level, patterns and commonalities between them start emerging. Some can be spotted by an experienced eye, some can be revealed by a quality portfolio-level TCA solution and still others can be explored through machine learning techniques. The application of machine learning is coming to the fore in execution performance analysis and can help traders create a representative peer group for their executions in similar market conditions. Use of a sophisticated data-driven approach at a portfolio level has the potential to close the feedback loop between expectations and outcomes with the right level of scientific rigor, all the while complementing the expertise of traders.
Pre-trade, the computational methods at hand can embed insights from trader experience with tick-by-tick market analysis, and the dynamics of key decision making metrics. This can re-frame a vague heuristic such as: “usually use an opportunistic algo in a trending, fast moving market when liquidity is good” to a model output indicating: “75% confidence for the use of a specific opportunistic algo with an expected performance of +0.1bps vs. arrival mid given the real-time liquidity, spreads, volatility, market trend and bespoke additional factors”.
Post-trade, the assessment of the efficacy of their forecast vs. achieved performance should become part of the process and ultimately reinforce the forecasting tools employed.
Art or Science?
In conclusion, TCA is not a means in itself, but a tool to be employed as part of the trading process, for performance analysis and creative exploration alike. The wealth of data techniques and models available in the TCA space are making the discipline an increasingly scientific one. How these models are employed, leveraged and improved remains in the hands of the market practitioners.
An exciting journey ahead for the buy- and sell-side alike.