It’s no secret that trade or transaction cost analysis (TCA) has become one of the most important components of the buy-side trading desk. In light of market structure changes due to increased regulation and the all-encompassing endeavour for best execution, it has never been more critical for traders to get to grips with in-depth analysis of exactly what has happened with each and every execution.
The idea is that with effective TCA processes, not only can traders confidently showcase to regulators that they are indeed meeting best execution requirements, but they can also gain once-shrouded insights into broker routing decisions, algorithms, and trade performance across the entire order lifecycle. Armed with such knowledge, buy-side trading desks assert that TCA has had a quantifiable impact on execution performance and, in some cases, produced alpha in trading behaviour.
Despite its impact on regulatory compliance, Daniel Nicholls, head of trading at Hermes Investment Management – winners of Trading Desk of the Year in 2018 and Mid-Cap Trading Desk of the Year in 2019 at The TRADE’s Leaders in Trading ceremony – says TCA is not a box-ticking exercise but an ever-evolving process, and if you want to connect the dots, you first have to collect the dots.
Hermes has honed its TCA process with a dedicated and impartial TCA team that works very closely with the trading desk. It has separate oversight and is part of the investment office reporting to the best execution committee, but remains key for the trading desk in terms of leveraging TCA for impactful results.
“There is more demand now than ever before in investment management for increased transparency at each stage of the order execution process,” Nicholls says. “Documented proof of skill, accuracy and integrity in trading are now necessary and this analysis has become a tangible asset on the trading desk. It is an evolving process which looks to identify and scrutinise any patterns delivering the best and worst executions.
“We don’t expect our traders to have a crystal ball to predict the future, however, they are now armed with strong historical analysis for guidance. TCA gives our traders the information and understanding that they need to engage with brokers about how we are using their tools and why we are getting these results. We want our traders to take the risks they choose to take. We are trying to eliminate the risks we don’t want to take where possible.”
The critical step
Every order that hits the trading desk at Hermes must have a pre-trade benchmark that is documented by the trader. Subscriptions have a benchmark specific to the time the money arrives in the fund, and if the order doesn’t have a time-implicit target, the trader notes this as ‘implementation shortfall’.
It’s at this stage, the execution phase, that the order ‘belongs’ to the trader. Be it high- or low-touch, ownership of the order does not get transferred to the broker to achieve best execution, meaning it’s now up to the trader to select the most effective means to beat the benchmark. At this critical step the trader has the ability to create alpha in their execution through various means such as pre-trade market impact analysis, market timing, broker selection and algorithm choice.
“It is necessary to try new brokers and new algorithms to further improve our processes,” adds Nicholls. “If we didn’t do this, we would never improve the quality of our executions. To misquote Einstein: ‘Poor execution is doing the same thing over and over again and expecting a different result’.”
When trading low-touch market impact orders, timing is perhaps most readily the means by which the trader will generate alpha. Like many institutional asset managers, Hermes does not have proprietary algorithms, but uses broker algorithms for orders that constitute a low percentage of the daily volume. Orders must be categorised into high- and low-touch, and a low-touch order should never be given to a high-touch broker due to higher commission charges. It’s very likely they would use an algorithm anyway.
For high-touch orders, traders at Hermes aim to cross blocks by trading against the natural liquidity with scraping networks or high-touch brokers, both of which demand higher commission. If the market environment is tough and no natural liquidity is available, working with a broker that trades the most in the security is considered the best way to save the cost of the spread.
Like-for-like
Data is key to the TCA process, and it is widely agreed that the more data, the better. At Hermes, the order management system feeds extracted tick data from each transaction across asset classes to the TCA platform. The measurable parameters span a six-month period or longer, as analysis of a period less than this could produce false positives or outliers due to volatile days in the market that may cloud any drivers of performance.
This can be a headache. But Hermes aims to address the mountains of data and numerous benchmarks required by streamlining the data and asking specific questions on whether the trader’s decisions, algorithms, brokers, venues, and timing helped the trading desk beat the benchmark.
“We use the insights gained from our analysis as a tool to help replicate good trading behaviours and reduce any bad ones. If the data is measurable and meaningful, it should be analysed. TCA must of course be understandable and actionable otherwise its purely academic, which would clearly be less helpful for the traders,” Nicholls explains.
Hermes uses this information on the quality of executions it receives from brokers and their algorithms to inform routing decisions in the future. Nicholls adds it is imperative to understand the nature of the flow that the desk is giving to a broker by considering certain factors such as the percentage of the average daily trading volume, spread size, expected impact, and volatility when orders are routed to them.
Comparing like-for-like performance is another important part of the process, therefore all data points are separated, and similar brokers and flow are grouped together. When comparing similar brokers, the analysis goes beyond which broker out- or underperformed the benchmark the most. All contextual metrics are measured, including spread capture, fill size, momentum, impact, and reversion, for example, in order to paint a fuller picture of broker performance.
A weighted ranking score system is then produced to reveal the broker winners and losers in the analysis, before trends or potential outliers that are driving the ranking are identified. For the losers, Hermes consults with the brokers to verify the results of the analysis and work towards preventing any continued poor performance. If this low scoring occurs over a small sample size, the broker is placed on a watchlist, which is closely monitored until the results are either proved or discredited. A desk review is also undertaken in circumstances where there are difficult market conditions or flow.
The process for high-touch and broker analysis is similar to the analysis carried out for algorithms, but Hermes compares algorithms side-by-side for actionable insight not only for selection, but for informing any changes made to the algorithm itself. If the algorithm is tweaked in partnership with the broker, Hermes diverts flow there to monitor performance closely and to ensure the adjustments are having the desired impact. This is considered to be of immeasurable value to the buy-side trading desk in terms of understanding how algorithms operate in certain market conditions. If a trader is missing a benchmark in volatile market conditions, for example, the TCA will spot this and the desk can monitor the results.
Daniel Nicholls, head of trading, Hermes Investment Management
Refining workflows
Typically, venue-level performance metrics have been considered as being ‘nice to know’ for buy-side traders. This is now shifting as venue analysis is increasingly being deployed as an explanatory factor for macro-level performance evaluation.
“We don’t dictate the venues which brokers choose to access as this is up to their own smart order routing logic to decide,” Nicholls explains. “It is important though, for us to understand why and which venues are selected, as this impacts execution performance. This in turn effects which brokers and algorithms we route to.”
Hermes looks at venue performance in two ways. If a broker or algorithm is missing intended benchmarks, or if a venue is experiencing adverse price movements relative to the rest of the market, then a deeper analysis is carried out. The venues are also categorised by type, such as lit, dark, systematic internaliser (SI), electronic liquidity provider (ELP), or request for quote (RFQ), to allow the trading desk to outline broker routing behaviour.
For example, forensic analysis can show Broker A is not interacting with SI liquidity whereas Broker B is, and they are outperforming. In fact, Broker B is trading with SI liquidity and achieving larger fills than the rest of the market, having a lower impact and capturing the spread. If this holds over a suitable sample size, there is a credible case to defer this segment of the desk’s flow to Broker B. If the sample size is low, it is added to the watchlist until the trend is confirmed.
There are several actionable steps that the trading desk can take following intense venue analysis. If the analysis shows high reversion in a dark venue but large blocks are being filled, then Hermes doesn’t move to shut it off, but to avoid the high reversion on small size prints the desk may request that the broker applies a minimum execution size to avoid that in the future. Likewise, if other brokers with similar flow are accessing liquidity in a better way then Hermes will divert flow.
The desk also adjusts the smart order routing hierarchy to avoid recurring bad performance. Outlining one such instance, Nicholls says the trading desk had a series of liquid low average daily volume US names that needed to execute at the open. These were intermarket sweep orders, spraying all venues, NBBO (national best bid and offer) or not, and Hermes found that too much was being executed on the regional US venues at suboptimal prices. The trend was backed up by the desk’s TCA in-depth analysis, indicating that the asset manager had to act.
“Without impacting the time to completion, we wanted to avoid the regional US exchanges,” Nicholls adds. “Armed with this information, we were able to highlight the need for a solution with the broker and amend the smart order routing hierarchy to fill these small liquid orders in dark pools first. Capturing half the spread by executing at mid and executing more quickly than aggressing the lit where spreads are widest immediately post-open.”
The final actionable step upon reviewing venue analysis via TCA is to turn the venue off. This is considered an absolute last resort for Hermes – turning off a liquidity source is no easy decision. But if a venue’s performance is unsatisfactory, it will be avoided through routing decisions, smart order routing hierarchy adjustments via the broker, or trading strategy selection, which is the preferred choice for Hermes.
Despite the focus on venues, brokers and routing decisions, it is also important to measure performance across the entire order lifecycle, alongside the actions of traders, and use this as an opportunity to refine workflows. Considering slippage and time accrued between the trader receiving the order and placing it out the market, for example, could reveal other areas that the trading desk needs to improve on. Looking at the opportunity cost of not trading in the days leading up and after the execution can also prove to be valuable to portfolio managers.
Huge potential
Like many buy-side firms, it’s not just equities performance that Hermes is analysing. Fixed income TCA is widely-considered to be the most difficult asset class to efficiently measure performance, and a lot of time can be spent, or wasted, verifying the results of analysis against external pricing sources.
Nicholls explains that the benchmark his trading desk is primarily measured against for fixed income is the far touch of the spread at arrival to the trader and execution, but this is considered to be inadequate benchmark data, lacking standardisation and transparency on the benchmark price. Being aware of these limitations means that Hermes relies heavily on data it truly trusts.
Similar to the process for equities, broker performance is measured against the benchmark and categorised into those with the most and least orders inside the benchmark spread for a deep dive into the figures to reveal each broker’s performance. Embodying RFQ to calculate hit rates and quit rates, for example, can be valuable input when assessing broker quality and identifying those brokers that are perhaps often seeing the trading desk’s intentions but failing to deliver.
“We continue to scrutinise performance to the best of our capability whilst leveraging alternative data points to review execution quality. Until we see the development of a consolidated tape with a consistent data collection process and enriched data points to identify addressable liquidity, which needs to be driven by the regulator, this is the best we can do,” Nicholls adds.
For foreign exchange, Hermes has found limited use in deploying TCA to analyse brokers as the desk’s FX flow is primarily executed at the mid to save the cost of half the spread, with standardised charges across its brokers. But TCA is applied to FX to ensure that brokers are charging the costs agreed, reviewing which brokers maintain the fee on unusually large or small orders and then using this as intelligence on where to direct those types of orders going forward.
The insights generated from effective TCA as outlined above are then used to build in-depth scorecards, ranking the performance of brokers, algorithms and venue accessed for every asset class. The broker scorecards can also reflect additional statistics such as settlement rates, and less quantifiable metrics like quality of service, are also scored.
There remains a huge amount of future potential for TCA once more data becomes available. Nicholls concludes that at the moment, the market tape reflects where and when executions took place. But if the tape reflected all the liquidity available, particularly that which is posted but unexecuted, his trading desk would be able to conduct further in-depth analysis, which in turn could generate alpha.