On one level, the concept of measuring costs and trade quality supposes that there is a robust access to data that is both homogenous and accurate, said James Singleton, Chairman and CEO of Cürex Group, an ECN (electronic communication network). But in a credit sponsored market like FX, measuring costs can vary firm to firm.
On another level, he states, the measurement of cost and trade quality is impacted by the very trading environment and methods chosen by a trader. The impact of last look on an order, for example, is hard to capture. Moreover, the continued use of voice orders, single dealer platforms or RFQs can create market impact before a trade is even executed, which is often ignored or not calculated at all.
“The accurate measurement of costs and execution quality will remain difficult as long as FX market participants stick with past practices,” Singleton says.
TCA provision itself has gone through an evolution. From price comparison across custodians, to comparing manual or electronic execution processes, to ensuring platform liquidity and, most recently, to comparing algos.
Howard Tai, Senior Analyst at Aite Group, said that even though TCA has gone through these different progressions, that doesn’t mean that the majority of market participants have gone through all phases of that evolution.
To some extent, it risks being little more than “ticking-the-box” exercise for compliance purposes, however, a small minority of firms can and do appreciate the true application of best execution, Tai noted.
“Progressive firms do a self-examination that the process to get the best outcome is in line with what they are trying to achieve,” he said. “It becomes a self-examination of the firm’s investment methodology itself.”
According to research earlier this year from Greenwich Associates, traders increasingly turned to FX TCA following the 2013 fixing crisis. Almost a third, 31%, of institutional asset managers now use TCA as part of their trading process. Even among corporate FX traders, 8% now use transaction cost analysis.
“This may not sound like much, but considering that these companies are not financial institutions, it is a strong sign of increasing sophistication by FX end users,” wrote Richard Johnson, VP in the firm’s Market Structure and Technology practice.
Such signs are welcomed and noticed across the community, but it’s still a quote-driven market.
FX is, at the moment, a “hybrid gray world”, said Pete Eggleston, Co-founder and Director at BestX, a TCA provider, and the same rules as in equities and futures just don’t apply.
“You have a lot of business that is still quote-driven, but doesn’t have the same fingerprint around the whole order or lifecycle of that trade,” he said.
Over the last five years, the conversation about what best execution means has itself changed, he added. It used to be that taking the best price among five executing parties for an RFQ was considered a job well done, but that thinking has moved on, Eggleston said.
“People understand there’s a lot more to Best Execution than just hitting what optically looks to be the best price.”
In order to calculate cost, a firm must first decide what it’s measuring itself against, so picking the appropriate benchmarks is the first goal, said Guy Hopkins, Founder of FairXchange, a TCA platform provider and bespoke execution consultancy.
In some cases, this is prescribed, like in the PRIIPS (Packaged Retail and Insurance-based Investment Products) regulation, but in others it requires choosing benchmarks suited to the mandate of the trading desk and the firm.
“Once the appropriate benchmarks are chosen, you then need to source the appropriate market data to be able to compute those benchmarks and your performance against them,” said Hopkins. These two pillars will result in a reasonable estimate of cost. But measuring execution quality is a far more complex exercise. “You are then seeking to understand that cost in a wider context: was the cost reasonable – for me, for my client, for my counterparty? What were the drivers that led to that cost, and what can I do to manage those variables in the future?”
And new analytical solutions are helping FX trading firms in those efforts to measure, record and justify best execution.
FX isn’t wanting for data, but until now it hasn’t really been presented in a systematic way that facilitates understanding and insight into cause and effect, said Hopkins.
“This is true for both traditional risk transfer execution and also more complicated strategies like algos,” he added.
The top eFX market makers, both banks and non-banks, have been using this structured approach to data analysis for some time to manage their businesses. Now there is more widespread adoption of the tools and techniques across a greater variety of market participants to manage liquidity providers, understand algo strategies, timing and sizing of trades, among other factors.
Hopkins noted that although spread remains one of the core metrics to measure performance, implicit costs such as market impact are equally important.
To make meaningful sense of execution performance, he explained, there needs to be an assessment of the relative contribution of four variables: trader actions; strategy used; liquidity the strategy interacted with; and the wider market conditions such as time of day, and currency pair effects, for example.
At FairXchange, that means analysing the different events associated with FX trading alongside the actual fills. The firm’s analytical framework, Horizon, provides metrics that help clients gain insight into each of these drivers. Horizon takes a large dataset of any trading event that can be captured electronically and is timestamped, and compares it to market data at the time.
“Ultimately it means taking that data and building up a picture of what’s happened within those trading events,” he said.
In algo execution, for instance, focus tends to land on the parent order benchmark — performance against the arrival price and performance risk transfer price or TWAP.
And though these are important, they don’t provide much actionable insight into the performance or otherwise of a given algo strategy, said Hopkins: “When you’re looking at that result you think well, is that down to the market or the algo?”
In spot FX, most TCA providers were using the slippage methodology based on the mid-market price of single venues, said Willis Bruckermann, Analyst Consultant at GreySpark Partners, a capital markets consultancy firm.
That has a number of problems that are yet to be solved. For one, a trade in motion is much more likely to move that particular liquidity pool than the global liquidity for a given currency pair, so a measurement ends up being like an “echo chamber”, he noted. “The truth is it’s the mid-venue price, so the mid-market price would have to take on all the bids and offers, and serve all executions that are undertaken globally across this fragmented venue landscape.”
FX TCA providers are now starting to look at aggregating from across a large range of venues, both single dealer platform and brokerage execution venues, which is a new development for FX, he added.
Buy-side players, corporations, and retail aggregators need to take best execution seriously because of shifting market structure dynamics and upcoming regulatory changes, and that means investing the needed resources to attain and be able to demonstrate best execution, according to a recent report from Aite Group.
Senior analyst Howard Tai cautioned however that taking best execution and TCA seriously means looking at the investment lifecycle process from a holistic angle. Without doing so, any analysis is only examining at the moment of the trade itself, or “at-the-touch” execution alone.
GreySpark’s Bruckermann said that at-the-touch makes sense for liquid markets, but for less liquid pairs, once an order gets worked by engaging either one or more brokers there may be information leakage and market impact that moves a price against the order just by indicating interest.
“Asking from the point of execution or order placement would no longer be appropriate,” he said. “You would have to measure price deterioration at the point at which you begin scoping out the market for the trade.”
Bruckermann expects that other methodologies will come into play, and though that process is ongoing, it isn’t mature.
In the FX market, there are those direct costs that everybody sees, like broker and settlement fees, said Tom San Pietro, CTO of trading service, FXSpotStream. But depending on execution style and on which venue trades are conducted, those costs may or may not be apparent. And then there are indirect costs, like market impact.
“If you are trading electronically, price is an important factor, but fill ratios and speed of execution are also key factors that affect your cost of trading,” he added.
As a fully disclosed service, FXSpotStream facilitates a bank’s provision of pricing in processing orders, and is now evaluating the addition of entire algo suites that a bank supports. The team is also working on an analytics suite, FXInsights,that includes some elements of FX best execution: measuring decay, price of execution, latencies and other factors.
FXSpotStream CEO Alan Schwarz said that the service helps clients solve, among other things, the “hard costs” aspect of the execution equation: takers don’t pay to use the service, and makers pay a quarterly fee that costs less the more business
“On the one side, we solve it on the hard cost. On the other side, we solve it by giving people a transparent view that they can then have the discussion of market impact and trading related costs,” he said.
Anonymity and analytics
The pursuit of best execution requires a commitment to evaluation and analysis that includes both pre- and post-trade analytics, says Cürex’s Singleton. On the Cürex platform FX traders can survey the live market in 200 currency pairs, and the streaming analytics shows spread development across its entire book, which is 100% executable. There’s also historical spread analytics and a volatility tool.
MiFID II requires a pre-trade market check that is systematic and embedded in an investment firm’s trading policy and practice, explained Singleton. He added that having the ability to survey the marketplace anonymously is advantageous, particularly when considering the execution of larger sized trades or less liquid currencies.
That’s because streaming pre-trade analytics show when a skew in the market exists – more buyers than sellers, or vice versa. A volatility tool alerts the trader when the market is moving quickly or slowly, helping them choose the best timing for their trading decision.
“We think it is obvious - a better informed trader makes better trading decisions. Combined with robust post trade analytics, the end-to-end use of market and data analytics can improve the trading results of every FX trader,” he said, adding that, at the base of successful analytical solutions that prove best execution, are reliable and accurate data that are not distorted by aggregation, redundancy and non-executable pricing.
Moreover, the very concept of best execution is not a static one, rather it is a continual process that compels FX market participants to identify, use, evaluate and judge different execution alternatives in the marketplace.
On the post-trade side, certain TCA providers have developed helpful solutions that specifically assist FX traders with their mandated reporting under MiFID II, says Singleton. On the pre-trade side, meanwhile, FX traders are struggling with complying with MiFID II’s guidance on obtaining a pre-trade market check that is systematic and embedded in their trading policy and practice.
“Very few firms have developed pre-trade analytics that directly answer the regulatory mandate,” notes Singleton, adding that analysing execution performance on a meaningful basis requires access to lots of hard to obtain data.
Some of the metrics that would be useful to measure true execution performance would include order fill rates, order misses, venue hold times, post-trade market impact, price slippage from time of risk origination until time of order execution, and counterparty performance.
“Getting access to some of this information is difficult – analysing it is challenging for investment firms without adequate human resources or bandwidth,” he said. “And some of these metrics depend on access to precise time-stamped data. Without such precision, a supposed analysis is not worth much and could actually be misleading.”
To more optimally manage the key issues that drive performance, trading firms need to seek to quantify as much as possible, said Guy Hopkins.
“If they can see where or when they are having demonstrable market impact they can then do something about it,” he said.
Hopkins added that this is why it is so important to attempt to differentiate between the different drivers of performance, using market impact as a case in point.
“If a client is using an algo and the market moves away from them, was the strategy responsible, or was it just unlucky? It’s very hard to tell without analytics to look at the microstructural performance,” he said.
Similarly, for traders to be able to demonstrate that they consistently add alpha by timing when to trade is enormously valuable for the individual concerned, for the desk, the firm and its clients.
BestEx’s Eggleston said that the “justification” piece is becoming more important among buy-side firms, which execution desks need to do both before and after the trade.
“That justification is not just to their own boss or compliance group, or best ex committee,” he noted. “What we’re seeing more is when asset managers are trying to win mandates to raise their AuM, more and more the asset owner is asking to see evidence of their best execution process.”
That means evidence of how best execution is being monitored, and buy-side firms are now seeing from the asset owners’ RFPs explicit references to detail how that’s being done independently.
“It’s no longer adequate on an in-house basis. They need tools and to have an independent entity or technology firm to do that,” he said.
And this kind of change in the industry reflects an ongoing evolution in the drivers for needing TCA and execution analytics.
Aside from regulatory or compliance driven factors, the performance aspect is where quantitative research in market microstructure is starting to embrace next generation technologies such as machine learning to deal with data. Market microstructure is the term used to represent the way exchange happens in markets, and research into it generally looks at how the functioning environment and processes of markets affect a variety of factors such as costs, prices and trading behaviours.
For example, machine learning algos can be trained to figure out when a certain venue delivers a good result given the time of day, a philosophy that can be applied quite broadly across the execution process.
“We are really keen on smart data, not big data,” says Eggleston. “You can have all the data in the world, but unless you can interpret it and actually draw conclusions from it and take action, there’s no value.”
Hopkins believes that you can’t begin to optimise execution performance without understanding market microstructure and how it affects trading.
Top market makers have been using analytics and related techniques for years to manage their businesses, because that has been the optimal way to drive their revenues - it is largely a commercial decision as opposed to a regulatory one, Hopkins said.
“In our discussions with market participants, on the buy-side in particular, there is an evolving awareness that this sort of analysis provides enormous opportunity for driving the next wave of innovation over the next five to 10 years; not just in satisfying the all-important regulatory obligations but also to gain a competitive advantage,” he said.
This doesn’t, however, mean people need to deploy complex, black box quantitative models that are difficult or impossible to understand.
“We believe this is counterproductive, as it just moves the complexity from the trading tool to the measurement tool. It is essential that people are able to understand the information they are presented with, both so they can act upon it and explain it to other stakeholders, whether it be risk and compliance, management, regulators or their clients,” he said.
“TCA is the first step along a ‘grand path’ of evolution with analytics,” said John Crouch, CEO of Ideal Prediction, an independent trading analytics and data science company.
Ideal Prediction provides bespoke TCA analysis based on the client’s own as well as sourced data. Its platform works on a Measurement-as-a-Service basis for FICC.
Those measurements are then used to determine actions to improve P&L. For market makers, that could mean optimizing client tiers, for market-takers, it could mean how to route flow to the best counterparties, and for ECNs, results could include improving volumes and client experience.
Crouch is one of the first to admit that there is “mendacity” in the machine learning space, and that the most important consideration to clients is whether the technology is actually performing the optimizations that firms want to get done.
That process begins not with technology, but with the definition of best execution for the firm.
“Having a systematic approach to evaluating your goals is essential, but you first have to start by defining your goals,” he said. “The market is starting to understand that there is not a single best execution. There are going to be multiple ways to measure execution and each of these methods is useful for a different type of client.”
Then, it’s a matter of accumulating as many independent variables as possible, which in FX are factors that drive performance such as liquidity provider selection and what kind of algo is being used.