Some thoughts from NCFX on using conflicted data to measure FX trades.

The problem with conflicted data, or why you can’t mark your own exam papers…

The last few months have seen a change in pace for the FX transaction cost analysis (TCA) business.  Financial firms have recognised that steps must be taken to understand FX costs ahead of the various regulatory deadlines falling due in January 2018.  Products are popping up left and right, but are they window dressing for brokers, some kind of execution analysis, or TCA?  It’s actually very easy to tell the difference.

The TCA battle is being fought over pie charts, graphs and calculations, but we see this as all being completely irrelevant if the source data is not independent in the first place.  A good calculation based on bad data is entirely worthless after all.  One of our competitors put it best in a blog post of March 2016:

‘Even if the data you are provided with is accurate, you still need to compare it to data from other providers. The provision of a level playing field, upon which performance can be objectively measured and compared, can only be provided by an entity which has no vested interest in any of the results. No desire to increase trading volumes or market share, no desire to cross-sell other financial services or products, no Chinese Walls at risk of getting breached. Simply sitting above the market and allowing market participants to compare apples and apples, without trying to sell any apples or anything else for that matter’.

So, the goal of TCA is to independently compare apples with apples.  This matches the regulatory aims of the new TCA rules, which require a simple, absolute calculation of cost based on independent, ex-ante data.  The penalties for misreporting costs are unknown – new regulation will mean this isn’t just a box ticking exercise. Yet far too little attention is being focused on how the use of conflicted data hides transaction costs. We would not expect students sitting their final exams to mark their own papers nor award their own degrees.

The quality of FX execution is more often than not judged against reference rates that have been directly supplied by the venue or liquidity provider handling the trade. The same firms that are competing for market share of the daily FX business. To counter this point, some providers have begun to obtain reference data from external sources. However, if the source of this external data is an alternative trading venue that is known, then it is easy to influence this external reference rate to flatter your execution.

The widespread use of conflicted data frustrates efforts to achieve transparency on how financial intermediaries get paid, and how much. If the data used to measure costs is conflicted, it does not matter who is doing the analysis, or how pretty the graphics.

The problem of conflicted data is evidenced by the experience of an NCFX client who wanted to understand their transaction costs. Using the ‘conflicted’ approach they derived a mid-rate from the bids and offers received from their counterparties – never less than 7 per trade.  This approach led to them to underestimate their transaction costs by an average of US$ 100 per million per trade, which resulted in a misstatement of annual transaction costs of over US$ 10 million dollars.  It’s clearly not sensible to risk this kind of error, so removing any potential conflicts from the reference data is key.

So, if using a mid-rate constructed from taking their own best bid and offer is likely to be wrong, what mid-rate should be used?  There are, after all, so many, even on the same platform. In the example overleaf we highlight some of the problems faced by market participants

Consider the following simple examples:

Client 1 and Client 2 both sell GBP/USD through the same FX platform. The platform aggregates prices from bank a, Bank b and bank c. The Platform also provides TCA based on observable trades and on its calculated mid rate.

Client 1 has credit with bank A+B                                           

Client 2 has credit with bank B+C


Bank A quote: 1.2150–56

Bank B quote: 1.2152–58               

Bank C quote: 1.2154–60


Client 1 is quoted: 1.2152–56 (mid 1.2154)                              

Client 2 is quoted: 1.2154–58 (mid 1.2156)

However, The Platform observes all quotes, GBP/USD is 1.2154–56 (mid 1.2155)


Client 1 sells GBP/USD at 1.2152                                               

Client 2 sells GBP/USD at 1.2154


Possible TCA results:

Client 1 against own mid – 3 Pips                                           

Client 2 against own mid – 3 pips              

Client 1 Against platform mid – 5 pips                                  

Client 2 against platform mid – 1 pip

Client 1 on Platform observable data – 0 pips                   

Client 2 on platform observable data – 0 Pips

So, which TCA result is the correct one to use?

The answer simply is none of the above. in each case the data used is conflicted. Here’s why:

  • If either client calculate TCA by using their own mids respectively, they simply reflect their bid/offer spread and do not capture the cost of skew in terms of the underlying quotes.
  • By calculating TCA by using the platform mid they are only comparing the aggregated pricing of 3 bank marketmakers. This is hardly representative of the whole market. they will not capture the market impact of using the platform If by comparison it happens to be particularly latent in comparison to the wider market.
  • Calculating Transaction costs by using data observed on the platform, (The respective best bid each client sees), only CONFIRMS THAT the client dealt on the best price that was available to them. THIS IS CIRCULAR AND MISREPRESENTS THE ACTUAL COST OF TRADING.

 What then is The correct approach?

The only way that the clients in our example can objectively assess the cost of execution on the platform is to measure the trade outcome against an unbiased external market reference price. however, It is critical when choosing external measurement data, that the reference price selected is an aggregation of the whole market and not formed from a small number of trading venues whose prices are easily manipulated against our clients interests.

Uniquely, NCFX offers truly independent, forensic Transaction cost analysis that is completely objective. All NCFX TCA results are calculated impartially by using highly aggregated real-time data from multiple sources to capture the full dispersion of live pricing at any given moment. NCFX create and publish data in real time and store all outputs for future audit purposes.

In an equal world, we might expect market impact cost to have an equal probability of being positive or negative, but in the world of trading as principal to principal, market makers have the option of whether to trade or not. Adverse selection ensures that the skew always goes against client’s interests. If it did not, the market maker would not be profitable.

As demonstrated in the example above, to achieve an unbiased, objective measure of transaction costs, we need to be able to use a proxy of the best market mid-rate available at the time of execution. The best approximation of this is to re-create a normalized rate, from a variety of sources. Using data from numerous venues increases the chances of capturing a large dispersion of best bids and best offers. This variety helps to offset the skew embedded in prices from individual venues at any given moment.

As we saw with the manipulation scandal concerning the use of LIBOR and WMR, reference rates that are constructed by the price setters, who are pricing against the interest of their clients are open to abuse. This weakness is structural. The issue arises because the reference rate is being generated by the same people that handle the execution, which is then used to measure the execution quality. The case of WMR hit the headlines because it demonstrated collusion among the price setters.

The imaginary students in our exam (the WMR abuse) were cheating by sharing their answers. But the students don’t need to collude to cloud the integrity of the examination. As long as the students are producing their own reference rates there will always be ambivalence in the results.  As we saw in the example cited above, in FX this ambivalence can hide the true extent of costs and flatter TCA outcomes.

In an environment when regulatory scrutiny has never been greater, using conflicted data seems a risky strategy.   By choosing to use independent data, generated ex-ante, and indirectly, market participants offer themselves a greater guarantee that their transaction costs are fully identified. What is more, objective data offers market participants certainty that the more exacting standards of regulatory cost disclosure will be met.

Xavier Porterfield CFA

Head of Research

Mobile: +33 (0) 6 37 72 16 06

100 Pall Mall, London, SW1Y 5NQ