In 1967, Robert McNamara’s war command was generating impressive data. Body counts were rising. Kill ratios were improving. By every headline indicator, the United States was advancing. The problem — which took years and enormous cost to acknowledge — was that the measurement had become a substitute for the outcome. What could be counted was counted. What could not be counted — strategic coherence, institutional will, the structural conditions on the ground — did not appear in the reports, and therefore did not officially exist. The sociologist Charles Handy later named this the McNamara Fallacy: the tendency to measure what is easily measurable, disregard what is not, and mistake metric improvement for progress.
The FX industry is not fighting a war. But it may be making a structurally similar mistake — and the metric it is optimising is TCA. As it celebrates the maturation of Transaction Cost Analysis, a more uncomfortable question is emerging: are we optimising the metric while the underlying problem quietly deepens?
The feature: Why Data Analytics and TCA Are the New Currency in FX Trading, which was published last year in e-Forex reflects a genuine industry consensus. Over half of the 400 FX professionals surveyed globally are actively investing in data analytics and execution measurement. TCA has evolved from regulatory obligation into competitive differentiator. By every headline indicator, the industry is winning the execution quality battle — spreads are at historic lows, fill ratios are improving, and venue analytics have never been more sophisticated.
So why, in conversations with institutional execution desks, does a quieter and more troubling narrative persist? That large-clip execution is getting harder, not easier. That adverse selection events are more frequent, not less. That the market increasingly looks liquid at the point of enquiry and isn’t at the point of fill.
The answer, I would argue, lies not in a failure of TCA — but in a fundamental limit of what TCA was ever designed to measure.
1. The metric and its blind spot
TCA is an audit of what happened. It measures slippage against arrival price, spread capture, VWAP deviation, and fill ratios — all anchored to the moment of execution. It is, by design, a retrospective instrument.
What it cannot measure is the structural environment that preceded the execution. It cannot quantify the probability that your liquidity provider’s pricing engine had already detected the directional intent of your order flow before you sent it. It cannot record the depth that was displayed but non-actionable — the quote that existed until 400 microseconds before your order arrived. It cannot capture the difference between available liquidity and committed liquidity in a market where AI market-making has made that distinction commercially significant for the first time.
The GBP flash crash of 7 October 2016 is the cleanest illustration. At 00:07 BST, sterling fell approximately 6% in under two minutes during thin Asian session liquidity. Pre-event TCA for every institution that had executed in the preceding hours showed excellent spread capture, strong fill ratios, and unremarkable slippage. Not one of those reports contained a signal that the structural conditions for a liquidity cascade were already in place. The event did not appear in the retrospective data until it had already happened — by which point the damage was done. TCA was not wrong. It was silent.
The e-Forex outlook observed that “tight spreads generated by overly aggressive skews may not necessarily be in the favour of the broker.” This is a careful acknowledgement of a structural reality: the number that TCA reports as a spread improvement may be, in practice, the price of informing your counterparty of your flow characteristics.
2. What AI market-making actually did to execution quality
The spread compression of the past decade is widely celebrated, and on the surface it should be. But it is worth asking what drove it. AI-driven market-making has simultaneously achieved two things that appear contradictory: tighter top-of-book pricing, and greater sophistication in detecting and responding to informed order flow before a fill is committed.
This is not an accusation — it is the rational commercial behaviour of any liquidity provider managing inventory risk. But the consequence for institutional participants is a cost structure that has shifted from visible to invisible. Spread costs are down. Adverse selection costs, market impact on larger clips, and the systemic erosion of depth during structural volatility events — none of these appear cleanly in a TCA report, because TCA is calibrated to the execution, not to the conditions that made the execution inevitable.
Esteban Mora, Chief Commercial Officer of 26 Degrees Global Markets has stated directly in e-Forex that “liquidity analytics will continue to transition from a passive, post-trade reporting layer to a real-time control system.” This is precisely right. But the transition is impeded by a structural problem: when TCA is both the regulatory standard for best execution and the commercial benchmark by which execution desks are evaluated, every participant in the ecosystem has an incentive to improve the same metric — which means the same collective blind spot persists.
3. The regulatory amplification problem
MiFID II best execution obligations were designed for a world in which post-trade TCA was the fastest available feedback mechanism. The framework is not wrong — it reflects a genuine commitment to transparency and accountability. But it was calibrated to a market cycle measured in hours and days, not microseconds.
The result is a regulatory architecture that, perhaps inadvertently, reinforces retrospective measurement as the standard of excellence. Firms investing in post-trade analytics to satisfy regulatory requirements are, structurally, investing in the wrong timeline. The structural fragility events that generate the most significant capital damage — flash liquidity withdrawal, adverse selection cascades, the phantom depth that characterises hollow-book conditions — do not wait for the end-of-day TCA report.
The FCA’s own thematic review of algorithmic trading noted that firms had given insufficient consideration to “how they would detect and respond to algorithms behaving in unintended ways” in real time. That observation applies with equal force to how institutions monitor the market conditions into which those algorithms are deploying capital.
A note on what this argument is not
This is not a case against TCA. Post-trade analysis remains the correct instrument for post-trade questions: regulatory best execution reporting, counterparty performance review, venue selection, and historical slippage benchmarking all require retrospective measurement, and TCA performs those functions well. The argument is narrower: that the industry has extended TCA beyond its original design scope — from audit instrument to execution quality standard — and that the gap between those two things is now commercially significant. A physician who uses a thermometer to diagnose a fever is using the right tool. A physician who uses a thermometer to assess cardiovascular risk is using the wrong one. The thermometer has not failed. The diagnostic framework has.
Conclusion: A different question
The industry does not need better TCA. The data is already excellent, the benchmarks increasingly granular, the analytics impressively sophisticated. What the industry needs is a different question.
The transition from “how did we execute?” to “should we have executed at all, in those conditions, at that moment?” represents the genuine next frontier for institutional execution intelligence. It is not a critique of TCA — it is an acknowledgement that TCA was built to answer one question, and the market has evolved to demand another. As the FX industry moves toward the real-time control infrastructure that its own practitioners are now describing as inevitable, the most important investment decision is not which TCA vendor to use. It is whether execution quality measurement begins at the point of the fill — or before it.
Charles Glah, CFA UK (ASIP), is a quantitative researcher and founder with over 20 years of professional experience in financial markets, specialising in market microstructure, machine learning, pre-trade intelligence, and algorithmic execution. He writes in a personal capacity. His website is: gateway.deepalgo.co.uk