Investors and institutions trading foreign exchange learned in June 2026 that the Foreign Exchange Professionals Association (FXPA) had formally defined FX spread grids as indicative pricing tools, not firm quotes or contractual benchmarks. A new analysis from FM Intelligence now asks the sharper follow-up: how far can the price a client actually receives drift from that published grid, and who absorbs the difference?
What the FXPA FX Spread Grid Guidance Actually Says
The FXPA’s June 2026 document addresses a persistent source of dispute. Liquidity providers, trading venues, buy-side institutions, and conduct regulators had each been reading spread grids differently. The guidance clarifies that a grid is a symmetric pre-trade reference, not an executable commitment. FinanceFeeds reported that the FXPA specifically warns treating grids as executable pricing benchmarks can lead to mismatched expectations between liquidity providers and clients.
FM Intelligence maps five distinct readings of the same document. Liquidity providers, venues, the data-rich buy side, and conduct regulators each interpret it as supporting their own position. The simplest reading is the one FXPA states plainly: definitions were disputed, an industry body clarified them. Because FXPA’s membership spans buy-side, sell-side, venue, and data firms, a document that serves several constituencies at once reflects genuine consensus rather than capture by any single interest. None of the five readings requires bad faith.
The FXPA has since issued a related document, Definitions and Best Practices for FX Internalisation in Algo Execution, dated July 2025, indicating the body is systematically working through structural ambiguities in how FX orders are routed and priced.
The Gap Between Grid and Fill: What FM Intelligence’s Model Shows
FM Intelligence’s proprietary Grid-to-Fill Gap model tracks the distance between the advertised spread grid and the spread a client actually realises on execution. Under normal liquidity conditions, the firm puts that gap at roughly 10%. Under stress, large order sizes, or thin liquidity, the base case widens to roughly 200%, with a high-case scenario above 300%.
The firm describes these as illustrative modelled estimates, built from FXPA’s own statements rather than a live transaction dataset, and revisable as actual data arrives. Divergence between a grid and a fill is, by construction, expected. Execution is asymmetric; a published grid is not.
Part of the explanation lies in how FX dealers operate. Dealers internalise about 80% of spot orders, matching client flow in-house rather than sending it to an external venue, according to the Federal Reserve Bank of New York. The NY Fed’s Liberty Street Economics blog also notes that FX remains the largest financial market in the world by trading volume. Internalisation lets dealers apply last-look rights and pricing skew, both of which affect the fill a client receives relative to a pre-trade grid.
Last look, as defined under Principle 17 of the FX Global Code, is ‘a practice utilised in Electronic Trading Activities whereby a Market Participant receiving a trade request has a final opportunity to accept or reject the request against its quoted price.’ The Code, maintained by the Global Foreign Exchange Committee (GFXC), does not impose legal or regulatory obligations on participants; it supplements local rules rather than replacing them.
The GFXC’s December 2024 three-year review identified enhancing transparency around certain execution activities and improving transparency around the use of client-generated FX data as explicit priorities. An earlier review in 2021 had already produced separate guidance on last look and pre-hedging, along with templates for transaction cost analysis (TCA), an independent measure of what execution actually costs versus a benchmark.
FM Intelligence is careful to separate the grid-to-fill gap from misconduct. The GFXC’s 2017 last-look response paper drew the same distinction between disclosed and undisclosed practices. Where FM Intelligence sees latent risk is narrower: the distance between a pricing representation and the fill is the same ground that older enforcement cases turned on. State Street settled for $382.4 million in 2016 over hidden FX markups paired with best-execution assurances. Barclays paid $150 million in 2015 over its last-look engine. Both cases involved undisclosed conduct, not a published grid. FM Intelligence stresses that no enforcement action between 2024 and 2026 has targeted dealer pricing against an advertised grid, and the analysis names no current firm.
The broader shift running underneath all of this is who gets to measure execution quality. Every move to benchmark on realised fill data rather than a grid directs attention toward the venues and analytics firms that capture and analyse those fills. FM Intelligence estimates only about 25% of institutional FX participants currently run independent, multi-LP transaction cost analysis. Its base case projects that share of execution cost judged primarily on realised data rising toward 62% by 2027, though it presents that projection with bull and bear ranges rather than as a certainty.
The timing matters. As the GFXC continues pressing for greater execution transparency and as TCA adoption grows, the gap between what a grid shows and what a client receives is likely to become harder to leave unexplained.
