Quant Memo
Core

FX Liquidity Aggregation and Client Tiering

Not every client sees the same FX price at the same bank — pricing engines sort clients into tiers by how profitable and predictable their flow is, and the spread you're quoted is really a measure of how the bank sees you.

Prerequisites: FX Order Flow and Price Discovery, FX Quoting Conventions

Ask two different clients at the same bank for a EUR/USD quote at the same instant and, in electronic FX markets, they may well see two different spreads. This isn't an accident or a bug — the bank's pricing engine deliberately sorts clients into tiers based on how the bank expects that client's trading to behave, and quotes each tier a spread calibrated to that expectation.

FX pricing engines don't quote one spread — they quote different spreads to different client tiers, based on how predictable, well-informed, or price-sensitive each client's order flow tends to be. A client whose orders rarely move against the bank afterward gets tighter pricing; a client whose orders often precede an adverse price move gets wider pricing to compensate for that risk.

Why tiering exists at all

A market maker's real cost isn't just the bid-offer spread it collects — it's adverse selection: the risk that a client's order arrives right before the price moves against the position the bank just took on. A real-money asset manager rebalancing a portfolio on a predictable schedule is relatively benign flow; a fast trading firm whose orders are frequently followed by the market moving in the direction of their trade is expensive flow to internalize, because the bank is likely to lose money holding the resulting position even for a few seconds. Tiering is the pricing engine's way of charging each client roughly what their flow actually costs to service.

Tier 1: tightest spread Tier 2: moderate spread Tier 3: widest spread, via aggregators
Each layer down adds markup, both for expected adverse selection and for the extra hop through an aggregating platform.

Worked example

A bank's pricing engine quotes EUR/USD at a raw interbank mid rate with a 0.1 pip spread to its top institutional tier. A corporate treasury client one tier down is quoted a 0.4 pip spread — wider, reflecting smaller, less frequent, but still fairly benign flow. A retail broker accessing the same bank's liquidity through a third-party aggregation platform receives a 1.2 pip spread — wider still, because the aggregator adds its own markup on top of the bank's tiered price, and retail flow is priced with an extra buffer against adverse selection from any fast traders mixed into that retail flow.

  1. Notional traded: $10 million by each client type.
  2. Top-tier spread cost: 0.1\text{pip} \times \10{,}000{,}000 \approx $100$.
  3. Corporate-tier spread cost: 0.4\text{pip} \times \10{,}000{,}000 \approx $400$.
  4. Retail-via-aggregator spread cost: 1.2\text{pip} \times \10{,}000{,}000 \approx $1{,}200$ — twelve times the top-tier cost for an identical trade size, purely from tiering and the extra intermediary layer.

What this means in practice

This is why "the" FX spread quoted for a currency pair is really a range, not a single number, and why sophisticated clients negotiate directly for better tiers or aggregate their own flow across several liquidity providers to try to qualify for tighter pricing. It also means a client's own trading behavior feeds back into future pricing: flow that consistently looks like it's trading ahead of price moves gets re-tiered wider over time, even if the client never intended to signal anything.

A wide spread quoted to a particular client doesn't necessarily mean the market itself is illiquid at that moment — it can simply mean the pricing engine has classified that client's flow as costly to internalize. Checking a different platform, tier, or execution method can reveal much tighter genuine liquidity sitting one layer up.

Related concepts

Practice in interviews

Further reading

  • BIS, FX Global Code — Principles on Pricing and Execution
  • Cartea, Jaimungal & Penalva, Algorithmic and High-Frequency Trading (FX pricing chapter)
ShareTwitterLinkedIn