Mitigating clearing latency and routing leakage on non-bank FX liquidity pools
We are currently restructuring our liquidity stack for a mid-market multi-asset brokerage, and we keep hitting a wall with our primary B-book to A-book bridge transition. Specifically, we've noticed a non-trivial amount of latency slippage when routing spot $EURUSD and $GBPUSD flow during late NY/early Asia sessions.
We have been testing two separate Prime of Prime providers. Both claim direct cross-connects in LD4/NY4, but under load, our fill ratios on limit orders are dropping by nearly 8%. Our logs suggest that while the initial execution message is sub-millisecond, the downstream clearing validation is lagging, which seems to trigger some aggressive quote-rejection or wider spreads from the LP side. We're essentially getting pre-hedged against, which defeats the purpose of our current STP logic.
Has anyone here managed to negotiate strict non-disclosure or no-pre-hedging SLAs with their LPs, or is that a pipe dream for a broker under $500M ADV? Also, if anyone has switched from a standard MT5 gateway to a custom FIX engine to bypass the middle layer for liquidity aggregation, did the reduction in rejection rates justify the dev overhead?
Ah, the classic "direct liquidity" fairy tale told by PoPs. During that late NY/Asia transition, most of those "direct" feeds are just recycled internal matching engines that dry up the second a light breeze hits the spreads. Have you checked if they're pre-hedging your flow before the bridge actually clears, or are you just getting front-run by their own internal LP?