RFQ vs Streaming by Notional: When Each Mode Wins, and Why

Streaming prices are continuously published and quotable for whatever is at the top of the book. RFQ prices are quoted on demand for a notional you specify. The two pricing modes coexist on every credible institutional FX venue, and the choice between them is one of the most consequential decisions an execution desk makes — often without explicitly making it. Most desks default to streaming because streaming feels active and RFQ feels manual; the resulting cost asymmetry is one of the larger silent drains on institutional execution.

What the streaming top-of-book is, and is not

The streaming price you see on a Bloomberg or a 360T or an EBS is the LP's quote for the volume at the top of the LP's book. For majors, this is typically a few million units of base currency. For crosses, it is less. For exotics, it can be hundreds of thousands of units.

The streaming top-of-book is a real, executable price for that quantity. It is also not a price for ten times that quantity, or fifty times. When you send a marketable order larger than the top-of-book, you walk the book — your child fills land at progressively worse prices, your effective cost rises faster than the top-of-book spread suggests, and your TCA reflects the average rather than the headline number.

What RFQ is, and is not

Request-for-quote is the institutional alternative: you tell the LP exactly how much you want to trade in which direction, and the LP returns a single firm price for the full notional, valid for a short time (typically 10-30 seconds). The LP knows the full size and can price accordingly, hedging or warehousing as it chooses.

RFQ is not slower than streaming for the purpose of large orders. The total time from decision to fill is comparable; the streaming alternative replaces 'one RFQ quote' with 'fifteen child fills against fifteen top-of-book quotes' and the overhead lives in routing and signalling rather than in the quote round-trip itself.

The mechanics of how the streaming book is structured for institutional flow, and what determines the relationship between the streaming top-of-book and the deeper book, are covered in The Architecture of a Fair Spread. The short version: the streaming top-of-book is a function of the LP's adverse-selection budget at that quantity; deeper sizes have a different budget.

The threshold

The threshold at which RFQ starts winning over streaming is, for a given pair on a given venue, well-defined and stable. It depends on three things: the depth of the streaming book at the relevant time of day, the half-life of your flow (because longer half-lives make signalling more expensive and streaming therefore more expensive), and your tolerance for execution variance.

As a rough rule for majors during the European session, the threshold sits around 10-20× the streaming top-of-book quantity. For EUR/USD this is typically $50m-$100m; below that, streaming wins; above that, RFQ wins. For G10 crosses the threshold is lower — $20m-$50m. For exotics it can be near the streaming top-of-book itself; the streaming book is shallow enough that walking it has a near-immediate cost.

When streaming wins anyway

Discovery-mode trading

When you are not yet committed to a full notional and want to dip into the market to find out how much resistance the price puts up, streaming is the right mode. You buy or sell a small clip, see the response, decide whether to continue. RFQ for this use case is wasted — you are asking the LP to price a quantity you do not actually intend to trade, and you will get worse subsequent prices for that probe.

Strategy-driven children

If you are running a VWAP or TWAP algo, the algo is already chopping the parent into clips that each fit inside the streaming top-of-book. RFQ is the wrong tool for the children; it is the right tool for the parent if the parent is large enough.

Multi-leg structures

Swaps, forwards, and option-equivalent multi-legs have their own quote mechanics that streaming doesn't address well. RFQ is the default here; the question is whether you negotiate the legs separately or as a single structured request.

When RFQ wins

Large notional

The obvious case. Anything above the threshold described earlier.

Long half-life

If your flow has a long information half-life — a discretionary macro position you intend to hold for hours — the signalling cost of slicing through the streaming book is large. RFQ collapses that cost; the LP gets the size in one print and your residual intent is not visible in the streaming book.

Stress conditions

In a stress session, streaming top-of-book is the first thing to thin out. The same notional that streamed at 0.2 pips on a calm Tuesday streams at 1.5 pips during a release window — and the deeper book widens disproportionately. RFQ tends to widen less in stress conditions, because the LP is pricing a single discrete decision rather than an aggressive child fill in flight; the LP can absorb the position into a longer-horizon hedge if necessary.

End-of-day reconciliation

Hedging a daily residual into the market via streaming is a recipe for predictable, exploitable behaviour — the same direction, the same time, every day. RFQ randomises the price more (in your favour) and removes the time-of-day pattern; the cost difference is small per ticket and accumulates into a meaningful annual number.

How Drovix surfaces the choice

Drovix's institutional UI offers RFQ and streaming side-by-side on every quote, with a notional-specific spread indication for each mode. The threshold above which the RFQ price is tighter than the streaming-walked price is computed on the fly and shown to the operator. For algo execution, the parent-order spec can specify the mode threshold explicitly; the algo splits between modes based on the live computation.

The point is not to make every desk an RFQ desk. The point is to make the choice visible, defaulted correctly, and overridable per parent order. Most desks, once they see the breakdown, send 10-25% of their notional through RFQ where previously they were sending 0%. The cost savings are not glamorous; they are real.

Where to go next

→ Capacity Planning for Execution Strategies — how to size a strategy against the RFQ/streaming threshold of the venue you intend to deploy it on.

→ Routing Beyond the Inside Quote — the routing model that makes the streaming-walks-the-book problem less expensive when streaming is the right answer.