Sign in

Blog · Wallet share and penetration · Trading

Hit ratio by client and product: a trading desk's guide to flow share

How a sell-side desk computes hit ratio and flow share per client and product from its own RFQ logs and blotter, why the two numbers must be read together, and the three drifts that predict a client is leaving before the revenue shows it.

The short answerHit ratio is RFQs won divided by RFQs received, per client and per product, over a window. Flow share is your won volume divided by the client's total flow in that product, where you can estimate the total. Compute both from the RFQ log and the blotter, roll them up by sales-trader and desk, and watch the trend per client: a hit ratio that halves while volume holds is a client testing another dealer.

A trading desk knows its P&L by client. It rarely knows, per client and per product, what share of the client's requests it wins and what share of the client's flow it sees. Both are in data the desk already logs. This guide sets out how to compute them, roll them up, and read the drifts that matter.

The two measures

Hit ratio answers "when they ask, do we win":

Hit ratio = RFQs won ÷ RFQs received, per client, per product, per window

Flow share answers "how much of what they do comes to us":

Flow share = volume won ÷ client's total volume in the product

Hit ratio needs only your own RFQ log. Flow share needs a view of the client's total, which comes from venue reports, a benchmark provider, or an estimate. They must be read together: a high hit ratio on a small share means the client sends you the trades they have already decided to do with you, and a low hit ratio on a large request count means you are the price check.

The rows you need

  • RFQ log: one row per request, with client, product, tenor or sub-product, size, whether you quoted, whether you won, and the timestamp. From the venue exports and the internal quoting system.
  • Blotter: one row per executed trade, with client, product, size and price. The won side of the RFQ log should reconcile to it.
  • Coverage: which sales-trader owns which client, from the CRM or the desk's own list.
  • Client totals, where available, per product and period.

Client identifiers are enough; the measure never needs a name. On a desk where the data cannot leave the bank, all of this runs inside the bank's own tenant.

The roll-up

  1. Client by product. Requests received, quoted, won; volume won; total volume where known. Hit ratio and flow share.
  2. Client. Sum requests and volumes across products. Do not average the product hit ratios: a client that sends 400 swap requests and 12 option requests is a swaps client with an options hit ratio that is noise.
  3. Sales-trader. Sum across the sales-trader's clients. Assert that the sum of clients' won volume equals the sales-trader's blotter total.
  4. Desk. Assert that sales-traders sum to the desk, and that the desk's won volume equals the blotter.

The identity, as always:

won(desk) = Σ won(sales-traders) = Σ won(clients) = blotter total

A client covered by two sales-traders, or a product renamed after a venue migration, breaks it at exactly one level.

A worked example

Rates desk, one client, two products, last quarter against the quarter before.

Product Received Won Hit ratio Prior hit ratio Volume won Client total Flow share
Swaps 412 132 32% 34% 3.1bn 9.8bn 32%
Options 96 9 9% 21% 0.2bn 1.6bn 12%

Swaps are steady. Options have halved in hit ratio while the request count has held. The client is still asking, and increasingly trading elsewhere. That is the pattern that precedes a lost relationship by a quarter or two, and it is invisible in the revenue line until the requests stop too.

The three drifts to watch

Hit ratio falls, requests hold. The client is testing another dealer's prices against yours. Usually pricing, occasionally response time. Visible a quarter before revenue moves.

Requests fall, hit ratio holds. The client has moved a product to another dealer and stopped asking you. Your win rate on what remains looks fine. The request count is the signal.

Hit ratio rises, volume falls. You are winning the small tickets and losing the large ones. Common when a competitor has committed balance sheet to the client. Size-weighted hit ratio shows it; count-weighted hides it.

None of these need new data. They need the same log, computed the same way every quarter, with the prior quarter beside it.

Where it goes wrong

Quoted versus received. A desk that declines to quote on a request should count it as received and not quoted, not drop it. Otherwise the hit ratio measures only the requests the desk chose to compete on, which flatters it.

Venue migration renames products. A product code that changes when a venue upgrades splits one product into two rows and halves both hit ratios. Keep a mapping table and apply it before the roll-up.

Size in different units. Notional in one venue's export, contracts in another. Convert before summing, and assert the blotter total afterwards.

Joint coverage. Two sales-traders on one client double the client in the roll-up. The desk-level assertion catches it: sales-traders sum to more than the blotter.

Every quarter, without the spreadsheet

Mapped once, the same exports produce the same roll-up each quarter, with the prior quarter alongside and the three drifts flagged per client. The desk head gets the ranked list; each sales-trader gets their own book. Covirage does this inside the bank's own tenant on Enterprise, with client IDs only. The trading page describes the setup.

Questions people ask

What is a normal hit ratio for a trading desk?

It depends on the product and how many dealers are in competition. A five-dealer RFQ where every dealer is equal gives each a twenty percent baseline. Desks that are consistently above their baseline on a client are the incumbent; desks below it are being used for price discovery. The baseline is the comparison, not an industry number.

How is flow share different from hit ratio?

Hit ratio counts requests; flow share counts volume. A client can send you every RFQ and trade elsewhere, giving a low hit ratio, or send you only the trades they intend to do with you, giving a high hit ratio on a small share of their real flow. You need both to know which one you are.

Where does the client's total flow come from?

From venue data where the venue reports it, from benchmark providers for the products they cover, and otherwise from the client's own disclosures or an estimate. Record the source per client. Where there is no reliable total, report hit ratio alone and say so.