Sign in

Blog · Board and management reporting · Trading

Client KPIs for sales and trading desks: ten measures that matter, each with its formula and the export it comes from

The ten client KPIs a sales and trading desk should run on, each with its formula, the export it comes from and what it tells you: hit ratio by client and product, size-weighted hit ratio, inquiry share, the quiet client list, client tiering by flow and profitability, axe hit rate, traded unsent, coverage recency by client value, revenue concentration, and response time on inquiries. Also the three measures most desks miss, the figures to drop, the identities, and who owns what.

The short answerA sales and trading desk should run on ten client measures: hit ratio by client and product; hit ratio weighted by size; share of each client's inquiries seen; the quiet client list, meaning clients whose flow has stopped while the desk's has not; client tiers by flow and profitability; axe hit rate; clients who traded the axed instrument without being sent the axe; coverage recency by client value; revenue concentration; and response time on inquiries. They come from the inquiry log, the blotter, the axe and distribution records, the CRM and venue data. The three most often missed are the size-weighted hit ratio, because a client can trade every small ticket with the desk and every large one elsewhere; the quiet client list, which needs a client's own pattern as the benchmark; and traded unsent, which shows the distribution lists are out of date.

A desk sees a share of each client's inquiries and wins a share of those. The measures that matter show which clients trade size with the desk, whose flow has gone quiet, and whether the desk's axes reach the clients who would trade them.

The ten measures

# Measure Formula Export What it tells you
1 Hit ratio by client and product Inquiries traded ÷ inquiries received, per client, product and channel Inquiry log; blotter Where the desk wins and loses
2 Size-weighted hit ratio Notional traded ÷ notional inquired Inquiry log; blotter Clients who trade small with the desk and size elsewhere
3 Inquiry share Client inquiries seen ÷ estimated client inquiries in the product Inquiry log; venue or client-stated data Whether the desk is shown the flow at all
4 Quiet client list Clients whose inquiries in the last 20 days are below a stated fraction of their own norm, where desk volume is normal Inquiry log Flow that has moved to another dealer
5 Client tiers by flow and profitability Tier from trailing volume and revenue after cost of capital and hedging Blotter; revenue attribution Where balance sheet and attention should go
6 Axe hit rate Axes traded by the client ÷ axes received, within the window Axe log; distribution records; blotter Whether the distribution lists work
7 Traded unsent Clients trading the axed instrument, right side, in the window, without having been sent it Axe log; blotter Clients missing from the lists
8 Coverage recency by value Revenue of clients with a two-way contact within cadence ÷ revenue CRM or chat logs; revenue Top clients nobody has spoken to
9 Revenue concentration Top ten clients' share; largest client share, by product Revenue attribution Dependence on a few accounts
10 Response time on inquiries Median seconds or minutes to quote, by product and channel; hit ratio by response band Inquiry log with timestamps Trades lost to speed, not price

Every one of these is computed per account, per salesperson and product, and in total, and every one carries an identity that must hold before the table is shown.

The three most desks miss

Size-weighted hit ratio. Count-weighted is the default on every venue report.

The quiet client list. Revenue reports show who traded, not who stopped.

Traded unsent. Nobody joins the axe log to the blotter for clients who were not on the list.

A worked line

A client sent forty inquiries in the quarter and traded eighteen: 45 percent. Notional inquired was $400 million and traded $32 million: 8 percent. Every trade over $10 million went elsewhere. Another client sent twenty inquiries, traded six, and did $70 million of $200 million: 35 percent by size. The first client ranks higher on the venue report. The second is the better relationship by a wide margin.

What to drop

Inquiry count as a measure of relationship. Price-checkers inquire most.

Axes sent. Volume of distribution is not effectiveness.

Voice and electronic hit ratios blended. Two different competitions averaged into one number.

The identities

Table Must hold
Hit ratio Inquiries = traded + traded away + passed + no quote
Axes Matched trades are a subset of the blotter; each trade matches at most one axe
Tiers Every client is in one tier; client revenue sums to the desk total
Coverage Clients = covered + uncovered

A table whose identity fails is a table with a row missing or counted twice. It is not shown until it is fixed.

Who owns what

Measure Owner Reviewed
Quiet client list; coverage recency Head of sales; salespeople Weekly
Hit ratio, both weightings; response time Head of sales with head of trading Weekly to monthly
Axe hit rate; traded unsent Head of sales Monthly
Tiers; inquiry share; concentration Head of sales and trading Quarterly

A measure with no owner is a metric, not a KPI; see KPI versus metric versus measure.

Go deeper

The short version

Ten measures from the inquiry log, the blotter and the axe records. Weight by size, compare each client with itself, and join the axes to the trades. Covirage computes all of them from the exports desks already produce, files only, with the definitions stated and the identities checked. See Covirage for sales and trading desks.

Questions people ask

Why weight hit ratio by size?

A client at 45 percent by count and 8 percent by notional is using the desk for small tickets and price discovery, and trading size with someone else. Count-weighted ratios rank that client near the top. The size-weighted figure, beside the count one, shows which relationships are real.

How is a quiet client identified?

By comparing the client's recent inquiry and trade count with its own trailing pattern, in products where the desk as a whole has been active. A client that usually inquires thirty times a month and has inquired four times, while desk volumes are normal, has moved its flow. The rule excludes market-wide lulls.

What does traded unsent mean?

A client took the other side of an instrument the desk had axed, in the window after the axe was sent, without being on the distribution list for it. That client would have traded the axe. It is the clearest evidence a list needs changing, and it is found by joining the axe log to the blotter.