How to Read a Broker's Execution Statistics

Why the headline number is nearly useless

"Average execution 34ms" is a real measurement of the wrong thing. Averages hide the tail, and the tail is where money is lost: the fills that happen during a data release, at the session open, or when liquidity thins. A broker whose median is fast and whose 99th percentile is two seconds looks identical, in a headline, to one that is consistently fast.

The second problem is scope. Speed is measured from the broker's server, not from your platform, so it excludes your own latency and any queuing before the order arrives. It is a useful internal metric and a weak comparison between firms.

What is worth looking for

Four figures say more than speed. The percentage of orders filled at the requested price, versus better, versus worse — a broker reporting symmetrical improvement and slippage is behaving differently from one where deviation is always in its favour. The rejection or requote rate, which reveals how often a quoted price is not honoured. Percentile latency rather than the mean. And the same statistics broken out by instrument, because majors and exotics behave nothing alike.

Brokers under tier-one supervision often publish some of this because best-execution rules require it. Where a firm publishes nothing at all, the absence is itself informative — not proof of poor execution, but a signal about what is measured internally.

Measuring your own fills

You do not need the broker's cooperation to build a small sample. Export your trade history and compare, for each order, the price you requested with the price you received, and record the time of day and the instrument.

Three patterns are worth naming. Slippage that is consistently negative — worse for you, on both entries and exits — is a costing problem regardless of the cause. Slippage that is negative on entries and neutral on exits usually points to how orders are routed rather than to malice. And rejections clustered around news events tell you which strategies this broker cannot support, which may matter more than the headline spread.

Reading the spread alongside it

Execution quality and spread trade against each other, and a comparison of either alone is misleading. A tight advertised spread that widens sharply at the exact moments you trade is more expensive than a slightly wider spread that holds. Sample the spread yourself at the hours you actually trade rather than at the quiet hours a marketing screenshot was taken in.

What to do with the conclusion

If your own sample shows consistent negative deviation and the broker publishes no statistics, the sensible response is a smaller allocation rather than an argument — execution disputes are hard to win and easy to avoid. If the numbers look symmetrical and rejections are rare, you have something most traders never obtain: evidence about your own broker rather than an opinion from a review site.