Skip to content
CyclepediaThe trading encyclopedia
ITENESPT-BRFRDE
Learning path Gold Professional operator

Execution quality: price, fills and speed

Execution quality compares an order's actual outcome with declared benchmarks: price, effective spread, price improvement, speed, probability and executed quantity. It is not captured by the fill alone.

On this page

In plain terms — An order that was “executed” was not necessarily executed well. Evaluating it requires stating which price and time it is compared with, how much quantity was filled, how quickly and at what total cost.

Execution quality describes how favourable the order's actual outcome was relative to a benchmark and its stated objectives. No single measure is suitable for every order: an urgent exit prioritises speed and fill probability, while a large, non-urgent order may accept more time to limit cost and market impact.

This analysis is distinct from best execution. The former is a set of measures; the latter, in regulatory frameworks that require it, is an obligation concerning the process and the result pursued across several factors. Good statistics for a sample do not guarantee the best outcome for every order.

Main dimensions

Dimension Question
Price is the fill favourable relative to the chosen quote or benchmark?
Cost have spread, fees, slippage and impact been separated?
Speed how much time passes between receipt, executability and execution?
Fill what proportion of the order was executed?
Probability are comparable orders filled, cancelled or left pending?
Stability does the result persist across instrument, time, size and conditions?

A venue may be fast but provide a worse price; a limit order may obtain a better price but remain partly unfilled. Quality comes from a trade-off consistent with the objective, not from maximising one column in isolation.

The benchmark must be stated

Comparing a fill with “the price I saw” is not reproducible. Depending on the analysis, possible benchmarks include:

  • the best bid and ask available when the order was received;
  • the quote midpoint;
  • the arrival price, observed when the decision or order reaches the system;
  • the closing or opening price of a defined interval;
  • VWAP or another period benchmark;
  • the signal price, if its timestamp and feed were retained.

Different benchmarks answer different questions. A full-day VWAP does not measure the same thing as the best ask when a market buy order was received. The feed, time zone and timestamp precision are part of the definition.

For a purchase, one elementary measure of price cost is:

unit cost   = average executed price − benchmark price
cost in bps = unit cost / benchmark price × 10,000

For a sale, the sign is reversed. With partial fills, the average price and statistics must be weighted by the quantity actually executed.

Effective spread and price improvement

In the U.S. equity market, Rule 605 uses standardised benchmarks and statistics to make different entities comparable. Its measures include quoted and effective spread, price improvement or disimprovement, speed and quantity executed.

The effective spread compares the execution price with the midpoint of the reference quote while accounting for the buy or sell side. Price improvement means an execution better than the reference quote under the applicable definition. It is not the trade's profit and, by itself, does not measure total cost.

Rule 605 definitions belong to the U.S. scope covered by the rule. They can provide a methodological model, but should not be applied without adaptation to futures, foreign exchange, crypto assets or differently structured markets.

Minimum data for an audit

A useful record retains at least:

  1. instrument, side, quantity and order type;
  2. decision, receipt, executability, acknowledgement and fill timestamps, where available;
  3. limit, stop and duration instructions;
  4. venue, broker or routing path;
  5. synchronised quotes and benchmarks;
  6. every fill with quantity, price, fee and currency;
  7. quantity cancelled, rejected or left unexecuted;
  8. relevant market conditions, without using them as an automatic explanation.

Results should be segmented. A single average can conceal deterioration at the open, around news, for large sizes or on a particular venue. Distributions and percentiles reveal tails and instability better than the mean alone.

Limits of the analysis

  • The price that would have been obtained through alternative routing is counterfactual and often unobservable.
  • An unexecuted order has an opportunity cost that is difficult to compare with an immediate fill.
  • Slippage, spread and impact can overlap in the measure if the benchmark is not selected carefully.
  • Small samples, unsynchronised timestamps and changes in size make comparisons fragile.
  • A correlation between latency and deterioration does not, by itself, prove causation.

Common mistake — Comparing brokers or venues using only nominal fees or one order. Evaluation requires comparable orders, explicit benchmarks and executed quantities.

Sources

Suggest an edit My proposals