"Execution quality" is used to mean at least four different things, and the confusion is expensive. Measuring the wrong one leads to routing decisions that make the experienced outcome worse.
The four measurements
Fill rate. The proportion of orders that fill at all. A high fill rate on its own tells you almost nothing — a provider can achieve it by quoting a wide spread that is always available.
Slippage. The difference between the price displayed when the order was sent and the price it filled at. Useful, but sensitive to how the reference price is chosen, and easily flattered by a slow feed.
Latency. Time from order sent to execution report received. Matters most for short-horizon strategies and matters much less than most people assume for everyone else.
Rejection rate. How often an order comes back unfilled. A provider with excellent slippage on the orders it accepts and a 12% rejection rate is not providing good execution.
Why they trade off
Tightening the price a provider is willing to show generally raises rejections. Reducing rejections generally widens the price. Routing to the fastest session can mean routing away from the deepest one.
The consequence is that a single execution-quality score is a weighting decision disguised as a measurement. Anyone who gives you one has chosen the weights for you.
How we measure
Per provider, per instrument, per session, we record all four measurements separately and show them separately. Then routing rules are written against the measurement that matters for the flow being routed, rather than against a composite.
A desk running short-horizon flow and a desk running position trades should not be routed by the same rule, and they should not be told they have the same execution quality.
What this does not tell you
None of this predicts what a provider will do next month, and none of it is a statement about outcomes. It is measurement of what already happened, which is the only thing measurement can be.