Every platform in this category advertises one. Under 20 milliseconds. Under 100. Under 200. Thirty.
We don't, and this page explains why — partly so you can hold us to it, and mostly so you have better questions to ask anyone who does.
The numbers aren't measuring the same thing
"Latency" in this category can mean any of:
- Time from receiving a signal to handing it to the broker's API
- Time from receiving a signal to the broker acknowledging it
- Round trip from signal to fill confirmation
- Internal processing time, excluding every network hop
These differ by orders of magnitude. The last one flatters the most, which is why it's the most common. A platform advertising a number without saying which of these it measured has told you nothing you can compare.
And they're not measured the same way
Median or 99th percentile? Over what window? Including or excluding the market open, when queues are deepest and it matters most? Across which brokers? A median on a quiet afternoon and a P99 at the open are different by a factor that dwarfs the difference between vendors.
None of the numbers in this category come with a methodology. Without one, a smaller number means a more flattering measurement, not a faster platform.
Why we won't join in
We measure this continuously and we care about it a great deal. We could publish a number today.
The problem is that an honest number loses. If we publish a well-defined figure with its methodology and its percentile, and a competitor publishes a smaller number measuring something else with no methodology at all, the trader comparing them sees only that ours is bigger. Publishing honestly would cost us against an unverifiable claim, and the only way to win is to publish something equally unverifiable.
So we've opted out. If this ever becomes a category where methodologies are published, we'll publish ours the same day.
What actually determines whether you get your fill
Milliseconds in the middle of the chain are rarely the reason a trade goes wrong. What usually is:
Whether the order was valid. A malformed alert never reaches a broker at any speed. This is, by a wide margin, the most common reason an automated order doesn't happen.
Whether the account could take it. Risk limits, existing positions and paused connections stop orders regardless of routing speed.
What the market did. Liquidity at your price at that moment is not a property of your routing platform.
Whether you found out. An order that failed silently costs more than one that failed slowly.
What to ask a vendor instead
- What exactly does your number measure, from which event to which event?
- Is it a median or a percentile, and over what window?
- Does it include the market open?
- What happens when my broker is slow — do I see an unconfirmed state, or does it look like a failure?
- When an order doesn't go, how quickly do I know, and does the platform tell me why?
The last two matter more than any number, and almost nobody asks them.