A vendor's millisecond figure may measure parsing, request submission or a private internal step. The trader experiences something longer: a source message appears, the copier receives and interprets it, a route prepares an order, and the destination accepts or executes it. Comparing different endpoints produces an impressive-looking table with little practical meaning.
This article offers a measurement method, not a tested speed leaderboard. We have not run a controlled benchmark across the products in the main comparison, and public marketing numbers do not supply one. Treat speed claims as questions about measurement until their start and end events are clear.
In this guide
Name each clock boundary
Separate publication, receipt, interpretation, submission, acknowledgement and fill. Some events may lack precise accessible timestamps, so record that limitation rather than inventing precision. A second-resolution source timestamp cannot reliably support a comparison claiming differences of a few milliseconds.
Clocks must also agree well enough for the measurement. Subtracting two unsynchronised machine times can produce a negative interval or blame the wrong component. Use consistent event identifiers and timestamp sources, and keep the raw records alongside calculated durations.
| Interval | What it can reveal | What it does not prove |
|---|---|---|
| Publication to receipt | Source delivery and reception delay | Parser speed alone |
| Receipt to interpreted instruction | Parsing and processing work | Broker execution time |
| Submission to acknowledgement | Destination request handling | Final fill price |
| Acceptance to fill | Order and market behaviour | A software fault by itself |
Pending orders complicate the stopwatch
A limit order can be accepted promptly and remain unfilled because its price condition has not been met. Counting that wait as copier latency misdiagnoses normal order behaviour. Likewise, changing a range-based instruction into a market order may shorten the wait while materially changing the trade.
Compare the same entry policy when testing products. The entry-range article explains the consequences of those choices. A faster result is not automatically a more faithful result.
Test representative messages and conditions
Use the same destination environment and comparable messages, including the formats your source actually posts. Record order type, symbol, size, connection arrangement and market session. AI extraction and rule-based parsing can have different processing characteristics, but neither label supplies a universal timing result.
Collect enough observations to see variation rather than quoting the fastest successful example. Report a typical result, slower observations and failures separately, with the sample conditions. Do not discard a timeout simply because it would spoil the average.
Locate the delay before buying infrastructure
A local MT5 terminal sleeping through a signal is an availability problem, not a marginal parsing optimisation. A broker rejecting an invalid quantity is a validation problem, even when the dashboard eventually reports it slowly. The troubleshooting guide separates missing, skipped and rejected instructions.
A VPS can change the operating environment of a terminal route, but it cannot guarantee a particular fill or fix an ambiguous source message. Read the cloud-versus-VPS guide before assuming that moving one component relocates the whole system. Establish the slow boundary first.
Price quality deserves its own record
Compare the actual accepted and filled orders with the intended instruction, including spread and available quotes. Slippage and processing delay are related in some circumstances but are not interchangeable measurements. A quick request can still receive a different price in a moving market.
Preserve the evidence and repeat after meaningful configuration changes. The useful result is a route whose timing and failure modes you understand, not a universal claim that one vendor wins every race. Speed matters most when it accompanies the correct instrument, quantity and trade lifecycle.
Put the route to a demo test
Use the current setup instructions and verify the resulting orders in your destination account. Start with a configuration you can inspect before adding more sources.
Open TelegramSignal in AlgoWay