Define what happened before measuring how fast
A trading signal, a submitted request, a broker acknowledgement, an exchange fill and a reconciled position are different events. Reporting the elapsed time between two of them without naming those endpoints creates an ambiguous benchmark.
A broker can acknowledge an order that is later rejected, cancelled or only partly filled. A platform can receive a transport timeout even when the broker has already accepted the request. The operational question is the state of the order, not simply whether one HTTP request returned successfully.
An unsafe retry can increase exposure
Consider a synthetic incident sequence. At time A, a platform submits an order. At time B, the broker accepts it. At time C, the response is lost. At time D, the platform sees a timeout.
Submitting a new order immediately may duplicate the intended exposure. A design review should ask how the platform identifies the original request, retrieves authoritative broker state and resolves uncertainty before another action. This sequence is illustrative; it is not a report of a Wealthoak incident or a certification of an implemented recovery mechanism.
Scope a latency report
A credible benchmark identifies the broker, account type, instrument, order type, deployment region, observation period and sample size. Clock synchronization and timestamp ownership matter when events are recorded by different systems.
Publish a distribution, including tail observations, rather than the fastest request. State whether timeouts and failures are included in the denominator. Keep rejection rate, fill completion and price slippage visible alongside latency.
Availability must be account-specific
Account permissions, API sessions, exchange trading hours and the broker’s supported operations constrain execution. A broker name on a website is not enough to establish that a given account can place a given order.
Read the execution measurement framework and integration scope. These documents define evaluation questions. They do not substitute for an observed, dated acceptance test against the exact integration and account.