Measuring SMS waiting time: several timestamps beat one stopwatch
“The code was slow” may describe network delay, waiting for a message, or a late interface refresh. One stopwatch mixes these intervals. When evaluating a SimAggregator integration, choose several observable points in your own application and define each one first.
Useful points include the user action, request transmission, result receipt by the application, and display on screen. These timestamps do not expose the internal delivery time of external participants. Do not call the whole interval “supplier latency” when only your client is being measured.
Use a suitable stable timer for durations within one process, and absolute time for connecting logs. When combining records from different machines, account for clock disagreement. A negative interval is more likely a measurement problem than exceptionally fast delivery.
Report observation counts, unfinished cases, and test conditions along with averages. Comparing client versions requires comparable scenarios. If a change only speeds up rendering, that is useful evidence about rendering; it does not show that the external message arrived sooner.