Mostly, while testing performance, we tend to measure average transaction time and, by only doing that, we miss the outliers. At some point, when GC causes pauses for a long duration, the application's response time increases drastically, which affects users accessing the application. This can go unnoticed, as we are only looking at the average transaction time. When the GC pause frequency increases, response time becomes a serious problem that we might have ignored by just measuring the average response time.