I think the idea is to measure average latency but to give extra weight to the worst 5% and 10% of latency cases. I've not seen such a measure before but don't

I think the idea is to measure average latency but to give extra weight to the worst 5% and 10% of latency cases. I've not seen such a measure before but don't think it's absurd.