More proof that speed as perceived by the end user matters. This time from a Google Research paper (PDF). Google's experiments demonstrate that increasing web search latency 100 to 400 ms reduces the daily number of searches per user by 0.2% to 0.6%. Furthermore -- and this is where it gets really interesting -- users do fewer searches the longer they are exposed. In other words, the cost of slower performance increases over time and persists.
To use Peter Van Dijck’s words:
In other words, if your website is a little slower, users will use it less (we knew that), but they'll also use it less and less over time, and when it speeds up again, they’ll still use it less than before the slowdown.
Based on their observations, Google suggests site builders to think twice about adding a feature that hurts performance if the benefit of the feature is unproven.