When you estimate what you can pull off in the next iteration, are you basing that on a running average? Or the reality of the moment?