cost per request is a feature, start measuring it on day one

we shipped something users loved and then watched the bill. now cost per request is on the same dashboard as latency. you cannot optimize what you do not see.