Also in this issue: why human in the loop is the best ML model metric, how consumption product pricing gets overcomplicated and how to properly dogfood your product.