Management technique we learned early on: short-term machine learning deadlines can be set based on inputs (e.g. high-quality execution on a set of experiments) but not outputs (e.g. reaching some level of performance). Science does not bend easily to the wishes of managers.