Certain aspects of human culture, certain aspects of engineering quality are hard to put into numbers, as it turns out.
When we talk about (for example) writing a book, a simple productivity metric would be pages-written-per-day. Not that this isn't a useless metric, but the actual metric you want would be more something like good-pages-written-per-day, because throwing some garbage text out quickly is easy, the hard bit is doing good writing.
But then the obvious question is, what makes a good page? How do you judge the quality of a process' output aside from simple to create metrics? And many managers don't have the understanding to make that judgement. Quite frankly, if you're a manager and you cannot reliably judge the quality of the product you're working on, everything becomes hit and miss like throwing pudding onto the wall and seeing what sticks. In that case relying on a simple metric is worse than e.g. trusting the judgement of experienced engineers or users of your product.