This problem often comes up and always ends with "it's an art" or some outright wrong or ymmv conclusions with a heady mix of variations on ford's faster horses

This problem often comes up and always ends with "it's an art" or some outright wrong or ymmv conclusions with a heady mix of variations on ford's faster horses metaphor. One effective way to "get" the product is to be a real user. Not just pretend as one but to actually use the product. This gives you a concrete foundation in one usecase. You can then branch out to other usecases by asking and observing others. It's not as easy or sexy as collecting user data and plotting fancy graphs. And it's very very hard to justify time spent on this to your manager if they don't understand this inherently. But it's very hard to fool yourself as a user of your own product.