I am going to admit this is quite against my beliefs, but very intriguing. I am starting to suspect this strategy, based mostly on gut-feeling, works well mostly for products that can be 100% dog-fed — where you develop also for yourself and you are the customer. There is less pressure to get to know the customer better (you are the customer, after all).
At the same time I believe measuring even non-financial impact of new features makes sense. Without that you lose that feedback loop mentioned here earlier. How do you know the feature you implemented actually solves the problem for majority of people? Is it possible that it is just you that finds that solution good and it creates issues to others?