linux on the desktop failed for the same basic reasons many startups fail. They didn't listen to their users or focus on making a product that their users wanted.
By users I mean the core market of Windows - the business desktop. How often did linux outreach staff come to a large business and listen to the concerns of the customers? Never (in my experience). How often did Microsoft do this? Often (again in my experience).
It's confusing when you address Linux this way. Most Linux distributions are not for-profit, so they're going to have any staff going to businesses. Plus, even if they did, they'd have to offer support, which they aren't able to.
The only distributions that can do that are the enterprise paid ones, and those are mostly meant for the server side. AFAIK there isn't really an enterprise desktop distribution of Linux really intended for that kind of use. Linux on the desktop is what it is -- it's an option for those who want it. Market share is not a big consideration at all.
Another element is that anything above the kernel is subject to change at whim of some primadonna developer.
MS has in the past bent over backward to maintain binary compatibility.
Some of the same attitude is present in the kernel mantra of not breaking userspace, but userspace devs seems all to happy to break stuff at the drop of a hat.
By users I mean the core market of Windows - the business desktop. How often did linux outreach staff come to a large business and listen to the concerns of the customers? Never (in my experience). How often did Microsoft do this? Often (again in my experience).