I tend to base this decision on the level of pain and disruption it's going to cause.
You see... these decisions are far easier for an individual who's actual hobby is the computer itself. It's new, lets give it a whirl!
But when you are talking about a productive enterprise that's running smoothly it's a little harder to make the decisions. There are hundreds of mini-decisions ... will current software work in the new OS, will people need mass hardware upgrades, how much is operator re-acclimation going to cost, how many service calls will it entail... and on and on.
Many corp. and SoHo setups can't afford to be down for any length of time (usually measured in minutes). The system disruption of changing over to new operating systems, learning new UIs, perhaps even new software can seriously hurt a major corporation and it can kill that little 5 man travel agency.
Any new OS that goes into that environment has to be a "drop in" fit (what we used to call "turn key") and it has to work from day 1 and keep right on working. Unlike a hobbiest where the fun is in finding and fixing, these outfits suffer horridly from even the simplest complications in their day... they aren't going to sit back and let me do a clean install or even tweak... they want their computer working NOW!
Companies like these are not only disinclined to update, many resist it right up to the moment a computer fails and the replacement comes with the new OS preinstalled. Heck I've got whole networks I care for that are still using Win2000... Their logic is sound... It's working... Leave it the he11 alone!
So, over the years I've developed this habit of weighing a new OS value on the mass outcry of human suffering I may have to endure...