It's struck me as I populate my DF game's setting that there are really two broad types of enemies in a campaign. You have your Important Villains, and your Fungible Monsters. I mentioned this a while back, but decided I'd try to write out my ideas about this more fully to get them clear, and maybe help others populate their worlds more easily.
(You could say Fungible Villains, but I think Fungible Monsters sounds better.)
Let's look at them in reverse.
Fungible Monsters are just NPCs placed for opposition. How tough they are or aren't, or how important they become or don't become, isn't critical. The idea behind fungible monsters is, you can swap them in or out. It doesn't really matter if they turn out to be easier or harder to kill than you expected. You can always deploy more or less next time. Or the PCs can relax and fight them more casually, or drill down and fight more ruthlessly, the next time. If those monsters turn out to drop dead like flies from a simple spell or are utterly harmless against the PC's front rank fighters, well, it happens. If you’re playing a game with an acceptance of high lethality, then even a TPK isn't really a problem. The next bunch of PCs will know not to mess with those monsters.
The monsters in and of themselves aren't very important as individuals. This is decidedly not the case with Important Villains.
Important Villains are NPCs that, in and of themselves, are important as individuals or as groups. They cannot easily be swapped out for another NPC within the setting, and what they can and can't do is important. They are important and useful for a sandbox game, and critical for a story-driven game.* These are the opposition which are inherently interesting and touch. They aren't random, and the GM needs to know how tough they are from the word go.
If the PCs decide they want to overthrow the Mad Wizard of Wu Tang, he's an important villain (primarily player driven). If the GM sets the Mad Wizard of Wu Tang after the PCs, he's an important villain (primarily GM driven). If the whole game is "Guys, we're going to play The Quest to Overthrow the Mad Wizard of Wu Tang," he's an important villain (story driven - pre-game decision driven).
If that aforementioned Mad Wizard of Wu Tang goes down to a simple spell you didn't plan for, or his Killer Kung Fu skills don't actually work that well, you're going to ruin the verisimilitude of the story. It just strikes a bad note if you didn't forsee some elementary defense he'd need to have survived long enough to be the Mad Wizard of Wu Tang. The more stuff the Bad Guy is supposed to be able to do, the more thought you have to put into it. It's totally lame if the whole quest is to fight an important NPC and he's taken down in a single second because you didn't build in the right defenses. It's a letdown if the "greatest swordsman in all the lands" is less of a challenge to a PC swordsman than that wandering group of goblins was. While these things can happen in reality, fiction and gaming isn't meant to be real. Conan doesn't have an easier time with the Big Scary than with the Big Scary's Henchmen.
I personally find it nice to run a game where the vast majority of critters are Fungible Monsters. It greatly simplifies my game prep. But the idea that some monsters really do need to be important and need to be crafted to explain their in-game position and reputation is still one I need to keep in mind. It's amusing if the trolls die in a few seconds while the goblins put up a real war of a fight. But it's lame if the Dragon of Felltower is a one-shot kill . . .
I also think these divisions help you to not over-think some encounters. If the monsters really are Fungible Monsters, and could really be anything, just relax and stat them up and see what happens. Save the worry for guys who legitimately need to be Important Villains.
* Note that story-driven does not imply a railroad. It merely implies that there is some over-arching plot or story, whether setting-, GM-, or player-driven, that provides the backbone for the campaign. Railroading is a way of running a game, not a game style. You can railroad in a non-story based game ("We go to that dungeon over there." "Suddenly, a teleport trap sends you to the dungeon I wanted you to go to!") as easily in a story based game ("We kill the important NPC!" "He miraculously survives and uses his alter reality action to undo your mistakes!"). It's a question of freedom of choice and action, not of story.