I still haven’t run a Megadungeon, even though I have bought several: Stonehell, Rappan Athuk (twice), Castle Whiterock, Tomb of Abysthor, … I feel like I would like big dungeons. When I prep for my games, however, that’s not how it works.✎
If you look at my Swiss Referee Style Manual you’ll note that it is practically without advice for dungeon adventures. That’s because my dungeons are small. Usually they are one page dungeons. How come? When I prepare a game, I usually start from the answers I got at the end of the last session. I almost always end by asking: “So, what are we going to do next time?”✎
I write down the relevant non-player characters: people to meet, people to oppose, people with jobs to hand out, with quests that need resolving. I usually end up with one to four characters.✎
The bad guys are usually in a defensible position so I prepare spells, minions, rooms, a map… a one page dungeon, a ship deck plan or a village with a few important buildings (in which case I won’t prepare floor plans for the buildings). I usually end up with four or five buildings or ten to twenty rooms, a sketch of a map.✎
I think of complications. This is usually something that works in layers. Every two or three days I have a lame idea that I mentally add to the adventure. After two weeks, however, five lame ideas make a cool complication. I usually don’t think of a solution. The gargoyles want the player characters to leave, consider themselves to be Übermenschen… The elves are petriefied by a gorgon bull and the players have neither the saves nor the spells to survive a direct encounter. As it turns out they managed to get the gargoyles to accept their commands and the gargoyles brought all the petrified elves to the survace.✎
I think of rewards. I usually start by rolling on the appropriate treasure tables and embellish the magic items, if any. A armor +1 and shield +1 turn into the golden halfling armor of Priamus Bullfighter who disappeared 21 years ago from Elfenburg. The sword +1 turns into the blade of the herring knight, smells of fish allows the wearer to feel how far away the next air bubble is… I make a mental note to add the remaining equipment of said knight to the dungeon or future adventures… If my players decide to follow up I will place his city or temple on the map, and add the protector saint of all fish, and his paladins, and their special abilities, and there will be rescue missions, and favors to be granted…✎
Working iteratively is important for my process. I try to pull in non-player characters from very old sessions. Wespenherz, the new hireling, is an elf that they had rescued from bad guys in a previous campaign up in the north. This ring they just found was forged by Qwaar the Axiomatic and didn’t Muschelglanz write a book about the rings of Qwaar? Yes he did and as far as he know he had decided to investigate the Barrowmaze and never returned… It gives depth to the campaign, some players remember and start digging through the campaign wiki, older players explain newer players what happened back then, … I love it!✎
This sort of thing doesn’t come easy. As I said, every two or three days I have a lame idea, but after two weeks I’ve had enough lame ideas that together they make the game better. Much better.✎
The process also shows why it’s hard to integrate megadungeons. When I look at them, I want to skim them for interesting non-player characters my players would want to contact, for prison cells my players would want to rescue interesting non-player characters from, for interesting rewards my players might want to be looking for. And that’s so damn hard.✎
I still remember placing the Barrowmaze in my campaign and having an important non-player character flee towards it. I was using Nualia, an evil fighter with an evil sword from a Pathfinder adventure and decided that her dad was a priest of Nergal who lived in the Barrowmaze and was involved in the power struggle. Then I had an evil authority person from Kerealia fleeing towards the Barrowmaze because the players had ousted him. The dungeon itself was ok, not overwhelming. We rescued a dwarf from a pit and he’s still with us.✎
Once the players fell into a bottomless pit and dropped into the astral sea, they never had the urge to return, hower. Having Muschelglanz disappear in the Barrowmaze is my new attempt at letting the Barrowmaze play a role in my campaign. It hasn’t managed to make itself important. I was unable to find or emphasize anything in the megadungeon itself that would motivate my players to return again and again.✎
The reason I thought of using the Barrowmaze again was that one player decided to offer a reward for the return of Muschelglanz. I proposed to my players to play a random party of first level dudes trying to claim this reward by going into the Barrowmaze and finding Muschelglanz. They liked the idea so that’s what we’re going to do.✎
When I started preparing for the next session, I did what I usually do. I looked for cells to put him in. I looked for the headquarter of a faction that held him prisoner. I tried to find the important non-player characters and I tried to find a thing that Muschelglanz might have been looking for. I know that there is supposed to be a tablet somewhere. But everyhing else has been tricky. I really need to skim it again. Gaaah.✎
Now you know what I would appreciate in a megadungeon. Just in case you’re writing a megadungeon. ✎
Greg Christopher says that he’d like to see more GM Style Manuals. An intriguing thought. This would be something that is different from the mechanics, it would be an instruction manual on how to to run the games the way you do. I guess it will relate to your preferred campaign style or a simple Know Your DM flyer, but it would go beyond it: it would tell you and teach you how to achieve the desired outcome.
I wonder if I could write something like that. Others have, too.
Let me try. What would I tell a beginning referee if they ask for my advice?
I should find a catchy name for it! This is the… uh… the Swiss Referee Style Manual! I’m not Swiss, but I live here. Think of the Alps, the forests, the rivers, the valleys, the ruined castles, the army bunkers, the halberds and mercenary traditions, and… I dunno! It makes no sense! But it sure beats Running A Campaign According To Alex. (Does it?)
Anyway, I’d love to keep it short but it ended up being a massive wall of text. Sorry about that! At least it’s less than 300 pages.
Check it out → Swiss Referee Style Manual! :attention:
(I’ve continued working on this style manual and decided to move it to a separate page.)
In the comments, he also talkes about note taking. I wanted to say something about taking notes after the game; I had also heard something like that on Fear the Boot. I googled for the opposite of preparation and guessed that somebody must have used <b>postparation</b>. And indeed, somebody has:
Pre-paration is what we do before we start a job. Post-paration is what we do after we do the job. Completing all the postparation immediately after the job is what generates the feeling of well deserved relaxation. 
Taking shorthand notes during the game, and making them readable after the game are mandatory for my games. Otherwise I’d forget. Since everything new happens at the gaming table, I can just add these little notes to our Campaign Wiki after the game. If the players have met ogres once or twice, the ogre page on the wiki may be very short:
The hills between Dangerous Jungle and Great Roaring Jungle seems to be infested with these giants. Then again, their minds seem to be weak. They seem to have a fascination with skulls. 
It’s great. It works.
It’s not “prep” but it’s still a little bit of work. But since it comes <b>after the fact</b> it doesn’t railroad anybody.
I just read the pages three to six of Robin’s Laws of Good Game Mastering by Robin D. Laws. It talks about Player Types and lists them:
I think I qualify for Storyteller. When I prepare sessions that I’ve written myself, I usually plan to do plot development via roleplay at the beginning, leading up to a fight or two. I try to balance these two elements over two or three session arcs at the most. Thus, if a session was almost all about an intrigue, then then there will be a big fight and some form of resolution the next session, or the session after that.
Within a story arc of three or four sessions, there will be a number of hooks that offer players an in-game choice of how to continue the game. (I’d love to play much longer mini-campaigns and adventure paths in the future.) As an example: The players fought the orcs at Eralion’s castle in CrucibleOfFreya. The orc leader escaped and was not tracked. Random encounters alerted the players to the fact that there were gnolls in the area and they expressed interest in pursuing this. So instead of leading them to Tomb of Abysthor and more orcs and evil clerics, I’m leading them against the gnolls in TheDistraction. This allows players to determine the future of the campaign in-character.
I took Law’s Game Style quiz, and it confirmed my Storyteller type. This also effects my DM style: I don’t mind meta-gaming at all, if it makes for a better story! Conversely, I have to work on better tactics of my minions, and how to provide my method acting players with more opportunity for internal turmoil. ;)
Looking at Georgios’ GM Types (the original in German: Georgios Spielleiter-Typen), I guess I’m “Director” first and “Provider” second. I do like the occasional “Plotmeister” session. I used to be a “World Builder”.
Update: The only quantitative analysis ever, apparently: Breakdown of RPG Players.
I like categorizing players by asking them to rank combat, interaction, and riddles. Personally, I hate riddles. They break the flow, and to be honest, I don’t get to see many riddles in real life, so riddles protecting things are not very believable.
Some players do enjoy riddles, however. What are good riddles to use?