I'm conflicted about the 2nd edition "Completes". For every good idea, there was a bad one, and I think there was something damaging about the mentality of endless rules-based customization (as opposed to imagination-based customization) they promoted. Instead of just imagining that your thief was a burglar, you now had a mechanical justification for it, which actually functioned to restrict options by sending the message that in order to do something interesting with a character class there needed to be a TSR-sanctioned way to do so. Rather than extremely broad character classes functioning as a canvas on which a player could paint almost whatever he wanted, suddenly people were rushing to the breast of Mummy TSR to suckle on her dangling teats for their imaginative milk.
However, one thing that the earlier Completes had in common was promotion of random generation as a means of sub-creation. This was an unqualified good. And it gave me an abiding love for the random generator which I don't think will ever go away.
A good set of random generator tables is a game unto itself. It makes DM prep genuinely enjoyable, because it allows you to roll lots of dice and surprise yourself with the results. This is intrinsically fun, but it does something extra - it, in turn, sets off creative sparks in your own mind as you attempt to weld the results together into something that fits. This gives the campaign world detail and throws out, like confetti, hooks for the PCs to get caught on.
Let's illustrate this through reference to the
The Complete Thief's Handbook's Random Thieves' Guild Generator. I will now use this generator to create a thieves' guild for the town of Swiftly, which I have just thought up (and stolen the name of from the title of a book I recently read). Swiftly is a river port, and it's a decent sized town - we'll say a population of 8,000. Here goes:
First, we find out how wealthy the town is. It's on a trade route, it has a port, and it's a major town, so it gets some bonuses for this. The final result of 19 - it's very wealthy. Naturally Swiftly is plush with cash from all the goods and money changing hands.
Next, we determine the attitude of the law. I have to know the dominant social alignment, but we'll call that neutral. I roll a 20, which means the law is corrupt. Obviously, the wealth in the town makes it easy to buy off judges. Or maybe the judges are just merchants - they're one and the same thing, and merchant courts run the legal system itself.
After that, we find out the relationships of the thieves to the merchants, and to other guilds. Again, we get some modifiers here because the law is corrupt, the society is rich, and so on. The result is a "standoff". The merchants are not in cahoots with the thieves' guild, but nor are they hostile. Their relationships with the assassins and bards' guilds are indifferent, but there is hostility with the beggars' guild. (I love, by the way, how it is taken for granted that these other guilds exist.) Beggars and thieves don't get along in Swiftly.
We now find out how many thieves there are in the town. The population is 8,000, and it is wealthy, and the law is corrupt, so the number is 2d10+3+10%. So that makes 15, rounded down. Some of these may not actually be in the guild itself - we'll find out shortly.
Now, the characteristics of the guild. I roll a 4, which indicates the leader is a guildmaster. A little disappointing, because if you get a 20 you get to roll on a special sub-table including results like "the guildmaster is a dragon". But you don't want to overdo things. We then turn to the nature of the rule, which is based on three axes: strong/weak, cruel/just, and despotic/populist. The leadership rolls reveal it to be weak, fairly cruel, and fairly despotic. A further roll, on another table, indicates that the membership is cohesive.
From this a picture emerges of a rather pathetic guildmaster who resorts to cruelty and despotism in an attempt to hold sway. The thieves in the guild remain cohesive, perhaps because life is good in Swiftly (it is so rich, and the law is so corrupt), and perhaps because none of them are particularly willing to challenge the guildmaster. They'd prefer to avoid being head of the organisation and thereby becoming targets themselves. They like the easy status quo - the guildmaster is not powerful enough to be anything more than a nuisance - and they are willing to let him be the fall guy if necessary.
The number of thieves in the guild is 75% of the population of thieves as a base rate, with -10% because leadership is weak, making it 65%. That's 9.75, rounded up to 10. The members are neutral to the 5 thieves who are not in the guild. The picture of a relatively lazy, laid-back guild solidifies. The thieves of Swiftly have it so easy, it isn't worth fighting - except, apparently, with beggars.
We now find out the levels of the thieves in the town. There are 2 of level 4, 1 of level 3, and 5 of level 2. The other 7 are 1st level. It makes sense for one of the 4th level thieves to be the guildmaster, with a the 3rd level guy/girl as his lieutenant. The remaining 4th level thief, and 4 of the 2nd levellers, would be independent - being of a certain level of competence, they don't need the guild and prefer to keep all their loot to themselves rather than pay guild fees. The 1st levellers are all good members - probably mostly youngsters.
And there you have it. The next steps would be to flesh out the names, personality and equipment of the higher-level thieves, think up a reason for the war with the beggars' guild, perhaps jot down some other NPCs - notably corrupt merchant/judges, high-up beggars, and so on - and there you have it: a finished thieves' guild. About 10 minutes' work, and something for the PCs to interact with should they ever end up in Swiftly.
Incontrovertible evidence that random generators are your friend.