The Central Concept
The main concept of Leadership for my long-standing campaign was rather simple. The way I saw it, hired people are less loyal and more willing to jump ship than followers gained through Leadership. Followers tend to be people who see things in their leader that they like a great deal (for whatever reason), and to such an extent that they'll do a lot more for their leader than other people. So one key element to my concept was the idea of attraction. Romantic and sexual attraction can come into play between leaders and their followers, but a follower's attraction can spring out of many other motives. A follower can be drawn to a leader due to envy, admiration, love, friendship, and other things besides. Another key element is extent - followers will do more and probably ask for less in return. Followers are far more willing to relocate, perform labor, offer spells, share information, keep an eye on the player character's interests, guard the player character's stuff, and so on. Followers will not necessarily work for free but they are likely to charge less and to be happy with other rewards (more contact, praise, sexual attention, what have you).
In my game, however, NPCs are always their own people with their own desires, fears, and minds. This might seem like a stupidly simple thing to say, but it heavily influenced my portrayal of Leadership. For instance, players never got to design their own followers or cohorts when I ran. I initially made that decision based on my own thoughts about social interactions. I believed that people are a mystery to one another until they start to socialize, and even then you never know anyone else completely. In gaming, I feel that it is part of my job to make the NPCs feel as much like people as possible; relatedly, I feel that it's important to leave mysteries between the PCs and the NPCs. In the end, I figured: the players don't usually get to decide how the NPCs are made and they aren't generally privy to character statistics - why should that change because of Leadership?
Thus, I did not usually allow players to directly control cohorts and followers gained through Leadership, nor did they generally gain access to character sheets for their people. Players would learn what their people coud do as they got to know them. Generally speaking, I designed followers and cohorts to be good at what they did (within reason). Cohorts traveled with the group while followers stayed behind as friends, support, business partners, information gatherers, and so on. This meant that there wasn't a horde of NPCs traveling with the party and the player characters didn't lose much of the spotlight. I gave my players the choice of controlling their cohorts in combat; some took the chance, and some didn't. I continued to portray cohorts and choose their statistics to keep them distinct from the player characters. It was important to me that NPCs of all kinds remained their own people - they're not just extensions of the player characters, or cardboard cutouts.
I based such decisions on my concept for Leadership (as well as my overall DMing philosophy) and they struck at the heart of the most heated complaints against the feat. Some DMs disallow the feat outright because they have seen players build unreasonably powerful cohorts and/or followers; sometimes players get so out of hand that their followers outshine the player characters. Other times, DMs shy away from Leadership because they don't want players to be able to play another class and gain the associated benefits, for only the price of one feat. There are certainly other reasons for disallowing the feat, but many of the complaints I've heard might be remedied if DMs were willing to take a more active role in how the feat is adjudicated.
In my game, leaders could only try to recruit certain people, or types of people - in the end, it was up to the NPCs to decide to follow. Leadership was definitely based on Charisma and powers of attraction cannot necessarily be controlled, or reliably controlled. A leader could discourage people from joining them but they didn't get the final say on who showed up in the first place. Followers would also leave if they were treated poorly on a consistent basis, regardless of whether or not the player character's Leadership score had changed. The first characters to take Leadership in my long-standing campaign got to pick their people from NPCs who showed up in the course of the game. I did not come up with complete stats for every possible choice; that would have been crazy. Once the leader made an "official" choice and the NPC agreed, I would fill in the NPC's stats. There were several followers that my boyfriend designed for other players to take and his builds were reasonable, so I allowed them. One leader showed up later, at ninth level, so her followers and cohort were given details before she entered play. I found that when PCs gathered followers along the way, the whole process had a more organic feel.
Since I wanted a more personal approach, I made some adjustments to the sheer number of followers. First, I allowed players to level up followers they already had instead of gaining new followers at higher levels. Next, I came up with a way to slightly reduce the number of followers without disadvantaging the PC. After a leader got 10 first level followers and was entitled to 5 more, they make a decision. They could gain five new first level followers, or they could add one level to five of the first level followers they already had. I did this at regular intervals - once between a Leadership score of 13/14, once between 15/16, another time between a score of 17/18, and so on. This left fewer first level followers and built up established characters. Most of my players chose to reinforce their current followers rather than gain more first level folks. Additionally, I ruled that leaders could only have one cohort apiece. We typically had two characters with Leadership in our group of three to four PCs; more than one cohort apiece would have taken too much away from the PCs.
I found that using different races could add a sense of variety and excitement to followers. I had been happy to use one follower with a level adjustment (a tiefling), some followers from "half" races (a half-drow, a half-orc, some half-elves), as well as two followers with +0 level adjustments. My players were pleased, too. At the same time, however, I never wanted a leader to have a zoo, nor did I want them to "collect" rare races with little rhyme or reason. I did my best to make the variety of characters reasonable; since the PCs traveled often and far, however, they had more chances to meet exotic people. In my main campaign, the majority of followers were generally going to be from the most prevalent races and classes. I always said that i1f I ever get around to doing a Planescape campaign, I would use monster levels to foster variety and play up the exotic quality of the setting.
I kept the default modifiers intact, for the most part. I was willing to give up to a +2 modifier for truly outstanding generosity on a leader's part, but I was also willing to impose a -2 penalty depending on the type and measure of a failure. I felt that a penalty should be applied if leaders got their followers killed, rather like the penalty for getting a cohort killed. Since it never happened in our game, I did not had the opportunity to test this, but I had been considering a -1/2 point penalty per follower. These penalties were not necessarily intended to be permanent, however. A leader who took pains to make amends would have been able to work off their penalties. A leader who was able to outrun their reputation might have be able to start fresh and avoid the penalties, at least until their old deeds were uncovered.
If you want a much better version of Leadership, purchase the third chapter of my drow book!