large convention theory
Magic things happen to SF conventions at around 3000 people. Where exactly is hard to tell; Boskone and Minicon both jumped from noticeably under 3000 to substantially over in a single year. But two things happen. First, management tricks to shoehorn the staff into a structure appropriate for a 2500-person con stop working. Another level of indirection, of convention management, is needed; you need a larger cadre of experienced managers, and they and the chair have to think in a more abstracted manner. At the same time, your volunteer base gets distracted; there are just as many or more of them, but they're not willing to give you their weekends.
Put more concretely, a convention has a run-time volunteer need of a little less than two hours per attendee. Arisia meets this need by having 100 people who work 25 hours each, and 150 who work 12 hours each. We push our management structures to be able to handle that many people; this model might work for 2700 but it wouldn't work for 3000, because three-level management can only handle 250 or so people. But with new incentive and a four-level management structures, we could have the same 100 people working 25 hours each, and 500 people who worked 4 to 8 hours, and that would take us to 3000 people pretty comfortably. The doubling in the size of the staff requires rethinking how volunteers works, but that's a rethinking you have to do to get past 2500 people anyway. Once you do it, the management reasons to keep your volunteer count down are obviated. By adding more casual workers, this management structure scales from here to probably around 5000 attendees before you have to worry about incenting your workers to work longer hours. Which is good, because it's a lot easier to find people who'll work a couple of hours than people who'll work all weekend.
Will we reinvent volunteers this way? We don't have to; at this size we have the luxury that the current structure does more or less work. But if we get a lot bigger, we'll have to lower
our incentive requirements, and that's one of the counterintuitive conclusions that I think other big conventions like Boskone and Minicon missed.