When you produce, say, 42 separate miniatures SKUs in a year, you have to get the numbers right 42 times. (And since the products are distinctly different, getting them right is a real challenge. Guessing which of the Heroes SKUs is going to be insanely popular and which is going to languish in the shelf is virtually impossible on a regular basis.) Every item that doesn't sell through 100% is a tax on those that do sell.
My wife works in retail, and you see this a lot in toy merchandizing... Star Wars Episode I hits the screens and the Qui-Gon, Obi-Wan and Darth Maul action figures fairly fly off the shelves and sell out in a week and a half. Six months later, all the Amidala, Jar-Jar and Palace Guard #3 action figures are languishing on a clearance rack for 75% off.
It makes me wonder, though... They've already gone sort of half-way, with non-random PC packs, and a single visible Large creature in each monster pack. I wonder if it would be worthwhile for them to take some of the (more or less) guaranteed sellers and make themed packs out of them.
For example, something like a non-random goblin pack with 1 of each non-minion, and a half-dozen of the minions... or something similar.