Cost Of Delay
Usually it's just linear (value/week), but sometimes there are dates-of-significant-change
- "this feature is much more valuable if done before Sept01 when school starts"
- conversely, that often means "there's no point in having this feature done before Sept01"
- which means you have to kinda simulate your queue so you don't build that feature too early (at the expense of features that would be generating linear value in the meantime) while not being late. Which means you're counting on accurate feature-level Estimating, Planning your Development Queue schedule out in the future, etc. Which gets ugly. So SWAG conservatively.