Joe is asking for ideas regarding time estimates.
First of all, read Planning Extreme Programming by Martin Fowler and Kent Beck. Even if you don’t want to do XP it provides excellent ideas for planning.
The book promotes “the gummi bears principle” meaning that estimates should not translate to calendar time. Instead stories should be measured relatively to each other. So if story A is a 2 and story B is slightly harder, then it is a 3. This requires an upstart period where you “calibrate” your measures. But after a while you get quite good at it.
To really get away from calendar time thinking, we give stories “time points” which can be from 1 to 5. This helps developers to not think in terms if time, but instead sort of grading the stories.
At the end of each iteration, we sum up how many time points we did. This is our velocity and stories for the same amount of time points are planned for the next week.
It is by no means perfect, but better than any other technique I’ve tried.