When you estimate tasks, should the estimates be done in hours, or in days?
As I see it, the big advantage of estimating in hours is that if you THINK in hours, you tend to get a more accurate estimate. There are lots of development tasks which will seem like they should take "no more than 2 days", but if you think about all the individual steps (I have write create the page and the new service. And the stored procedure. And I'll have to get a security review and a code review. And I have to remember to do the unit tests. Oh yes, and save time for bug fixes), the total comes out a big bigger.
As I see it, the big advantage of estimating in days is that it's quicker and simpler. If you team is sitting there arguing whether a task is 3 hours or 4 hours, then you're wasting time -- after all, development estimates are never THAT accurate anyway: we always need to allow for the unexpected.
Considering these, I could be persuaded to do it either way. What is NOT useful is to think how many days it will take, multiply by the number of hours per day, then spend time arguing about whether it is one more or one less than this number.