The Risk Of Accurate Time Estimates
Getting accurate time estimates may not always be a good thing. The reason is that the accuracy may come at the cost of productivity, much much beyond the productiivity losses due to spending time creating estimates. Time estimates may get more accurate because people are actually doing less work.
Imagine you are a manager. Imagine you put pressure on the people you manage to give you accurate time estimates, for how long a task or a project will take.
Imagine that things like promotions, pay increases, firings, etc are based on how accurate people's time estimates are.
Let's say that this pressure works, and causes all the people you manage to give accurate times estimates That's a good thing right?
The answer could actually be, no. Here is why....
Steps To Get Accurate Time Estimates, The Wrong Way
This is how someone, faced with this kind of pressure, could get most of ther estimates to be very accurate.
Step #1: Over estimate everything.
Step #2: If you complete your work before you use up all your time, don't go ask for any more work or move on to the next task (yet); just slack off and let the time run out. (And, after the time runs out, then go ask for more work or move on to the next task.)
(The 2nd step -- step #2 -- is really the important one.)
And then time estimates are accurate!
The problem is that the amount of work being done has gone down!
In any place of work, that isn't completely dysfunctional, this is not what you want.
Therefore, it is important to not prioritize accuracy of time estimates above all else.
If you punish people for finishing early, don't be surprised if they stop "officially" finishing early.
(BTW, this isn't just some theoretical musing. I have already seen this happen at one large bureaucratic company.)
Don't seek accurate estimates unless you actually really need them. Is this a project for a client where they are demanding to know how long the project will take? If not, what is the point of getting time estimates?
(If you need to, there are better ways to motivate the people you are managing. Or if possible, only hire people who are self-motivated and don't need to be managed.)
If you still want to do predictive analytics on how long things will take on your own, then measure how long the people you are managing are taking with tasks and projects, and then you (the manager) should make your own times estimates for the people you are managing based on the data that you (the manager) collected. (And remember, don't use these predictions that you (the manager) are making as a "sword", to give pressure, else you run the risk of having the problem described in this article happening.)
Also be aware that accurate time estimates may not be possible in every situation. (P≠NP would suggest this by itself.)