Those Cursed Project Estimates

 

I have been thinking and writing much about project estimation these days.

It seems to be an area where so much confusion occurs, starting with the guy who is typing right now.

Here are some of the pitfalls of project estimation, several of which I have fallen prey to myself:

Anchoring

This occurs when estimates get biased (deliberately or not) by some influence other than the knowledge and experience to yield a reasonably accurate estimate.  Teams may be asked to give a rough "back of the envelope" estimate just so management can make a decision and then when it comes time to do the real estimates, a bias from a false starting point is already there.  In my experience, these estimates tend to be way off the mark.

Anchoring can also occur by the way you ask for estimates.  I have recently realized that I sometimes prompt team members by venturing a guess myself when I feel fairly confident about the technical aspects of the work.  By doing so, I have been anchoring my teams' estimates unnaturally.

Overconfidence

I was once in the office of a customer discussing my team's estimates and my range estimate was a -5%/+10%.  One person commented that this was way to uncertain, after all, it was for a project only 3 years long working with complexities and lots of unknowns.  In retrospect, I was far too overconfident myself in those estimates.

Another form of overconfidence can come from the tendency to forget about the human factor involved with project estimation and rely too much on tools and fancy math.  If you have lots of well-recorded history then these models can work very well, but plugging in estimates without regard for the human influences (like anchoring) can lead to big deviations and general lack of rigor in your project estimates.

A few of the things I'm working on now include the following:

• Uncovering the best method for acheiving reliable estimates when you are in a dynamic environment with little to no past history to rely on.
• Heading off the anchoring problem, in all it's forms and all the sources where it can be a problem.
• How to accurately reflect the level of confidence and low/high ranges on project estimates.

What are your struggles with project estimation?  Do you have any thoughts to offer by way of extending what I've already said or responding to one or more of the points made?

Image by !!!! scogle via Flickr