Problems arise when people try to estimate tasks that are new or difficult. In web development, for example, I can sit down and estimate with a fair amount of accuracy how long it takes to put together a struts-based jsp form backed by an ejb call or two. Why can I do this? Because I’ve done about 60 of them in the webapp we develop at work. I have already worked through all the insipid monkey wrenches that have burned through my time in the past like browser incompatibilities, css annoyances and many other time-wasters that I know as well as I know the plot to a repeat sitcom.

So, what screws me up? Anything new. Last month I had to modify a bunch of VBA macros in an excel worksheet. I had worked on it before about 2 years ago, but not since and had basically “forgotten” VBA. My initial estimate: 2 days. My actual time: 2 weeks.

I imagine that I would get similar horrific mis-estimates if there was a twist in the web-app, like, for example switching IDE’s, moving to struts 2, using a new javascript library, etc.