Sridhar Vembu’s Blog

One of our ServiceDeskPlus customers posted http://forums.adventnet.com/viewtopic.php?t=3899 (scroll way down) in response to the news that our release slipped

Yes, I agree, it is painful for us too to have to slip. Honestly, I personally hate to give out dates for this very reason. As a company, we also try hard not to promise release dates, because the added pressure just subtracts from quality, doesn’t get work done any faster. Software is just inherently unpredictable. One “small silly” bug might suck days and days to fix, throwing any kind of schedule into the trash. It is often unknowable in advance that the bug is not so easy to fix.

There is a fractal like or scale-invariant nature to this uncertainty. Even in fairly small scales (”This feature will be done by this evening̶ the chance of hitting it that evening are pretty low. It could be done a lot sooner, or could take an extra day or two. In other words, the range of uncertainty is often as big as the initial estimate itself. Due to the interdependent nature of software modules, things actually don’t average out - a slip in one module doesn’t cancel out the ahead-of-schedule finish in another module. Instead the worst case slips along the way tend to add up, and they end up determining the schedule. I am pretty sure someone could compute the probability distributions and show the scale invariance of a schedule, where the worst case delays add up at each step.

After a decade of experience, I have come to the painful conclusion that “Software Project Management” is an oxymoron. I don’t want to make fun of Microsoft, but as a software person, I feel their pain Their main sin is announcing dates years in advance, not slipping by a year (or two). Why announce something like “Office 12 will ship in 4Q 2006″? I scratch my head - how on earth can they possibly know that?