RSSDid the waterfall model fail?
Most people reading this blog will recognize what sort of waterfall I am actually talking about in the title, and it's not the sort you see in the picture. The waterfall model of project management is the one which we grew up with, nurtured and cherished, and ultimately turned on accusingly, blaming it for all our heartbreaking failures to complete projects on time, on budget, and to the satisfaction of the stakeholders. We've since cast about for new models, dallying recklessly with RUP and Agile, but still seem to fall back into the seductive simplicity of the charts and timelines of our first love (only to scorn it again, with horror, as we realize what we have devolved to).
Now Eugene Nizker at CIO Magazine introduces the thought that perhaps the good old waterfall, chaste and lovely in the morning light, was never really the problem.
Nizker's thinking is that the model itself actually has good applicability for certain processes, and that much of the blame it has garnered for failures have been the result of attempts to use it outside its realm of utility. This is hard to argue with; if you're using something and it doesn't work, it probably wasn't useful. But Nizker goes further and says we need to consider that maybe much of this is individual failure that the process gets blamed for. I suppose you have to make the distinction between having a hammer and being unable to hit the nail (you are a lousy carpenter) and having a hammer and being unable to turn a screw (you have the wrong tool).
But then Nizker goes on to define the nails and screws of the IT project world, and here I think is the most valuable takeaway from his article: the separation of project tasks into "deterministic" and "partially deterministic" types, the former of which may be succesfully governed with a waterfall methodology and the latter of which might be better managed with an iterative development process. I highly recommend reading his post for further details. While I suspect that his classification scheme may still be too simplistic for the problem at hand, it's an excellent starting point.
