If you have a project with an estimated cost of $1MM and you expect a return of $1.1MM then strict governance and process controls are critical to success. A variance of 10% will take you from profitability to a loss.

If you have a project with the same estimated $1MM cost but an expected return of $50MM then project management is less critical. a 10% variance will not really matter in the big picture. Given my fictional example even a 100% to 500% cost overrun could happen and keep the "success" label.

So the take home message is that project management is critical for useless projects. If you have a great project then even with terrible project management you will still be successful.

In my own experience we run a might tighter governance on the weekly 'Features and Maintenance' sprint team as opposed to the month long 'bigger feature' sprint.

So put your bad PMs on key initiatives and shuffle the best to the menial tasks.

And get yourself on a 'no-fail' path by ensuring that all your projects have large ROI with logarithmic (hockey stick) growth.

I love reading articles by fearless people with 'tenure' at IEEE!

http://www2.computer.org/portal/web/computingnow/0709/whatsnew/software-r