Agile and Fixed Cost Projects

Tom Breur

23 October 2016

People have employed “Agile” with some success in fixed price contracts. Or so it seems. I have read claims from (in particular large) consultancy firms that they apply Agile within the context of fixed price contracts, and they use this to get to the desired end product quicker, more efficiently, etc. My motto is: “Don’t argue with success!”, but I have always wondered how those dynamics are supposed to work. To me it seems that Agile fits the “time and materials” model really well, helping the client prioritize. But how should it work with fixed contracts?

The essence of a fixed price contract is that the client knows upfront how much they will be paying for some deliverable. What puzzles me is that a customer knows how much “it” will cost, but Agile does not appear particularly well suited to specifying what “it” is! The whole idea is that you fix schedule and resources, but that you vary the scope! With fixed scope, you are “stuck” in the old ways of spending excessive effort trying to specify beforehand what “it” is, exactly, without having done much of the work. I often find that it is only through the building and the partial delivery that those requirements, specifying “it”, become clear.

With some government and military contracts there is sometimes even a legal requirement to have clarity upfront how much any deliverable will cost. This makes intuitive sense form the outside, but those involved in large and complex projects know how futile those efforts can be. And many big military projects, like developing jet fighters, have spectacular cost overruns to prove my point. With planning of public infrastructure the track record isn’t much better.

To the public eye, these massive cost overruns don’t look good. So their concern for accurate estimates is entirely legitimate. Personally, I don’t quite see how you can get to “better” estimates in any rational way, other than by reducing the size of the project, and beginning piecemeal delivery early, to get valid empirical data on completion rates. But that’s just me, I guess…