In my experience I haven¹t seen Agile approaches used for custom software
development worth much more than $2m, in those cases the key was in the
customer having utmost confidence in the solution and development team.
Yes, many times and by more than one order of magnitude.
I've seen systems of equivalent complexity to healthcare.gov built with agile techniques in 10% of the calendar time and at something like 1% of the cost. How much of that was a function of using agile instead of waterfall I don't know. The key here may be that the developers were respected employees of the business customer, not contractors. They (we) were able to participate in requirement design, ship useful functionality incrementally when it was ready, prioritize features, and push back on unreasonable or unwise requirements.
I suspect the fundamental issue at the heart of healthcare.gov and a lot of government projects (by no means just in software development, or just in the U.S. Federal Government) is the reliance on outside vendors instead of developing qualified in-house teams and talent. Not that building qualified teams is easy, of course, especially when you're limited by government salary scales and regulations. But privatization and outsourcing of core competencies is far from the most cost effective approach imaginable.
