Why are government software systems so expensive?
Well, the agency publishes an RFI, and vendors respond saying how much they think the system should cost. Agencies take those responses, average them, and tell their legislature or a federal grant-maker “it’s gonna cost $50M.” So $50M is given to the agency—as a matter of public record—and the agency issues an RFP. The quotes are all going to be for real close to…$50M. And where did that price tag come from? The very vendors who are bidding.
😳
So how do you intervene here?
One way is to stop regarding the cost of major software as some sort of a mystery price that cannot be internally validated. If four quotes for a 1,500 square foot house all came in at $5 million, the conclusion shouldn't be "this is a $5 million house," but instead "these quotes are absurd." Agencies need the knowledge to inspect quotes and determine if they're internally valid.
@waldoj What kind of metrics would you recommend?
If it were me, I'd want a function point analysis done before asking for quotes or ask the vendors to supply one. The number of function points could then be compared against known data (e.g., from the IFPUG) to get an idea of the cost of such a product.
@lwriemen One might reasonably object “but nobody can possibly estimate the cost of a project that huge with any accuracy.”
@waldoj I highly doubt "scrum-team year" is a standardized metric. If you haven't yet, you should read one of Capers Jones' books on software estimation.
As long as people keep recommending non-standard metrics, the software world will continue to be faith-based rather than scientific in choosing methods, processes, languages, developers, etc.
@lwriemen Function points are not useful to legislators or governors’ budget offices. As a species we have figured out how to estimate software complexity. That’s not the need here. It’s how to allow non-technical government leaders to grasp complexity for purposes of budgeting and oversight. Function point analysis ain’t it.
@lwriemen I think analysis is meaningless for existing software (COTS)—in that case, the price is set by the market and is unrelated to the complexity of the software. $100M software projects are custom(ized) software. A simple shorthand I recommend is estimation of the number of scrum teams and the number of years of work required to complete the work and comparing that to bids. How much are they asking for per scrum-team year?