I think there's an intuition and a sense that it's the way to go, but the data are extremely strong, showing that software projects should really not exceed.... Well, the limits that we give in our paper, which borrows from the U.S.'s General Services Administration, the suggested rule is a maximum of $2 million U.S. per year for no more than three years, and with no extensions beyond that. The reason software projects in particular demand that small scope relative to, say, big infrastructure projects or something, is that, when you're talking about building software and services that are built out of that software, you don't really know what you need until you test early with users.
There was a shift in how we thought about software development in the private sector, with what's called the “agile manifesto”—which you can find online—and it changed how the private sector and leading public sector jurisdictions think about software projects. You start small, perhaps with many contracts at once, and you bring these vendors together, in part, because you're not pooling and making tons of unfounded assumptions about what the end product's going to be.
Now, to enable that way of contracting, you have to look at how money flows through Treasury Board and how budget submissions are done. You have to create space for public servants to get that early feedback and adjust. However, it has happened in many other jurisdictions. It's really remarkable how far behind the Government of Canada is in moving their software development practices to what is well-supported in the data in both private and public sector corporations, which is that you have to keep it small and test and adjust as you go.