Compared to construction, where requirements are more often well-defined upfront, software is more ... fluid.

You don't know what the customers want.
You don't know what product to build for them.
You don't know what software architecture is best.

But as time goes on and we keep building, we'll learn these things.

That's why before a project starts, it's anyone's guess how long it will take.

But once you've been working on it for a while, things tend to get more well-defined.

You choose a technology.
You choose a problem to solve.
You get feedback from the users.

So, what we should optimize is learning, and the way to do that is through continuous delivery.

I've written about the importance of fast feedback loops before.

I got the best outcomes from asking my team: "What do we need to build to learn about [what the customers want, which technology is right, how to scale,...]?" and then let them come up with proposals.

Yours,

Taj