I think the "minimal" needs to be looked as as "relative to what you could conceivably build" not as "relative to the market." You always want to build the minimal thing you can, that enables you to get the validated learning that you're after. It's just that if there's an established market and existing players, you probably have to build up to that bar, in order to learn about the bit you intend to do differently.

Or something like that... :-)

> TLDR: In established markets, viable != minimal.

So it's not possible to target a niche with a specialized product that only has 20% of the competition's features, but does those 20% properly for that niche?

Agreed.