The MVP became a licence to ship things nobody wanted

Ries defined the minimum viable product as the smallest experiment that produces validated learning. In practice the term got used to justify shipping a thin product and calling the resulting silence data.

The distinction the book makes and most teams drop is that an MVP is an experiment with a hypothesis attached. Without a stated hypothesis and a threshold that would falsify it, you have not run an experiment. You have launched something small.

There is a second problem. Iterative validation works when customers can evaluate what you show them. It works badly when the value only appears at scale, or in categories where users cannot articulate the need until the thing exists.

Where have you seen the method fail honestly? Not because it was applied badly, but because the product was the wrong shape for it?