In a lazy-by-default language (Haskell) you absolutely have to use those kind of approaches, because you can't write code where effects are important any other way. You have no control over evaluation order (other than dirty hacks), so you can't rely on that to control effect sequencing.

Of course in a strict language with a decent type system like Scala you eventually find you want to use that style anyway. But you can come to it in your own time and see the benefits for yourself rather than having it forced on you.