I like this answer, but it leaves one important thing out. Mathematicians have been studying abstraction for thousands of years. When they decide something is u

I like this answer, but it leaves one important thing out. Mathematicians have been studying abstraction for thousands of years. When they decide something is useful, that probably means it has broad applicability and enough functionality to get interesting results out of. Mathematicians decided monads are useful. Take advantage of that. Even if you're working in a language that can't express the abstraction, you can allow it to inform your library design. If something could be a monad, that's usually a better choice than something ad-hoc instead. (Sometimes it's overkill and a monoid would suffice, but both are better choices than an ad-hoc interface without any broadly accepted theoretical basis.)