Practically, that's not too far off. Arrows, at least as implemented currently, have a number of significant restrictions based on the need for the `arr` combinator to accomplish much of anything; Applicative has, as a result, largely taken over the spot in the Haskell ecosystem that Arrow was originally intended to fill.

My understanding is that there's a rework of Arrow in progress that may change this in the future, since *theoretical* Arrows are more distinct, flexible and useful than the current implementation.