If you were taking a sub-d approach, I don't see anything in this particular model which requires tricky transitions from high-density to low-density mesh, especially if you can model separate parts as separate parts, which is the general rule, unless you need a manifold surface (for example, for printing).
You would be aiming for the lowest possible number of edge-loops to capture the curvature of the piece .. let's say, the seat-base...
..shown here with a) the model as made, as a cage over the subdivided surface, b) the cage conforming to the subdivision modifier, optimal display on, and c) the underlying subdivision, level 2.
So, if you chose to go that way, you would dissolve nearly all the edge-loops, keeping only those needed to capture the curvature and sharpen edges, assign a Subdivision Surface modifier, and apply it at the last possible moment, if at all.
If you were really pushed for your poly-count, completely flat areas isolated from the surrounding curvature can be N-gons. You could apply the subdivision, and dissolve all the edges in the flat area (maybe insetting the N-gon to protect the surrounding curves), and cut in just enough edges to get rid of smoothing artifacts resulting from triangulation.