I was under the assumption that Hardware Instancing was different from standard batching - for instance, in that link it looks like Static and Dynamic batching methods are CPU calculated and sent to the GPU whereas the Geometry Instancing API is the actual hardware instancing, and the Vertex Contants somewhere in between.
I've only just started looking into dealing with the GPU, so please forgive my ignorance.

Here's a practical demonstration of combining meshes and textures to reduce the number of draw calls and improve performance as a result.

Note the comments on Occlusion Culling (not applicable in our case).

draw-call-minimizer

Static or dynamic batching and hardware instancing are other ways of trying to maximise the number of objects drawn per draw call resulting in performance gains.

Thanks Pixel, that helps.