It’s hard to improve on Prof. Ed Kitch’s critique of the weak chain of inferences that warrants using commercial success evidence as proof that an invention would have been nonobvious:

But how is commercial success relevant to non-obviousness? … This argument involves four inferences. First, that the commercial success is due to the innovation. Second, that if an improvement has in fact become commercially successful, it is likely that this potential commercial success was perceived before its development. Third, the potential commercial success having been perceived, it is likely that efforts were made to develop the improvement. Fourth, the efforts having been made by men of skill in the art, they failed because the patentee was the first to reduce his development to practice. Since men of skill in the art tried but failed, the improvement is clearly non-obvious.

Each inference is weak. … This seems a fragile thread on which to hang a conclusion of non-obviousness, particularly in a case where the patentee shows only commercial success but does not show that the commercial potential was perceived or that attempts actually were made that failed.

Edmund W. Kitch, Graham v. John Deere Co.: New Standards for Patents, 1966 Sup. Ct. Rev. 293, 331-32.