The internet is chock full of anecdotes about various nootropics, but very little empirical data. Gwern (linked above) tries to blind himself in some of his experiments, which is good, but still N=1.
What needs to happen is, IMO, people interested in nootropics sign up for a "trial", pay for the drug and any minor equipment needed to test efficacy, the trial administrator gets a pure sample of the drug and sends it out blinded to the participants. Then the results could be collected and possibly even published (in a journal that doesn't have an IRB requirement).
It could go even further if people could cough up the cash for a cheap genetic variation test like 23andme, and then the efficacy could be correlated with genetic variants.
But as it is, we are in the absolute dark ages about our knowledge about nootropic efficacy in normals. It is complicated by the fact that the most likely candidates are often controlled substances.