Imagine that an architect were tasked with building a house, came up with a design, had a team of builders construct the house and then asked the prospective owner whether anything needed to be changed. The prospective owner tells the architect that he doesn’t like a part of the house and would rather it were done differently. To please the owner, the architect knocks down that section of the house, rebuilds as directed and then repeats the process until the prospective owner is satisfied.
To any sane person, this situation would seem completely ridiculous. It would be stupid to go through all the effort of constructing a building without first confirming whether everything was acceptable at the planning stage. Strangely enough, this is not far off from the common practice in science. We design and perform experiments, collate and narrate the results and then send our write-ups to journal editors, who in turn distribute the manuscript among a coterie of our peers. These peer reviewers supposedly check the manuscript for publication-worthiness, advise us of errors in execution or interpretation, and recommend various changes and additions. Frequently, this turns out to be a prolonged exercise that incurs significant expense, but does not add much value to the central theme of the manuscript. Without a doubt, the scientific process can benefit from well-performed peer review. However, extant peer review processes are more wasteful than beneficial as in the analogy above.