There is one simple technique that would allow nearly any forensic analysis technique to be assessed. It would detect the difference between good science (DNA) and junk science (forensic hair matches). It would work down to the level of the individual laboratory (as in this example) or individual laboratory researcher (Annie Dookhan). And it would only cost 10% extra.

That technique is blind testing with published results. 10% of all submissions to the testing lab would be test samples -- a mix of a bunch of known negative samples (from individuals completely unrelated to the crime in question) and a few known positive samples (perhaps from someone already convicted with overwhelming evidence). The lab would not be told which samples were tests and which were "real", and the test samples would be evenly distributed (10% of the submissions from each department; 10% of the "rush" submissions, etc.). Finally, the results reported on these test samples would be made publicly available.

Although this simple, effective, and cheap method would completely transform forensic practice, it will not be implemented. The reason is incentives. The parties involved in forensics are the police or prosecutors and the labs. Both have an incentive to identify the guilty. The accused have a countervailing interest (in correctly determining innocence), but they are not part of the transaction, so their interests are not represented. In fact, the entire system is incented to find people guilty, not to correctly assess people's guilt. So most in the system are actually DISincented to discover and debunk junk science.