A very nice and complete paper on the subject was written by Damon A Muller at the Australian National University1.
> [...] empirical evaluations of the effectiveness of profiling are equivocal at best. Where such analyses do exist, they tend to be limited, and feature flawed methodology and low numbers of experimental subjects.
The paper goes in-depth in explaining what types of profiling exist, what challenges are there to actually measure the effectiveness and pretty much debunks profiling as a last-chance option not supported by hard scientific fact.
* * *
Reference
1: <