| Iambob doesn't account for the fact that interviewing candidates falls into a Simpson's Paradox for this instance. You have two cohorts which comprise a majority of your applicants: 1. Unemployed, X% qualified. Has time. 2. Employed, 100-X% qualified. Does not have time. X is small. If both candidates are qualified, tests for X also find how qualified they are so they can be compared. His methods are objectively better, provided that both candidates have the same amount of free time. If you compare two unemployed engineers, you'll pick the right one. If you compare two employed engineers, you'll pick the right one. If you compare an unemployed engineer vs an employed one, however, you'll probably pick the unemployed engineer, even if the employed one is better. Because he doesn't have as much time for your homework or to maintain a Github profile. Thus, if you're comparing 30 engineers, half of whom are employed, you'll pick the most qualified unemployed candidate, rather than the objectively most qualified candidate. That's why, even if the test is slightly worse, with an interview room only test you'll consistently pick the better candidate: a large, large segment of talented candidates are employed. I would love a statistical analysis of Simpson's Paradox and technical interview methods if anyone's ever done one. |