Whenever this topic arises, I always find myself returning to the idea of 'hire lightly, fire lightly'.<p>That you can 'measure talent' by arbitrary nonsense like whiteboarding or regurgitation of CS curriculum has been demonstrated as a fallacy so often that it <i>should</i> be understood by now--but it isn't.<p>So: change the shape of the talent pipeline. Hire lightly, give them a chance, evaluate, learn. Fire if necessary. Sure, go ahead and look for patterns and potential metrics, but don't overthink it.