
Calibrate on real signal
Judge profiles the way you would before submitting to a client. Consistent feedback on a handful of profiles moves the ranking more than scattered feedback on many.When to refine the brief instead
If most of the shortlist misses in the same direction, the brief is the problem, not the ranking. Add the constraint that’s implicitly missing and run again.Calibration tunes within a brief; it can’t rescue a brief that describes a different role.
