Aleksandr Kogan has been talking down the value of his data. Is he right to do so?