The values of names, addresses and dates of birth for identifying individuals uniquely depends on several factors. This note shows how to combine the inherent power of an identifier, its availability and its error rate to give a practical measure for ranking its usefulness. The theory is developed in terms of selecting a number of non-unique identifiers to form a composite key with sufficient redundancy to identify individuals in an automated system. The note concludes with an indication of other classification, allocation and decision processes to which the technique may be applied.