When proposing the Analytic Hierarchy Process (AHP) as a means of prio
ritisation of a number of issues, T.L. Saaty [1] suggested that a flat
10% inconsistency tolerance be permitted in a pairwise comparison dec
ision matrix. This 10% figure has since been utilised almost universal
ly and without question. In this paper it is shown that any such limit
, based on a fixed percentage, is inequitable to matrices of different
orders to an extent that might call into question the final result. A
n alternative criterion, which incidentally provides greater flexibili
ty to the user, is presented.