Choosing m=97 as modulus for the computation of the check digits in an account number of a Belgian bank account, it was assumed that m had to be selected from the set of primes. The paper shows that this restrictive assumption is incorrect by introducting m=93 as a more efficient modulus than the current m=97. The advantage of using m=93 appears in the capacity of detecting two altered digits. It also presents a slight modification of the check system, yet more reducing the probability of an undetectable error pattern. The paper may be of practical interest for banking in those countries where a national standard structure of account numbers does not yet exist, but where its establishment is given serious consideration to.
