We rank support departments when using the step-down method. Alternate ranking criteria include department size, number of other departments serviced, and arbitrary ranking. Which method will usually produce the smallest errors, relative to the answers obtained from the conceptually accurate reciprocal method?