Background and Objective: To describe donor behaviour quantitatively and apply the information on actual donations derived during observed changes in blood collection in Melbourne between July 1990 and June 1996 to construct a logical model to examine the effects of changes in donor management on the sufficiency of the blood supply. Method: Computerised donation data files were searched to determine time to next or previous donation for donors giving blood between 1990 and 1996 stratified by age, donation history and venue. Actual numbers of whole blood donation given by new and repeat donors, July 1990 to June 1996, and numbers of donors refused were used to construct and test a logical predictive model. Results: 558,682 donation intervals were analysed. Donor return rates at 2 years were shown to increase with donor donation history and donor age. The prediction model showed that a 25% decline in whole blood collections over 2 years could be explained by the cumulative effect of a decrease in donor return rates of 2-4%. Between 1990 and 1996 many donors moved from static city collection sites to suburban mobiles. New donor attendance correlated with repeat donor attendance. Donor complaints correlated with donor referral numbers. Conclusions: The model showed that large shifts in nett blood collections can be explained by relatively small shifts in donor return rates at 2 years.