This paper uses Data Envelopment Analysis to evaluate the operating and scale performance of Taiwan’s local banks for the year 2000, and explores its implication for the management. The input variables include the net worth, operating expenses, personnel expenses, interest expenses and the branch number; had the output variables are investment incomes, interest incomes and fee incomes. The banks are classified by the total assets and by the bank’s special features. The results how that: (1) Four kinds of banks based on total asset have significant difference in technical efficiency, pure technical efficiency and scale efficiency. Those banks with total assets between NT$210 billions to NT$800 billions have the best efficiency performance. Efficiency measures do not increase with size for the larger banks. So merging or expanding branches may not improve performance. Smaller banks are in increasing returns to scale, and should expand for better performance. (2) Five kinds of banks based on the setting up years and target customers have significant difference in total efficiency and scale efficiency, but no difference in pure technical efficiency. And the earlier to set up the banks, the better the scale performance. We also calculate the slack variables for inputs and outputs and make suggestions for those banks that are not efficient.