Although single-stage Data Envelopment Analysis (DEA) models with undesirable input-output have been extensively studied, there still lacks more systematic investigation of a two-stage DEA model with undesirable variables such as non-performing loan (NPL) in Malaysia. The NPL concept has been widely discussed by economists since the economic downturn in 1997, which had immense impacts on many countries, including Malaysia. This paper studies the performance of commercial banks in Malaysia from 2013 to 2017 using a two-stage Data Envelopment Analysis (DEA) model. The main objective of this paper is to measure the performance of commercial banks in Malaysia by considering NPL and next to determine the suggested value for the selected variables in order to improve the inefficient banks. The findings show that domestic banks were generally more efficient than foreign banks. The results for the whole process show that domestic banks also performed better than foreign banks in both stages. The two-stage model used in this analysis enables inefficient banks to look at where the problem occurs, either in stage 1 or stage 2, thus allowing them to determine the suggested values in that inefficient stage that can bring them towards becoming fully efficient banks.