In this paper we propose a new CFAR detector based on ordered data variability. The proposed detector, designated as ODV-CFAR, uses the variability index statistics to select dynamically the whole reference cells, the smaller ranked cells or the larger ranked cells to estimate the unknown background level. The performance of the proposed detector is evaluated and compared with those of the ACCA-ODV CFAR and the trimmed-mean (TM) CFAR detectors in various background environments. The results show that the ODV-CFAR detector acts like the CA-CFAR in homogeneous background and performs robustly in non-homogeneous environments.