Traditionally, statistical estimation of glycemic variability includes computing standard deviation of glucose readings or related statistics (eg, M value, mean amplitude of glucose excursions, and so forth). We advocate an alternative approach using risk measures of variability, which have substantial clinical and numerical advantages. In addition, continuous glucose monitoring (CGM) data have clinically important inherent temporal structure that should be taken into consideration. Thus, temporal variability methods are discussed for the analysis and interpretation of CGM output.