In [57]: df.groupby(['cluster', 'org']).mean() Out[57]: time cluster org 1 a 438886 c 23 2 d 9874 h 34 3 w 6