The groupby is the right idea, but the right method is cumcount
:
>>> product_df['month_num'] = product_df.groupby('product_desc').cumcount()
>>> product_df
product_desc activity_month prod_count pct_ch month_num
0 product_a 2014-01-01 53 NaN 0
3 product_a 2014-02-01 52 -0.018868 1
6 product_a 2014-03-01 50 -0.038462 2
1 product_b 2014-01-01 44 NaN 0
4 product_b 2014-02-01 43 -0.022727 1
7 product_b 2014-03-01 41 -0.046512 2
2 product_c 2014-01-01 36 NaN 0
5 product_c 2014-02-01 35 -0.027778 1
8 product_c 2014-03-01 34 -0.028571 2
If your really want it to start with 1 then just do this instead:
>>> product_df['month_num'] = product_df.groupby('product_desc').cumcount() + 1
product_desc activity_month prod_count pct_ch month_num
0 product_a 2014-01-01 53 NaN 1
3 product_a 2014-02-01 52 -0.018868 2
6 product_a 2014-03-01 50 -0.038462 3
1 product_b 2014-01-01 44 NaN 1
4 product_b 2014-02-01 43 -0.022727 2
7 product_b 2014-03-01 41 -0.046512 3
2 product_c 2014-01-01 36 NaN 1
5 product_c 2014-02-01 35 -0.027778 2
8 product_c 2014-03-01 34 -0.028571 3