How to group by the dates so I can filter out the data without aggregating in pandas?

This data contains the retention accounts with bonus loyalty offers. Metrics are aggregated for all accounts in 2017. Full year of 2016 is included to provide a baseline for account performance prior to their attempt to close.

Here is the dictionary just in case:

I am trying to find answers to questions like:

1. 2017 retention rate by month
2. Estimated profit logs from account closing.
3. Estimated profit generated by the program in 2017 in aggregate and per account.
4. Estimated payback period for retention account(like the amount of time to recoup the issued loyalty)

I don't understand the concept of TIMEKEY, how man I supposed to groupby all the timekey months and filter the data out in pandas? the groupby in pandas requires some kind of aggregation but I just want t开发者_开发知识库o group the dates.