I don't much like this approach but in one case we've created 12 separate input tables, one for each month, and then append them all into a central table. That gives you a lot of flexibility for building an identifier for each row including the date, looking up against the previous period's identifier, etc. and you'd then also want to filter to only keep the current month's row. This can bring problems with making sure you load the data in the right place each time, but that's one for your run book.