Even if IBM® Campaign created a temporary table for these customers, joining it back to the purchase transaction table would entail pulling over approximately 90% of the 4 Gb rows (and discarding all transactions except for the last three months) to execute a GROUPBY macro, for example.

Instead, you can configure an Extract process (placed at the purchase transaction level) to pull out all transactions within the last three months, put them into a table in the database, and then subsequently run multiple GROUPBY macros and other calculations against it (for example, min/max and average).