the insert is from a bpcs system. approx 27,000 rows of data used for calculations of stock levels. After which it is deleted and inserted again (every hour)

Since this put a large strain on redo - the plan was to remove at some point.

I figured since it was the biggest shared pool user, it would contribute the most to fragmentation - as it grabbed and released a lot of memory at intervals.

The insert is a standard insert - no parallel