Like I said, low hanging fruit goes first. Later efficiency gains will be more expensive.