They can't consider the impact of optimizations immediately or the interrelationships of items.
Assume you have a 30% profit margin goal. You decide to start by optimizing landing pages in your campaign first. Based on the tracking platform report, you would have paused Landing Page #3 (25% profit margin) as it falls below your desired campaign goal.
By removing Landing Page #3 from rotation, your campaign's overall profit margin would now be 51.67% based on the remaining landing pages. Put differently, the estimated impact of the optimization would be a 26% increase in the campaign's profit margin (from 41% to 51.67%).
It is worth noting that pausing Landing Page #3 does not impact other landing pages' performance (items of the same type) - just as pausing one placement (ie. "miniclip.com") would not impact another (ie. "yahoo.com").
Next, you decide to optimize the device types. Based on the tracking platform report, you would traditionally pause the mobile audience as its 25% profit margin falls below your campaign goal.
However, based on the prior optimization - and, for simplicity, assuming equal landing page distribution across all devices - the profit margin should have improved by 26%. If we consider the estimated impact of the prior optimization immediately, the profit margin on mobile devices should be 31.5% (25% mobile profit margin x 1.26% estimated impact of prior optimization).
By considering the estimated impact of optimizations immediately in subsequent calculations then, we would have prevented prematurely pausing an entire audience of mobile users.