People examine data manually in reports currently. If an item turns a profit, leave it, no yield optimization. If it's in the red, stop it. If it's around break-even, schedule it for examination to see if you can turn it green. No idea how close to optimal it is. I'd imagine not at all.
OK, I'm looking in from the outside, and I know all too well how these sorts of situations arise. However, it sounds to me a lot like the managers have largely lost control of the situation, and they are limping along doing the best they can with a system that has evolved from something system into a monster.
That happens. The challenge is how to make it better.
Thinking of it like software, there are no unit tests, and touching anything runs the risk of breaking everything. The usual way to proceed is to find a small section that you can write tests for, and then gradually transform into something you understand. The equivalent here might be to find a small collection of items that are largely independent of the rest of the product line, and see if you can bring that under control.
However, I don't know your context, and I do appreciate that wholesale changes might well be risky. The question is, can you even make a start on a small section? If not, all is lost and the system will limp along as is. But if progress is desirable, maybe there is a way to start.
Given that your senior colleague has actually spoken with you about this, maybe you can open that conversation, proposing it as a gradual way to make the system more efficient, profitable, and to reduce the existing risks.
Again, on the political side, make sure they don't feel threatened, bring them on board as a partner, make it clear that you want to learn from them, and couch it all in terms of reducing risk to the company in return for a small outlay in time and effort. See it as an opportunity for you to learn, them to gain credit, and the company to become stronger.
These are my thoughts - you must adapt them to your context.
