You can imagine an optimization problem as a landscape, where higher elevation represents a better result, such as higher enjoyment. You often face so many possibilities that you can't see much of the landscape. So as a simple heuristic, you might just keep heading higher until you reach a peak.

Obviously though, some peaks rise higher than others. Perhaps you're climbing up the wrong hill, such as with:

1. Competition plain water - you (too) serious?
2. Tinned seafood - we have fresh seafood now!
3. “Healthier” “natural” sweeteners - probably better to get fewer sweets and more vegetables

Like gradient ascent/descent in machine learning practice, you should consider big changes over incremental improvements every once in a while.