The seafood linguine sells forty a week and is priced on a recipe cost worked out in March. On paper it runs at a comfortable margin and it is one of the reasons the kitchen keeps it on.
What the March calculation does not include is that two thirds of the orders come with extra prawns, and that the prawn price moved twice over the summer while the menu price did not. The dish is now the least profitable thing on the menu and the most popular, which is the worst possible combination.
Nobody spots it because the property measures food cost monthly, as one percentage, across everything. A single dish drifting is invisible inside a blended figure — it just makes the whole number a little worse and gets blamed on suppliers.
Per-plate costing changes the question from "what is our food cost" to "which dishes are carrying it". Modifiers carry their own cost and their own stock movement, so a heavily-modified dish shows its real margin rather than the margin of its base recipe.
Why this matters to you: menu engineering is the highest-leverage change most kitchens can make, and it is unavailable to a property that only knows its aggregate. You cannot re-price, re-portion or retire a dish you cannot see.