Get the number per order rather than per month, because the monthly figure hides the thing that decides this.
For a representative delivery order, write down: menu price, minus commission, minus the packaging, minus card and platform fees, minus the food cost. What is left is your contribution from that order. Do the same for the equivalent order over the counter.
What people find when they do this honestly:
Some items survive and some do not. High-margin, cheap-to-make items still contribute after commission. Low-margin items — anything where food cost is already a large share — often contribute almost nothing or go negative once packaging is counted. Averaging across the menu hides this completely, which is why the monthly view feels ambiguous.
Packaging is bigger than expected. It is a per-order cost that does not exist over the counter, and for drinks and anything requiring insulated or leak-proof containers it is not small.
Labour per order is higher, because delivery orders arrive in bursts and cannot be batched the way a queue can.
The useful action from this is usually not leave or stay. It is change what you sell on the platform. Take off the items that lose money, put on the ones that work, and price the platform menu to cover the commission. Most people are running their counter menu at counter prices through a channel that takes a large cut, and that is the actual problem.