Real-time operations are costly to manage even though the software has become easier to build. Not being Amazon and not being able to control inventory, and access to that inventory, hurts. However, probably a bigger issue is establishing consumer habit - triggering that need to order at the moment the user wants it. This comes down to establishing a brand (expensive) and also evolving consumer behavior (more expensive). And, these consumers will have to be in the < 200 geographical regions of the US that are dense enough to facilitate < 4hr delivery profitably (its probably less than 50 regions where you could fulfill on 1-hr delivery profitably at any substantial volume).
You could say Uber is a counter example of succeeding in this space. I would argue that they replaced an existing consumer habit with a better one. I don't see that in the same-day delivery space - thus the need for education.
There have been a few generations of these services (Webvan / Kozmo; Licketyship; Ecourier / Shutl; Postmates). Eventually, the cost of tapping into consumer habit vs the small margins of brokering a delivery will create an amazing opportunity. In the meantime, the cost of educating and acquiring users is a big deterrent and if you can build the kind of company innovative enough to solve this problem, its likely you have a prohibitively high opportunity cost.