One of the more disappointing aspects of this go-around of delivery services is that they don't seem to have learned anything about logistics. Logistics is not just solving the Traveling Salesman problem.
If I could miraculously solve a TSP in linear time, that would be an amazing accomplishment. But what good does it do you if the fixed costs of loading and unloading at each node are in the 5 minute range? That limits you to a hard max of 12 deliveries per cycle with the quoted 1-hour cycle time. At that N, even a shitty algorithm with off-the-shelf heuristics would suffice.
Furthermore, logistics is far more complicated than just transportation; It heavily depends on sortation. If I have someone shopping for 1 person, there is no sortation problem. As soon as I tell that shopper to pick for 2 orders, I introduce a sortation problem, as those two orders need to be kept separate. And the sortation problem becomes a distribution problem: Do I have the shopper pick a list of items and then later sort them into orders? Do I have the shopper pick into different baskets on the same trip? Do I send the shopper on different trips for different orders? Distribution tradeoffs absolutely matter...distribution affects the level at which an atomic sortation action takes place. With wildly different costs for different types of sortation actions, ignoring this potential for optimization could be deadly.
They don't seem to have learned much about economics either. When considering the costs of making a delivery, you can be assured that the variable cost components is weight/volume/highway-distance, and the fixed costs represent the cost of the last mile. Using this simple rule, you can extrapolate (or interprolate) the costs of a given amount of market share. To appoximate the fixed costs of delivery, get the cost of sending a tiny item to your neighbor. For example, sending a postcard via US Mail (with US Mail representing an approximate 99.9% market share and delivery density) costs you $0.34. To do the same with UPS, you are typically running into the $2.00 range. UPS has a much lower market share (delivering to maybe 5% of all addresses in a given delivery day). And UPS has $50B in revenue! To make matters worse, these companies (at least instacart) are promising one-hour deliveries. This means that the delivery vehicle physically cannot contain all of the items of two orders where one is placed at 0600 and an order placed at 0700. The implications of this are huge. If UPS switched to this model, sending out a new truck every hour, you can rest assured that the $1-2 last mile cost would balloon to the $20 range, because you have now reduced their delivery density proportional to the number of trucks you have to send out. The costs of the last mile of delivery are so dominant that they are the reason why fixed-price postage works as a pricing method! If you need 100% market share to reduce your last mile delivery costs to $0.34, and 5% market share to reduce it to $2.00, where does that leave Instacart? They are away from UPS size by several orders of magnitude, and they promise hourly deliveries on top of it! They have the right philosophy (growth matters), but that is a bit like saying that if I keep building my sand castle, I can eventually reach the moon.
In my opinion, Instacart is DOA. Hopefully, the next time this trend pops up, they'll hire some experienced OR professionals from the very beginning.