Hacker News Reader

Comment by cs702 | Hacker News Reader

apoorvamehta: thank you for your prompt, articulate response.

The logistical challenges look daunting to me. My recollection is that the problem of coordinating and optimizing deliveries in such a point-to-point system with time constraints is NP-hard (I could be wrong about this, but that's my recollection). Then one has to deal with all sorts of real-world problems like order-fulfillment errors and substitutions (intentional and otherwise), constant shrinkage of merchandise, etc.

Obviously there's some price at which the service can be profitable, but I'm not sure it's $4. (Consider that the former CTO of Kozmo, Chris Siragusa, has been running a one-hour, point-to-point $3 delivery service for a number of years, MaxDelivery.com, but he has kept the service restricted to a relatively small, dense area in lower Manhattan, making the economics worthwhile.)

In any case, I sure hope you're right!

View on HN

cs702, your recollection is correct. Last mile logistics is a hard problem. Not to mention, order-fulfillment errors and substitutions do happen. However, we believe these are solvable. It comes down to two things - one is proper training of the drivers and using technology as much as possible to eliminate chances of errors.

That is too high level. Let me explain more clearly for one specific case of order substitution. Since we are focusing on a niche (groceries), our system has already calculated the substitutes of items. We know how to substitute a Store Brand Ketchup to Heinz IF needed.

A lot of what I know about logistics is from my time at Amazon Supply Chain, where I dealt specifically with the challenge of fulfilling packages to the customer from AMZN FCs (aka warehouses). And, at Instacart we believe we can have the same efficiencies IF we model the stores in a city just like AMZN modeled warehouses across the world.

IF you can somehow get a network of local retailers in each city, such that the stores together behave like and otherwise have similar operating and economic characteristics as regional warehouses, then, yes, I agree, it's possible. That's a big IF, though.

I have to take my hat off to you guys for attempting this!

if you can coordinate your own employee, or an employee at the supermarket to pull the order and have it ready just-in-time style to the curb, and then have the courier, again, be it your employee, a taxi without a customer, or another shopper going home to the same neighborhood you have gold. I would focus on the coordination of activities people are mostly already doing and by putting it all together you, and this specialized order fulfillment network/consortium make a few bucks.

I wish you luck.. that is a super cool business you get to design.

> And, at Instacart we believe we can have the same efficiencies IF we model the stores in a city just like AMZN modeled warehouses across the world.

My supply chain experience comes almost entirely from third-party shipping (somewhat) similar to this. In my opinion it's considerably harder and more problematic than dealing with your own fulfillment centers. You give up a lot of control.

There's a whole slew of problems: * Integration with other people's systems, which are often horribly dysfunctional, require manual intervention (such as re-keying), and can't give you the data you need (e.g. stock levels) * The fact that you have to rely on staff that don't work for you, and either don't care about you, or in some cases deliberately sabotage your orders * You're often relying on a vendor who is also a direct-competitor in this or other channels

I think a good model is Seamless. They act as the middleman just as you do, however it's the restaurant's name you see, rate, and attach reputation to. This gives them the incentive to provide good service. If you're shielding your vendors from the effects of providing bad service, then there's less incentive to provide good service.

I think it's an interesting problem to solve. Though here in NYC we're spoiled by a large range of next-day options; I personally use services such as Fresh Direct, Amazon (Prime and "Shop 'n Save"), Soap, and USQ Wines.

Interesting. Do you plan to replace the weekly trip to Wal-Mart in terms of range, or will you limit yourself to the smaller product selection you'd get in a local corner store?

What will you do if all your drivers are busy? Will I be able to book an order for an hour-long slot in the future, or is delivery now or never?

Sure it's NP-hard for an optimal solution, but the n here might be fairly small (how many orders in a 1 hour window?) or they can just use heuristics to do very well, if suboptimally.

[...] how many orders in a 1 hour window?

Depends on the city. In large metropolitan areas like NY, Chicago, the VA-DC-MD corridor, etc., potentially tens or hundreds of thousands at certain times of the day (e.g., 7pm, when people get home from work).

I see a very tight profit margin, but the key to pulling this off is good supply chain optimization.

Good luck

Reply on news.ycombinator.com