Self-Driving Grocery Car(t)s? Online Grocery Tech: $1.5B Raised, > Webvan’s $800M. Or Drive through Stores?
As the ex-technology guy at Webvan, I was tickled to read Zoe Leavitt’s (CB Insights) blog this weekend, noting that there are over 65 startups in the “grocery tech” space that have recently raised over $1.5B: almost twice the $800M raised by Webvan during the dot com boom. Not all these startups are in online delivery; in fact most are focused on in-store tech. Why?
I get pitched an on-demand grocery/food delivery idea about once a month, which is ironic: although Webvan, a pioneer in the online grocery delivery space had a product/service that worked at scale, the business plan didn’t work out.
Anyway, why so much interest in this space after 15 years? Well, it’s over a $1T vertical, and the ultimate recurring revenue business model: $100+/week or $5000+ ARPU. A cable company or wireless provider might get a tenth of that kind of ARPU.
There are models that work at sub-scale, but to make a model work at super scale, i.e. scale a $10B revenue company exponentially, is just hard. Here is my view on why, for three classes of business models.
Notes: Scale is calculated as follows: $150 per order*5000 orders/day (our goal at Webvan for a major metro)*365 days*30 major metros in the U.S. and you are at about $8B+ per year.
SKU count of 10K+ is important, because if a person can get milk and bread and diapers, but not corn starch or a pacifier in your “store”, they have to go to the grocery store anyway, and then you lose them.
Gross Margin can be as low as 20% for some items and as high as 40% for some, and depends heavily on volume discounts, so I used 30% assuming reasonably high volume. You subtract all costs from Gross Margin to get to the Profitability percentage.
I did not include Marketing/CAC in the above, that will be dealt with separately in another post.
1. Centralized Automated Warehouse models are complex to build, with high capital expenditures (Webvan would spend about $40M to go-to-market in a major metro). Ours was the size of 18 Safeways, with 5 miles of conveyor, multiple rotating carousel pods, and millions of lines of code just to make the routing and “tote filling” algorithms inside the warehouse work. The model definitely has a chance to scale profitably, being proven by Ocado in the UK (> $1B revenues), and FreshDirect in the East Coast, and I suspect Amazon Fresh (here’s a shout out to Doug Herrington, super Webvan alum from our team, now running Amazon Fresh).
2. In the In-Store pick up model, since you are not working directly with the supply chain, the business is paying close to retail prices and thus the gross margin is low. You have to “charge” the customer: either through selective price inflation on certain SKUs, or through a delivery fee to make a profit. Both approaches have friction, and thus scaling is challenging.
One exception. Tesco makes the in-store model thrive in the UK, for two reasons: If the store owner owns the home delivery service, as in Tesco, there are synergies in customer knowledge and merchandizing that enable “incremental profitable economics” for the parent company. Second, London has incredible delivery density. American suburbs don’t.
Only a few cities in the U.S. can offer the delivery density and population scale of London: Like New York.
3. Regarding the third column, I’ve seen a number of pitches lately, especially around organics, or farm to table, or specialty diaspora items, where there is no store and no warehouse. The supplier drops off the goods directly at a small cross dock (the “almost last mile”, and the driver consolidated all the different supplier SKUs and takes it from there to the home.
Again, this is hard to scale because there are competing three factors: immediacy, SKU count and delivery density. What suffers in this model is the combination of immediacy with SKU count. Why?
To get ALL the SKUs needed for the order, you need to wait for all the individual suppliers to deliver to the cross dock. Hence the delivery windows in these models are about 2 days out. But because they are often specialty plays, the gross margin is high. So they can do well as niche plays, if you want to wait for 2 days, pay higher prices, it works.
Last-mile Delivery in customer selected windows is the nut to crack, across the board. Self-Driving Grocery Car(t)s anyone?
No matter which way you slice it, delivery density has a huge economic role in all the models. This is why robot delivery mechanisms, self-driving car(t)s, or a fundamentally different crowdsourced model of delivery are going to be critical for scale. In fact, home grocery delivery may well be the first killer app for self-driving grocery cars, well before consumer transportation.
Alternatively the new “store tech” ideas: AR, VR sensors etc., enabling better in-store pick up, or Amazon Grab and Go stores, might take over, maybe with a “drive through” experience: darn it, make the user eliminate the delivery costs of the last mile. After all McDonalds did it :)
Send me your comments and thoughts firstname.lastname@example.org, or right here in the post’s comments sections. Anyone want to try to build a $10B company, I’m all ears, as long as you have a serious idea that addresses the challenges above!