Interview Transcript

Disclaimer: This interview is for informational purposes only and should not be relied upon as a basis for investment decisions. In Practise is an independent publisher and all opinions expressed by guests are solely their own opinions and do not reflect the opinion of In Practise.

Can you firstly just describe, in terms of the breakdown in rough square footage, how you'd look at Amazon's network between the different building types; fulfilment centers, sorting. What type of building should we understand that is in the network?

For starters, the way I look at it in the most simplistic terms is, there's really two types of buildings out there in the world of Amazon, as far as the fulfilment operations go. There's the actual fulfilment centers, which is where inventory is stocked and orders are picked and packed and put into their shipping carton. I call those the 'cash registers' of the business. That's where the money is made. In the world of Amazon, it's the equivalent of a retail store where you would go to the cash register to buy something; it's just a warehouse instead. On top of that, the other side of the company is an entire logistics business that rivals the likes of UPS. In that side of the business, you have sortation centers, and you have delivery stations that move the finished packages to the customer's doorstep.

Even before the fulfilment centers, there’s a whole network of facilities that are designed to hold inventory and to replenish the fulfilment centers, and those are known as inbound receiving centers. There's quite a few of those buildings globally, that just bring goods in from foreign suppliers, store those goods ahead of the need at the fulfilment center, and then when the fulfilment center needs the inventory, they send the demand signal back to the inbound receiving facilities. Those facilities send truckloads down the highway to replenish the fulfilment centers. As I mentioned, Walmart and Target and other large retailers have the same notion where they set up these large holding tanks, typically near the ports, that are inbound distribution centers that just store inventory to feed the actual distribution centers that, in turn, feed the stores. It's a multi-touch supply chain, but it enables much more efficient purchasing and reduced inventory assets across the business.

Let's walk through the system. Let’s say I'm a third-party merchant; I'm using FBA or I'm using Amazon. How does that work step by step?

It very much depends on the origin of the merchandise. For vendors that are positioned offshore, whether it's 1P or 3P, typically those goods are moved over the water, unless they are items that are high value, short shelf life. When I mean high value, short shelf life, think of a laptop or something that is technology oriented and, in six months’ time, there'll be a new version coming out. Those items are moved by air freight.

By UPS or FedEx?

For all the air freight between say China and North America, I'm not intimate with that information. But there would be some form of air freight used for any type of products that have short shelf life and high dollar value. You don't want a laptop sitting on the ocean for days.

I think 50% of third-party sellers are Chinese sellers. If you're in Europe or the US, do you use UPS and FedEx or just a third-party carrier or Amazon for the inbound leg?

For the inbound, the vast majority of merchandise is not moving by air; it's actually moving by ocean freight. Over time, Amazon has built out their own infrastructure and capabilities for moving ocean freight, using their own resources so that they're not reliant on other third-party carriers. They've actually been building their own containers and manufacturing their own ocean containers, and they've set up a lot of technology in and around sourcing and moving ocean freight from China, not only to the US, but to Europe as well. Anything coming from overseas, essentially, hits the ports, and then, by ground, moves to one of these inbound cross-dock distribution facilities. In Amazon vernacular, they might call them 'IXD cross-dock buildings', but they're basically holding tanks. The inventory would reside there so that they can purchase and move merchandise in larger quantities per container than if they were to source everything directly to all of their different fulfilment centers.

For example, you might have an IXD building, sitting at the Port of Los Angeles, that services the entire West Coast fulfilment centers. All of the purchases for those fulfilment centers along the West Coast would be sourced into that building, and then it, in turn, would send truckloads up the highway to the various fulfilment centers along the West Coast when they need the merchandise. This way, you're not you're not bringing Christmas wrapping paper into the fulfilment center in June, flooding the fulfilment center and wasting that space. You're holding that Christmas wrapping paper in the IXD building from June to November, when you actually need and then you replenish the fulfilment center. That's a tactic used by all the big retailers.

For domestic suppliers, it's a little bit different and it's a real mixed bag. If you're a 3P vendor and you're domestic, Amazon will tell you where to ship the merchandise to and it's a function of which fulfilment centers will carry that product. Winter wiper blades, for example, would not be shipped to a fulfilment center in Florida; that would go to Michigan instead. It's really a function of where the demand points are in the supply chain.

On this inbound part, I think Amaozn has 20 million square feet of inbound facilities.

I'd have to look up the numbers. FedEx and UPS are our carriers, so they're not retailers, so this concept of storing inventory to feed their fulfilment centers wouldn't apply. Walmart definitely has an inbound distribution center network. In some places like Houston, you're close to four million square feet of just import distribution center space. These are massive holding tanks. They're not particularly automated or exciting buildings to go into. They're just huge holding tanks; they act as a buffer between China and America and prevent the flooding of inventory.

How much advantage Amazon gets over others in this part of the process over competitors?

You have to have scale in order to do this, because if you're a smaller retailer, it doesn't make sense to have yet another node within the supply chain. This concept really only applies to the top five or six retailers out there, because it does add another touch point. You have to ship it from the port to the IXD building, then from the IXD building to the fulfilment center. There's an extra leg of transportation and an extra warehouse that you're introducing to the supply chain. But yes, it does create a tremendous competitive advantage because, if you didn't have this in place, you would have far more inventory that would have to be held throughout the network, because you need to purchase and transfer goods from China to North America in full container load quantities in order to get the best possible freight rates. You would not be able to achieve that if you were purchasing container loads into all of the different fulfilment centers or, in the case of Walmart, regional distribution centers, so it creates far more efficient purchasing practices and reduced inventory levels across the entire enterprise.

How does inventory get from the IXD inbound center to the FC?

The FC knows what it sells every day because every time there's a unit that's been picked and packed and shipped, they aggregate that demand quantity at the end of the day and they send the demand signal back to the IXD building to say, I shipped one unit of item x. I need you to replenish that to me. The IXD building picks the orders – similar to a warehouse transfer for any other business – and those orders go out in totes. You'll see these yellow totes that they pick out of those IXD buildings. They load the totes onto trailers; 53-foot trailer goes down the highway. It's another form of a middle mile, and it won’t only be done by truckload carriers; could be independents. Amazon has developed a trucking app that allows them to get spot rates from carriers looking to move the freight, but they also have relationships with carriers. They move that freight and they're increasingly trying to figure out how to get that freight move to be a two-way move, not just a one-way move.

In other words, not only get the carrier to take the container from LA up to Sacramento, but instead of having them come back empty, arrange for that carrier to haul freight for somebody else, almost as a third-party carrier. They're increasingly trying to look for ways to drive more revenue and efficiency through the system to help subsidize their own operating expenses. But that middle mile between the IXD building and the fulfilment center is typically FTL.

Nearly 100% outsourced. Why don't they do it themselves?

Trucking is a notoriously unionized business, and if you were to set up all these trucking assets and people to drive the middle mile portion of the activity, what would be the benefit versus the way they do it today, which is simply outsourcing suppliers to do it? They'd find themselves in a situation where they would probably end up with a huge unionized trucking labor force, which they really don't want.

Could they not roll it out with the similar Flex program with have contracted workers instead of unionized?

To a certain extent, that's what they have done. This transportation app that they've developed is an app that allows a driver to pick and choose which lanes he wants to work, and what rate he will be paid and so on.

{audio:0:12:23) Can you just describe a typical fulfilment center, small versus large parcels. How are they different?

They're very different. The fulfilment center that most people associate with Amazon is what we call the 'Amazon robotics, small sortable fulfilment center'. Products in those buildings are, typically, 18 inches or less, and 50 pounds or less. That's the general guideline. The idea is that the product can fit inside one of those standard totes that you see riding down the conveyors. In the early days, when Amazon got started, they would send people out to the floor, and the buildings were typically two or three stories high. Maybe the building was 40 to 60 feet high in the early iterations. Basically, they went out with a shopping cart and they batch picked orders. Once the orders were picked, the totes were put onto conveyors, they descended down to ground level where they were packed, and then labelled and shipped. Oftentimes a 12 to 13 mile travel day for anybody that was doing order picking and it was very slow. Maybe a person might be able to pick 100 lines an hour if they were working at top speed. Along the way, they bought that Kiva Systems company, back around 2013, and they ended up taking that technology and rolling it out across their entire global network of facilities.

We still have legacy sortable facilities out there. In certain markets, labor costs are so low that it doesn't make sense to upgrade to the automation. But throughout the North American context and European context, most of the small sortable facilities are now automated with this goods-to-person Kiva technology, which is now called Amazon Robotics. In that situation, the robots carry the pod to the selector, the selector stays at our workstation. Typically, they'll batch pick six orders at a time. They'll select the item from the incoming pod and put it into one of the six totes that they're working on then push off the tote.

If the order is being picked from multiple levels in the building, the tote has to travel to an intermediate station where the item is removed from the tote and it's put into a cubby hole. It gets married up with a tote being picked at another workstation to form a consolidation process. If it's a multi-line order that's coming from different areas of the building, there's a marriage process that has to happen there to get the order together. But if it's a single line order, the tote can go directly to packing. They have a separate packing process for singles than they do for multis.

Ultimately, everything ends up in a shipping carton and heads down to the SLAM process, where they put the label on the package and then off it goes to shipping. As far as shipping goes, there's generally two flow paths. One is – and I simplify when I say this – UPS is doing the shipment, they pick up at the building, so the package would end up flowing down a conveyor directly to the UPS trailer. Otherwise, if it's going to a sortation center, run by Amazon, then the packages will flow for that sortation center. Amazon operates a very large number of sortation centers around the country so, in fact, there's multiple sortation centers that the package might be destined to. Some of them are involving air freight; most of the time, they're shipping product by ground.

At the fulfilment center, what are the decision trees? UPS can pick it up and ship it to a customer. What percentage goes via UPS? What percentage goes through Amazon's own network?

80% of the packages, we think, are flowing into the sortation centers, which is the first port of call for a package that's being handled by either Amazon or the USPS. There's a middle mile that's typically involved here where a full truckload is moved between the fulfilment center and the sortation center at night. All the packages are unloaded to a massive conveyor type system. Think of it as a racetrack that's at the sortation center. Nowadays, there's automation here being applied through Amazon Robotics, but the idea is that the packages are sorted out by zip code, and palletized. Packages that are destined to rural zip codes, that Amazon doesn't deliver to, end up being transported from the sortation center to the USPS post office and each post office is responsible for a bank of zip codes.

They'll drop those pallets off at the back door of the post office, and then the post office will make their morning run and go and deliver those packages, based on their operating schedule. We think that's about 22% of the package volume, roughly; about 2.2 billion packages in 2022. If the customer is living in an urban or suburban market that is serviced by Amazon, through their delivery station network, the sortation center will aggregate the packages by delivery station and then send truckloads down the highway to the delivery station. They unload the packages there and now there's a secondary sort process that takes place whereby the packages are organized by outbound route, whereby a route is the equivalent of a van. On average, we think there's about 170 packages per van that go out the door, which isn't 170 deliveries. It's actually less than that because there's usually more packages per delivery; it's around 1.4 packages per delivery.

Ultimately, the vans show up in the morning. They're parked in a parking lot at the delivery station site. The drivers show up at the to-be-loaded at a scheduled time of the morning, from seven till nine. They arrive in platoons. They get loaded over a 20-minute process. The vans depart the site and go and make their deliveries. The idea is, on a 10-hour day, you don't want delivery drivers spending much time getting to the area that's being serviced. I don't want the delivery driver driving two and a half hours before he or she makes their first delivery. I want them to be making deliveries as soon as they leave the site so you set up these delivery stations to serve us about a 45 to 60 minute drive radius so they have a much smaller area of coverage than the sortation center. Sortation centers, we think, cover an area of about 200 miles; delivery stations more like 40 to 50 miles.

I want to get into the last mile and delivery stations in a moment. But the product goes to the import inbound facility, which it seems to be as if there is not much technology in there; it's kind of a dumping ground for inventory. Then it gets trucked to the FC and then, typically, to the regional sortation center. What's the difference in the amount of technology for picking/sorting in the FC versus sortation centers? What percentage of total picking is automated in small FC versus traditional sortation center at Amazon?

We haven't talked about the large non-sortable FC. There's 109 of those out there across the country, some of which are operated by third parties. But those buildings are generally very low in terms of automation. The small, sortable FCs are very high in automation; the vast majority of picking in a small sortable fulfilment center is done with the use of good-to-person automation.

Why is it so low, for the large non-sortable?

When you look at the style of operation and the type of merchandise that carry in those large non-sortable buildings, they don't lend themselves well to automation. Think of a treadmill. If you've ever been to a place that distributes treadmills – they're large, heavy, bulky boxes – or furniture that you might see at an Office Depot type store, or television sets or things that require two people to pick up and handle just because of the awkwardness of the shape and size.

You mentioned 10 billion packages a year, for example, US. What percentage is large versus small, roughly?

The vast majority of the package vine would be small. I say that because a small sortable fulfilment center, that's a decent size, might be doing 300,000 to 500,000 units a day outbound, as compared to a high volume, non-sortable facility, which might be doing 30,000 to 40,000 packages. There's a scale of 1 to 10 there. There was a number that sticks in my head that says 85% of what they ship is under five pounds, so that kind of gives you a sense of it.

What's the difference in the outbound units, per day, for FedEx and UPS from their FCs?

If you're talking about annual package volumes, the way I think of it is Amazon, we think, in 2022, delivered about 5.9 billion packages for the year, through their own Amazon logistics network. When you look at the numbers for UPS for 2022, they're actually below that; they're around 5.5. Then FedEx is below UPS for domestic small package volume in the United States; it's important to clarify that comment. If those numbers are true – UPS and FedEx do publish their numbers, Amazon does not – and the 5.9 billion number is anywhere close to reality and that's what our estimate is, then they are now the largest small parcel carrier in the US, by volume.

But that's also an average for the year. The average doesn't mean much because the peak times are probably multiples larger over Christmas?

That's why when you talk about daily numbers, you have to be a little bit careful because a peak day is 2x an average day in the world of Amazon, or more.

How do you compare the efficiency in parcels per day or cost per unit at the fulfilment center or the sorting part of the supply chain in Amazon, versus the legacy players, FedEx and UPS?

You have to understand a couple of things when you're trying to compare these numbers because people tend to oversimplify the story. When Amazon ships a package to UPS or when UPS ships a package on behalf of Amazon, firstly, they're picking it up at the fulfilment center, then they are bringing it to their own sortation center. Then they are moving that package a long distance by ground, or they're moving at by air to a market area that is typically far away from the fulfilment center, because it's an area that Amazon doesn't service themselves yet. For example, if a customer lives in Montana, and they're a Prime customer, and they deserve two-day service level, and the item they're ordering happens to be coming out of the Phoenix fulfilment center, UPS picks up that package at the Phoenix fulfilment center, injects it into their own network, moves that package within the two-day time requirement up to Montana, through their own sortation center network, and then does the last mile delivery.

So UPS does everything from a transportation perspective, and therefore the cost per package is going to be $8 to $9, as compared to a package that the post office might move, which might be, say, $3 or less. Why would the post office be so much less? Amazon is taking responsibility for all of the sortation process and getting that package directly to the post office back door. All the heavy lifting has been done by the time the package arrives at the post office, and the post office's job is really just that last mile delivery component. Maybe Amazon is paying them somewhere around $2.75, $3, or something in that range, for that part. Then when you look at the cost structure of Amazon doing that same work through the Amazon logistics umbrella, they have to not only move it to the sortation center, sort it, touch it, move it again to a delivery station, sort it, touch it and also do the last mile delivery.

There's multiple touch points in there that are involved. None of those three flow paths are necessarily comparable. UPS does all of the long-distance freight; much of it is air because of the urgency of the Prime delivery. The post office does all of the rural freight where there's very low density of deliveries. Let's say you were to send a UPS driver out to go and do rural area for a 10-hour day, he might be able to do 60, 70 packages a day. Whereas in an urban market, he might be able to do 225 in the same day. The cost per package is different when the post office is doing the work because their mandate is primarily to service the low-density areas that Amazon can't cost effectively reach. On that note, Amazon has opened up somewhere in the neighborhood of 40 delivery stations that are rural facilities, operating in very small markets. We think that they're making a play to start nibbling away at that volume that they're giving to the post office.

One of the difficulties, when you get into this whole world of package shipping is, Amazon wants more than anything to tighten up the speed of service; they want to go faster. They're obsessed with speed. When you have to accommodate different operating schedules that involve the weekend – USPS ships on their schedule, not on Amazon's schedule – you end up with more delay being built up into the delivery cycle when you have to dance to all these different drum beats. It's much simpler to have one flow path where everything moves through that full path and you're in charge of the schedule.

I spoke to a former FedEx guy. He spent 35 years at FedEx, built out the Express business and he was saying what you're just saying, which is almost counterintuitive. Where FedEx has so many different products overnight delivery, priority overnight and various others, it makes it difficult to optimize the network. It’s almost easier to have just next day, and it can actually be cheaper and quicker to do only one next day or two day or same day, rather than try and push all these different parcels that have different delivery times through the same process, which seems really counterintuitive. How do you think about Amazon's obsession with speed?

It can be a neutral issue, cost-wise, in many instances, but it can also add cost. Generally speaking, the faster you go, the more expensive it gets. I'll give you a very simple example to illustrate that point. The normal flow path, where you take a package from a fulfilment center, over to a sortation center, over to a delivery station, that's done in large quantities. Let's say that delivery station is doing 30,000 or 40,000 packages a day through that process. All of the sorting at the delivery station is done overnight, so that by 7:00AM, when the first vans pull up, they can be loaded out. That night crew goes home at seven, and the building basically goes quiet from 9:00AM until when the vans come home again. Now in order to get next day delivery, what's happening now is they've inserted a process, where you go from sortation center over to the delivery station, midday, around 2:00PM, and a bunch of packages now show up that would have otherwise been delivered tomorrow.

Had you not been obsessed with speed, they would have been one day later. But because you're obsessed with speed, now you're running another truck over to the delivery station, midday. All of your vans are out on the road. What do you do? You call up these Amazon Flex drivers, which are the John Q Publics with their own cars; they get paid for a four-hour block, whether they do 10 packages or 100 packages. They show up and they pick up all of these midday packages, in order to get that extra speed. That's expensive, because you're running an extra truck over the road to get to the delivery station. You're involving a more expensive form of transportation in the last mile with these Flex drivers, in my opinion, rather than just pushing everything through the normal flow. That's one example that's easy to understand. But the other is the use of air freight.

With the first example you just mentioned, you're putting more parcels through the same fixed asset.

Now, the totally different subject – and I mentioned the air freight – for the cost of moving a package by air versus ground, the multiplier there, in my mind, has always been about 7x. So if it costs you $1 per pound, ground, and it's $7 per pound by air, just as a rule of thumb. Amazon has over 50 airports that they are moving packages to and from, in order to get that speed that they're looking for, for the Amazon Prime member. What's happened is, there's a large volume of packages moving by planes throughout the country through these airport hubs, regional hubs that they've set up, and not always are the planes full. It's not an 100% efficient operation where you've got a full truckload. It's not a full plane load. You're shipping partial planes, in many instances, because not every market can be optimized and you've got this excess capacity now that you're trying to sell to the post office and whoever else and you have to try to make them leverage that capacity and drive revenue through it and you can't always do that.

What percentage goes through air from the FC?

We're not sure, that's the problem. We think it's about 6% of the package volumes, somewhere in the neighborhood of 400 to 600 million packages, but that's a guess at best. There's one other aspect of the supply chain that might be of interest which is called the 'sub same day fulfilment center', which is relatively new; it's come out in the last couple of years. It's a fulfilment center and a delivery station combined. With the process I talked to you about where you pick an order at the fulfilment center, you move it to sortation center, then you move it to the local delivery station; here, you combine everything under one building, and it's a smaller fulfilment center, and it's a small delivery station, together under one roof. When you place your order, the idea is that you receive your shipment within five hours. That's a hyper-fast service level. The intent is to compete against the Instacarts of the world. It's also a direct shot at Walmart because it's the equivalent of a Walmart store. The concept is top 100,000 items are stocked in these buildings; instead of you having to get up off your couch and go to the Walmart store, they'll pick your order, they'll ship it to your house in five hours.

The delivery, in this case, is made by a Flex driver, so it's a more expensive delivery, it's a non-optimized delivery. You're not going out in the van with 200 packages; you're going out with a driver that might have 30 to 40 packages. But the fact is, there's no middle mile and all that touchpoint that goes on between the fulfilment center and the delivery station gets eliminated. The limitation here is that they can only service these top 100,000 items, so it's not 300 million items, and they can only service customers within about a 45-minute drive radius of that building. They're about 170,000 square feet, on average, and they're putting these down in all of the NFL cities, what I call the 'big cities', over a million population. Right now, there's about 46 of them that they operate; we think they're going to head to about 150, before the dust is settled. It's the equivalent of a store, except you order from your home.

How would you compare a delivery station of FedEx, UPS, the legacy players and Amazon?

I'm not an expert at FedEx and UPS in terms of how they operate their delivery station operations. We have to remember, UPS is 113-year-old business and they've built up that business over the last century, and Amazon has built their business up since 2014. Today, they do more package volume domestically.

That was my question. For example, FedEx have 2,000 delivery stations and Amazon have 500, 600 and they do more volume. How is that possible?

It's a function of Amazon being B2C, and UPS being B2B. Also, Amazon skim the cream and they're really going after the big cities and the big markets, the high-density markets. Rather than trying to do every single zip code, they focused on skimming the cream and getting the big markets done. They move so many packages; it's just a function of the package volume. There's oftentimes one unit per package, and they're shipping billions of units every year.

How do they get it from 500 stations?

No, it's quite simply, when you when you're talking B2C as a flow path, as a market, the ecommerce channel is generating tremendous package volumes, huge package volumes compared to B2B. A UPS truck shows up at a manufacturing plant and drops off 10 boxes in one shot. In the B2C world, the driver shows up, drops off one package, so it's far more labor intensive and package intensive to get ecommerce delivered to people's doorsteps. Think about the sheer volume of units that Amazon is spinning off, we think, in the ballpark of 10 billion packages a year in the US in 2022, of which close to 60% of that is done by Amazon Logistics. The vast majority of the orders are being generated through the big cities, so the 600 or 700 delivery stations. They actually have two types of delivery stations. But those delivery stations are servicing that massive volume. It's just large volume divided by low number of delivery stations enables them to service urban and suburban markets efficiently themselves, versus hiring someone else to do that work.

What's the difference in the cost per unit of the last mile, if it goes through Amazon Logistics versus through UPS or USPS?

If we look at the UPS, cost per package, we think it's around $8.73 per package. UPS is flying a big portion of that freight in airplanes, and they're moving that package from the fulfilment center, all the way through their network of sortation centers and last mile delivery stations that are equivalent. They've got a lot more work to do than the post office where the package is delivered directly to the back door of the post office. In the world of Amazon Logistics, we think they're spending somewhere around $3.82 to move that package themselves. We think it's around $2.93 through the post office, and about $8.73 through UPS. But none of those numbers are easily comparable. Of that $3.82, we think that last mile costs about $1.75.

How much for their own last mile from the delivery station to the customer for Amazon Logistics? Take Shopify for example, if I'm a Shopify merchant, and I'm shipping via Shopify Fulfilment Network, which obviously uses UPS carriers to do the last mile, what is the cost of their last mile, if I'm a Shopify merchant, versus Amazon FBA last mile? You mentioned $1.75 for Amazon Logistics today, roughly.

Shopify, because they move large volumes of packages, will negotiate with their delivery partners, say UPS and FedEx. They're not going to get to Amazon levels. I don't know what the Shopify numbers are, that they have on average. It's really a function of the size of the package, etc.

They provided a price but it's pretty interesting to compare that with buy with Prime because it seems pretty competitive. I'm just curious of how much Shopify is really subsidizing this. For example, for a bicycle helmet, small parcel, $6.59 standard domestic delivery per item, which is what I believe the merchant gets if they ship, end to end, through SFM, which seems. pretty good.

Let's say that Shopify has negotiated a discount with UPS of 50%. They're making money on freight; they couldn't do it otherwise. They'll get their discount through UPS, based on their annual volume, and then they'll mark that up and sell it to the customer, for your $6.59. They might be paying $4.50 for that same package to be shipped, and they'll put money in their pocket, but the customer if they were to do it by themselves would be paying $8.50. So it's a win win.

If you were to break out the total cost of shipping the package from inbound, middle mile, sortation, and the last mile, what's the total cost per package, roughly, for Amazon Logistics?

You would have to say, what is the cost of the sortation? What's the cost to move the middle mile from the fulfilment center to the sortation center? What is the cost of operating the sortation center? What is the cost of the middle mile to go to the delivery station? What's the cost of the delivery station operation? What's the cost of the last mile? That would explain just the Amazon Logistics portion of it; it wouldn't explain the air freight; it wouldn't explain everything else. But like I say, we think that the average cost per package right now, moving through the Amazon Logistics network, in 2022, was about $3.82 and just the last mile component of that is around $1.75 and that can vary a lot too, just depending on the area being served. Picture a van going into Manhattan with 225 packages and making two stops at big apartment building lobbies. That's an efficient run. He can come back and do another run. Now picture the same van doing a 10-hour day in the suburbs of Des Moines, Iowa and he might have five to 10 minutes between stops and he might be able to do 130 packages, over a 10-hour day, so it really varies.

This is really low though, right? $3.50, that's really low, per unit, for Amazon.

$3.82 per package, and even then, you have multiple units per package. Again, these are numbers we're engineering behind the scenes and the truth is going to be different, so we're doing the best we can.

My question becomes, looking longer term for this and Amazon's strategy of Buy with Prime and trying to be the rails of ecommerce and third-party shipping. How do you look at any potential competitive advantage that Amazon may have with their network, versus the carriers or the alternative ways of shipping a product to customers?

What Amazon has that no one else has, is sheer coverage of geography. They have a massive number of facilities that no one else operates in a world of ecommerce. Just between the small non-sortables and the large non-sortables, they have over 200 warehouses located around the country. They've got the bases covered.

Shopify could say, I've got 10 3PLs that have the same network.

It's a whole different ballgame. Really, what Shopify need is six to seven fulfilment centers, in order to get two-day delivery capability to 90% of the population, in the US. You'd be somewhere around New Jersey, Pennsylvania, somewhere in the Southeast, like Atlanta, in Dallas, in Chicago, and LA, and then up in the Northwest, in Seattle, and then possibly Denver. If you had seven buildings operating, you could cover the most of the country in two days, but you're reliant 100% on a third party, and the third-party carriers are going to constantly be raising their rates. We did an interesting little piece of work that looked at the rate of inflation in the small parcel business over a period of 20 years.

If you go to the list rates that they publish, which you can download for free, you can actually see a consistent 5% inflation rate, in small package delivery costs, for just the base rates, never mind all of the add charges, such as the fuel surcharges and everything else. You're at the mercy of a duopoly and there's nothing you can do about it. Whereas at least with what Amazon has done, they've got total control over the portion of the supply chain that is run by Amazon Logistics, not only in terms of the operating schedule, but in terms of the capacity and cost controls. They've built that with a non-union labor force, where they can flex up and down as needed, which is just an incredible accomplishment, given the amount of time that they did it with. Now they hired a lot of FedEx and UPS people to make that happen back in 2014.

Again, let's say I'm a Shopify merchant, I run a B2C company. I have to ship my products to customers; I have my Shopify website. Now I can go and put my Buy with Prime badge on and I can use Amazon's back end or I can go and use SFM. How do you see that decision making between merchants?

That's a question a lot of people are trying to ask and answer right now, and I don't have the answer for it. A responsible answer is, I don't have an answer for it. I haven't done enough homework on Shopify, to say, they are going to be at least 10% to 20% more expensive than Amazon. They're using third party carriers, and they don't have the coverage of geography the way that Amazon does, so speed is going to be an issue. Amazon marketplaces and infrastructure is just night and day difference compared to Shopify, but the reason that you would want to go with Shopify isn't necessarily cost driven.

It's differentiation.

It's trust. If I'm a retailer, and I view Amazon as a competitor, do I necessarily want them to know which items that I sell are the real hot ones? Next thing you know, they're going to show up as an Amazon Basics item. Do I want to entrust Amazon with my data? Who are my customers and what are the products that they're buying from me? That becomes a very contentious issue for large retailers. If you're a small mom-and-pop retailer and you don't have any way of getting product to market nationally, or even a midsize retailer who might have one or two buildings for ecommerce – you're struggling in this area – it might be very attractive for you to leverage that Amazon Buy with Prime option. but I'm very skeptical that large retailers are going to take advantage of that and feed and support a competitor and trust them with their data.

Let's talk about the last couple of years with Covid. How have you seen Amazon pull back the square footage growth? What exactly are they doing to cut the space, sublease the space, what have you been seeing?

Prior to Covid, 2017 through 2019, globally, they were introducing somewhere in the ballpark of 35 to 50 million square foot per year, of which 25 to 30 million was in the US; the rest of it in the rest of the world. That was a decent pace for them; they were keeping on track with their capex spend relative to their sales and their returns. This huge burst of activity took place in 2020 and 2021 where they doubled their fulfilment center network in the space of two years. They doubled what they had previously built over 25 years. They were adding new space, globally, at a rate of about 125 million square feet to 137 million square feet in 2021. In two years, they added the equivalent of about 260 million square feet of space, globally. It's a hard number to wrap your mind around because I always say to people, Walmart took more than 50 years to build out their distribution network in the US and it totals 150 million square feet. They did that over five decades. That's just distribution, that's not retail. 85% of what's moving through retail stores is coming from distribution.

That's the FCs they have and the inbound facilities, you mentioned Walmart has?

Walmart's footprint, which includes ecommerce, distribution centers for the stores, is around 150 million.

How many DCs do Walmart have, roughly? Amazon have 250, 260 fulfilment centers. They've added the same amount of square footage in the US over the last few years as Walmart have in 50 years.

That's the way to think of it. Sam Walton built that business with the understanding that it would always generate a 5% net rate of return, net margin, and Amazon hasn't had to worry about that as much. In terms of distribution centers that are non-logistics in the world of Walmart – I'm going to just do the math here for you – I'm going to suggest Walmart has about 196 buildings.

DCs that are similar square footage?

Similar, but 150 million square feet is what they operate.

With no Kiva system.

They have other forms of automation. Historically, Walmart was very reticent to automate. Most of the buildings were conventional operations. That's changed dramatically in the last 10 to 15 years. They have been heavily, heavily investing in automation.

How irrational was the Amazon build out over the last few years?

There were two things going on. There was the build out of the fulfilment center network and then there was this one time build out of the logistics network that they were setting up, which is laying the railway tracks for tomorrow. The delivery stations, the sortation centers, the inbound cross-dock. One is geographically driven, which is the logistics side of the business and the other is driven by sales and product sales growth. The more product sales volume you sell, the more fulfilment centers you need just to handle the throughput capacity.

If you didn't want to build out the logistics, you could just build out their fulfilment centers as GMV grows, and then ship them out and then get UPS and the carriers to pick them up from fulfilment centers and they do the rest of it. On the last call, Jassy said, he's basically built FedEx in the last three years. He means the logistics side we just mentioned; sortation and delivery.

The fulfilment center strategy is really easy to understand. Going back to around 2008, when the tax laws changed, it used to be that you would have to charge sales tax as a function of where you shipped from, not where you ship to. Amazon, historically, was building all their distribution centers in what they call 'tax friendly' states, which would be places where they didn't have to charge a sales tax when they sold something. They were avoiding California, and they were avoiding Texas, and New York and these other states where they needed to be in to get close to the customer, because they had that competitive advantage relative to other retailers that had stores. They didn't have to charge sales tax.

They changed the law so that it became a function of where the ship to address is, not the ship from address. If you happen to live in a tax-free state, you don't have to pay sales tax when you order online, regardless of where it's coming from. Once that set of laws went through, Amazon started building out their fulfilment centers in all the right places, basically in the backyard of all the major metro markets. If you go down the list of markets by size of market, today, they're pretty much done with all the cities that are a million or more. There's a couple of exceptions like Buffalo, New York, but they're almost done with what I call the NFL cities.

How much more square footage to cover the whole of the US?

That's a question of how low of a market are they going to go to with these fulfilment centers? Are they going to put a $300 million building into a town of 50,000? Probably not. Are they going to stop at 250,000 or 300,000? Quite probably, yes. We're seeing fulfilment centers today that are being planned and built in markets as small as 300,000 to 400,000. We think the plan is to get a large, non-sortable and a small sortable fulfilment center in the backyard of every town of at least a half a million. Then from there, speed will dictate whether or not they go below that number, down to the to 200,000 to 300,000 range.

Up to half a million, how many years is that of expenditure?

We need to look at the plan that they have right now. There's still 28% growth being planned for the footprint that they operate right now. If you look at the fulfilment centers, there's another 42 million square feet that we know that is going up in the US.

Over five years, or three years, or one year?

I'd say until the end of 2024. We don't have much visibility beyond that, because we're dealing with public data.

How much have space have they cut then?

Yes, they're cutting and they're still cutting. It's important to understand what's going on there. I'll define what I mean by cutting because there's three flavors of activity taking place right now. There's closure of active buildings, which is one form of cutting. Then there is cancelling projects that were in the pipeline that they were planning to build.

FCs or delivery stations or both?

Both. That's capital preservation. When you close an active building, you're trying to reduce operating expenses, because you have duplicity; you have too much capacity. Let's say I have five delivery stations in and around Boston, but I really only need two to get the job done. I can close three and migrate their volume into the other two. In Boston, they've cut back five delivery stations, if I remember correctly, and they've had specific geographic areas where they just had too much capacity. During that Covid period of 2020, and 2021, the powers that were at work in Amazon went out and grabbed as much space as they could, in anticipation of this one-time bonanza. We have to think back to March of 2020. We were all hunkered down at home and the kangaroos were bouncing down the streets of downtown Sydney, Australia. Wildlife was taking over; we had no clue how long this was going to last.

They were going out and literally tying up stuff.

Anything that looked like a warehouse, basically. I don't know how the approval process worked but some people have said they spent like drunken sailors.

They must have, if you’re saying they had five delivery stations in an area where now they have two.

{audio:0:1:00:59} They just went too aggressive. Honestly, it's easy for me to be an armchair critic. They did the right thing at that time. They were really struggling to get the orders out on time and to get people to come work.

Where are we today then? How many do they need to cut?

In the US, there's been 31 closed buildings, primarily delivery stations, where they had this duplicity. There's been 33 projects that have been cancelled and there's been 35 buildings that have been built-to-suit for Amazon. The keys have been handed over by the developer, and Amazon's left the building dormant. They basically put a padlock on the front door, and they said, we're not going to open up the building and start it up because that means more payroll expense when we don't need it.

What can you do with these built-to-suit buildings?

With those buildings, they're going to leave them dormant for one to two years with the goal that, as growth and volume pick up, they'll eventually be able to turn those buildings on. That's excess capacity that they overbuilt. They have excess capacity within the existing active network. They want to fill that up first, before they activate these buildings and it all makes sense. They won't sublease these buildings because they're very specific to the needs of Amazon. The throw away money here is the rent, basically. You're paying a lease for nothing. But it's better than adding a $40 million payroll to building you don't really need to activate.

Just in the US, we've seen 99 buildings impacted by the austerity measures of 2022 and Q1 of 2023; total square footage is in the range of about 32 million square feet. When the first Q1 results came out in 2022, and showed major losses on the operating income, we did our homework and sat down and said, we think that they're going to cut back somewhere in the ballpark of 30 to 46 million square feet of space. That was our prediction a year ago. They are at 38 million right now, so we were pretty close. There's no doubt in my mind, they're not finished yet. They're just now starting to really take a hard look at the rest of the world in the UK, Germany, Italy, Spain.

That's mainly delivery stations and sortation centers?

No, there's fulfilment centers getting closed there too. It is both really, that's the truth.

How do you look at the utilization of the network in the US today versus three years ago, pre-Covid?

We measure capacity three ways and obviously, unless you live in the world of Amazon or have your favorite tips on all this data, it's very hard to really say what is the capacity within the existing network? We don't have that, so we have to invent other ways of trying to figure that out. The way that we have done this, is we've said, what is the sales per square foot that Amazon has historically achieved from their fulfilment center network, including the mezzanine space that they use as working space?

Do they include mezzanine?

They do not, no. But it's working space so you have to include it because, as time goes by, they started out at 40 feet high, and now they're up to 125. That's a lot more working space compressed into the ground square footage. They also don't include 3PL space that they use very, very heavily overseas.

Are you using revenue or GMV?

Revenue. Product sales, which is a combination of product and services sales. When you look at the sales dollar that is being driven through the square footage of the network over time, we think that the benchmark number should be around $900 per square foot, and it's actually around $940. Right now, as it stands in the US, they're sitting at around $772. That means they're about 23% to 30% excess capacity in the US, depending on how you slice it. That same number in the rest of the world is even worse; it's between 32% to 46%. Let's say that they grow sales in the US by a rate of about 8% to 9%. That means we're looking at about three years, before they fill this tank up and start to need to flip on more space.

My 10-K numbers, doing the per square foot stuff is obviously different because that you're saying Amazon don't include the mezzanine space and don't include the 3PL lease square footage? They give you lease square footage in your numbers through?

If the 3PL owns the lease, it's not showing up in Amazon's numbers; that's just a service agreement. It's like a supplier of any other sort. We've also done this excess capacity analysis a few other ways. We always seem to come back to the notion that they're at around 25% excess capacity right now in the US. Amazon calls that complete fiction, but they obviously don't want people looking underneath the hood in too much detail.

How can they grow into this?

Sales growth, or driving more revenue through the system. Either you saw more volume by driving more promotional events; you notice that there's more than one Amazon Prime Day, seems like Cyber Monday and Black Friday keep getting elongated. So you drive more promotions through the system, especially during non-peak time conditions like Amazon Prime Day. That's one possibility. But the other is Buy with Prime, to try to drive more revenue through the system. You get other retailers to bring their inventory into your warehouses, or they're now trying to auction space off to their 3P partners. They have this whole auction program set up where it's almost like Uber. If it's a low demand time period, the cost of the space will be lower than when it's a high demand time period.

How much of the population can Amazon serve on same-day and two-day today?

Same day that I talked about, where they're within five hours, our numbers are they are doing about 600 million packages a year through that channel, which represents about 2% of the overall volume. Now on top of that, you have to add what's coming through the normal fulfilment centers, sortation center, delivery station, and I don't think they're doing much of that by same day. Same day, I think, is still relatively small.

Do you think that’ sustainable, to push that same day strategy? Or is that kind of dying out now and we're going to move back to two-day?

Everything I'm seeing tells me they're continuing to push down hard on that speeding up of the service level. And by the way, if you really want to talk about how to lower operating expenses, that's the answer. You can take the speed out of it, you say, we're going to revert back to a standard of two-day, no more next day.

Why don't they do just next day and two days?

I don't know how much it would save. It would be a difficult question for me to answer. As you throttle the need for speed, you hit shipping expense much, much harder than what it used to be. If you were somebody who was purely an accountant, and you didn't care about all the customer obsession, you'd look at this business and you'd say, good heavens, get rid of the air freight, and none of this next-day stuff. Let's make it standard two-day and charge more for Amazon Prime.

How much is next-day coverage today? Let's say 2% is same-day, how much next day?

I think it's under 10%, but I don't know. Even two-day, we don't have a sense of the service level breakdown. They don't publish any of that.

With the square footage, and the coverage they have, how much more do they need to build out to get to this standard two-day all across the USA?

It's not so much how much more they have to build out. You have, let's say 350 million items that you're carrying, and you can't carry those in every building. What do you need to do to get all of those items available to the customers that are Amazon Prime in two days? What hoops do you have to jump through? How much product are you shipping at a really exorbitant cost to make that happen? With the network that they've got right now, they can get to the majority of the population in two days. It's just a function of, what's the price tag to do that? Are you going to try to get to Hawaii from New York in two days? You're going to be air freight? For an obscure item, maybe that's what has to be done. It's just how much are you willing to pay? If you allowed that to be moved in five days or three days, you'd have saved a bundle.

What do you think is the biggest risk for Amazon ecommerce, retail business coming out of Covid, over the next two to three years?

Very good question. I think there's a number of risks that we have to look at. The labor shortage is a big risk. In the world of distribution, we're all struggling to get labor to come to work and do the job. We're having to pay more for it, and it's far more turbulent, the turnover is high. I think labor shortage risks are a huge issue for Amazon. Going back to the beginning of this podcast, we talked about the Walmart parking lot having 800 cars, and Amazon having 3,000. It's just a far more labor-intensive process to do ecommerce than it is to have the customer shop at a store. You're doing the shopping for the customer, and you're delivering it. There's just far more people involved in the process of moving a single unit to the customer's doorstep. They've got to figure out ways to get the people out of the warehouses, and they're doing some of that now with robotics and things of that nature. But it's still a huge, huge number of people. In the US alone, they've got probably 700,000 people working in their warehouses and then on top of that, you have all the people doing the delivery.

How many employees can they cut at the warehouses with Sparrow and Kiva adoption?

Sparrow's a great invention. We saw that they can do the robot and it looks like it can do about 65% of the small sortable unit volume. When we do the math on that, we think there's probably savings in the area of about a billion dollars of labor globally, that can be taken out of their system, just through that one invention. It's definitely a step in the right direction, but there's still an awful lot of labor that's out there that is needed to get the job done. It's a very, very labor-intensive operation, at a time when the demographics of the country are changing dramatically. The country is getting older very quickly, and huge numbers of people are moving into retirement age, and there's just not enough bodies to backfill. We're moving towards the Europe's business model and in Europe, they've had 20% plus of the population of retirement age or older for decades already. We've always been at 14%, and by the end of this decade, we're going to be closer to 22% in the US, so there's going to be about a 5% to 7% drop in the labor pool over what we have right now. We're struggling right now to get the job done, quite frankly. Labor is going to represent the single greatest challenge for the business.

Any other risks that come to mind?

There's always government and there's a threat of unions. Staten Island unionized, and they want $30 an hour instead of $18.50. A pay increase of that magnitude would be crippling for Amazon and it's unrealistic to pay people those kinds of rate wages in a warehousing environment anyway, for that type of work. I think Amazon's fallback position here is that they have tremendous scalability. They've got non-union warehouses clustered in and around New York City. If ever Staten Island goes on strike, no problem, they deflect the orders over to Connecticut and Pennsylvania and New Jersey, and they get the work done from those neighboring facilities while Staten Island is idle. As long as they're a blend of union and non-union, I think threat of unions and strikes and all that is very mitigated. But if that grows in a big way, that could be another issue for them.

I guess that also becomes an issue for everyone else, as well. What do you think the biggest competitive risk to Amazon's ecommerce business in 10 years’ time?

Everybody's getting better at the game, as time goes by. Walmart really was behind the 8-Ball in the beginning, but they really picked up the pace on this, and they're getting good at ecommerce and efficient at the same time and automating and just doing a lot of the right things. Target is doing some pretty interesting things with the use of Metropolitan sortation centers and the acquisition of Shipped. We're seeing all kinds of interesting activity out there, that is continuing to increase competitive pressure in this space. Amazon, if you go on the marketplace today, there's things that really turn people off; it's just become a dumping ground for knockoff cheap merchandise. It's hard to find things sometimes, and they've got a real cleanup in front of them to just make sure that the vendors and the products are legitimate, the reviews that are posted are legitimate, that people can easily find what they need. There has been, I think, a degrading of the quality of the shopping experience at Amazon, because of the way they've set up their business model that a lot of people find irritating.

I think it's interesting times ahead; more competitive pressure, probably more government interference as well. I would suggest that, especially when you look overseas, where the rules are much stricter, the use of gig workers represents a big, big threat. Like all these Amazon Flex drivers, and what Instacart is doing with the hundreds of thousands of gig workers they have. How much longer do you think that's going to last before California and New York, actually start to legislate against that, because that creates tremendous inequity in terms of labor cost for companies that have brick and mortar versus the companies that exploit gig workers?

How much does Amazon use gig workers, roughly, Flex workers?

It's not so much that Amazon's using gig workers; it's the transportation delivery function. All of the people that are doing those deliveries by vans are employed by delivery service partners. A delivery station might have five or six delivery service partners that Amazon helps them get set up. How do they hire people? Are they running payroll? Or are they using gig workers? It's anybody's guess. I'm sure it's a mixed blend. All those Amazon Flex drivers that have their own cars, they're definitely gig workers. They really help to help cut off the peaks that take place, not only during this season, but during the hour of the day. You might have peak hours for the day where you can offload volume to those Flex drivers. It's a very clever technique to get the job done. Eventually, I think these are going to have to become regular salaried employees that have benefits, and there's going to be a big cost advantage that they lose when that happens. I don't think it's that far away. Government interference in the business and the legislation especially in Europe, where the rules are much tighter than what we have in the US. You can't just operate the way you do in the US with sort of the freestyle.

If you were running the retail network today with Jassy, what would be the big changes you would do to tidy up or to drive more profitability over the next few years? What would you like to see Jassy do?

First off, I want to say one thing, make one thing very clear. I think he's done a heck of a great job. I really do. He stepped in at the worst possible moment. Labor shortage, labor unions forming, sales are dropping, profits are in the red. He had to deal with it all at the same time. When you're talking about 100 buildings being taken out of your network, over 30 million square feet of space, painful decisions left, right, and center, layoffs and all the rest of it, and to do that in a very compressed period of time, in a year, this type of work is something that most companies would take a decade to try to figure out. He's been forced to do it in a very short period of time, in order to get the investor community back on track. Kudos to Jassy for what he's done so far. I think, first and foremost, you've got to get every business unit in Amazon to stand on its own two feet and be profitable. I believe that's his culture, and his psyche is to say, look, Amazon rest of world, if you can't pull your own weight, there's no more subsidies coming from AWS; you got to figure out how to make that happen.

Do you think he's going to do that really? Like India?

They've invested huge amounts of capital to develop new markets, to expand the oxygen tank, in effect. You are always looking for more sales growth. If you've put 10 billion into a new market, and invested all that capital, you don't pull the plug on it, because it's not making money. Unless years and years and years go by, and you realize, this thing's never going to make money; it's never going to be able to stand on its own two feet, because of the market itself. Not enough volume going through the facilities, geography is too big, sales are too low, that kind of thing. There are markets like that in the world of Amazon today, that I believe are forever going to be unprofitable. I think the right thing to do is to write off your losses and just walk away.

Is that India or do you know of any others?

I don't know the specifics of every market, but I'm going to suggest Mexico, Brazil, India probably some other markets out there. Even in Europe, Italy and Spain, I would question. You look at the average spend of an American on Amazon, and you're somewhere in that $880 range for every man, woman, child, grandparent and infant, and that same number in Mexico is like $13. How are you going to make money in a big country like Mexico, when you have such low volume, and you have to set up all this infrastructure and cost? You're anticipating there's going to be an explosion in the online channel in Mexico, so you have to invest today for something that might transpire 10 years from now. The reality is that that may never happen. You have got to look at some of these decision points and say, should we throw good money away?

Potentially, cut some of the rest of the world?

Look at the losses on the operating income statement. They're losing as much money in the rest of the world, if not more, than they are in North America. They've got to do something, otherwise, you're cutting too much into the bone in North America, which isn't good either, because you've got to save money somewhere. To my mind, it’s about making the very brave decision to walk away from a market. They did it in China back in 2019; they walked away. They couldn't make money there, and they pulled out. Alibaba was too much of a force. They're going to have to do the same thing in some of these other markets. I haven't seen that yet. I've seen them start to nibble away at some of the infrastructure they have in the markets where they are doing well, actually, UK and Germany.

But they haven't actually cut any? I guess it's harder to find out in India or stuff, what they've cut by square footage.

They've taken stuff out of India, but it's hard to quantify. They've walked away from food and they've taken out the wholesale side of distribution there and some other things like that. But now that's a big, big market with a lot of opportunity but it's the frontline of the holy war of ecommerce because you've got Walmart in there, you've got Alibaba in there and Amazon and they're all fighting for that online market. They don't want to pull out because that's the frontline of the war.

Anything else you would tell Jassy to do?

Jassy is coming from AWS and that's his fabric; that's his DNA. AWS has always been tremendously profitable for the company. Now you're getting into the retail side of the business where it's difficult to turn a profit and so the foremost thing is this obsession with speed is costing a huge, huge amount of money. I would question the sanity of having air freight, for any of this. Sending a $20 order onto a plane just makes no sense at all. You're losing your shirt. I think somebody has to really pull back on those reins and say, we're not going to try to be everything to everybody. If you're ordering a product from the other side of the country, even if you're an Amazon Prime, we're sorry. Either you pay a premium for that speed, or you wait. I think we can all wait; we used to wait forever anyway, so what's the big deal? No one else can do this in two days.

That could be a difference, if they start tweaking what parcels can be shipped Prime, and next day and so on?

They know, to the penny, what it's costing them to do business. They know whether they should take an item from one fulfilment center and move it to another fulfilment center and consolidate it before shipping in order to save money. They know all of the details of what's going on; they're data crazy. They're knowingly are losing money, when they do work at high speed. They should end that activity and change the business model.

Because there's same day, then there's next day. I think maybe they've pushed it too far.

Yes, they're pushing the envelope hard. Even two-day, what are you doing to make two-day happen? Are you blowing your brains out, because you want two-day to happen, and you're losing money? We have to remember a $10 to $20 transaction is where you might be spending $20 in logistics to get the product to the customer's doorstep, in that two-day window. You just can't do that in perpetuity; you've got to do something about the business model, and the rules around what it costs to get something fast. They have to be changed. Those would be the main things I would suggest would be, continue automating, really do a thorough review of which countries to pull out from a rest of world perspective, and then change the game on speed. He's been doing a great job, I have to say. I really like what he's done so far. We think that they've been cutting back operating expenses to the tune of about $5 billion a quarter, which is a tremendous achievement in a very short period of time.

What is the long run margin of this business? How profitable can the retail business get? We've seen in the past, in North America, it's been highly profitable. You've seen 4% or 5% EBIT margins.

I think North America has a strong potential to be sustainably profitable at that 4% to 5% rate because of the high amount of transaction volume, the high amount of sales per capita that they can achieve. People are obsessed with Amazon, customers love it, that convenience is great. When you go into the rest of the world, and you take a country like Italy, where it's down around $100 per capita, small geography. You're not moving goods thousands of miles. It's a very small area. Or UK, for example. Huge population, small geography, relatively low logistics costs to move goods.

UK is going to be profitable; Germany will be profitable. Getting those other markets to be profitable is predicated on getting sales volume to grow by 600% to 700%. It's very questionable whether you can make that happen because culture plays a role in all this as well. Not everybody in Italy wants to buy online. The age of the population also has an impact. Younger people buy online more than older people do. All these dynamics play a role; focus on unprofitable markets, like Japan, where you have a huge population on a small island. You have all these opportunities out there that you can focus on and just drop the ones that don't make money and just say, you know what? We're not going to get anything out of this. They've got plans to expand into places like South Africa, Nigeria. If it doesn't make sense to do that, then so be it. You're not going to be a global player, and that's fine.