Amazon sells more goods than any one person could count – but the e-commerce giant’s true “core product” is convenience, and how quickly it can get an order from customers’ virtual shopping carts to their real-life doorsteps.
Part of what makes it so easy for Amazon to offer two-day or even same-day shipping to customers is its vast network of distribution centers, which are located across the U.S. and store and ship products to their final destinations. New research from Wharton business economics and public policy professor Katja Seim takes a closer look at how significantly expanding that distribution center network over the past decade has been key to Amazon’s growth strategy.
Seim recently spoke to Knowledge@Wharton about her paper, “Economies of Density in E-Commerce: A Study of Amazon’s Fulfillment Center Network,” which was co-authored with Cornell’s Jean-Francois Houde and Penn State’s Peter Newberry.
An edited transcript of the conversation follows.
Knowledge@Wharton: Your research is about Amazon’s distribution network. What is a fulfillment center and why is it important?
Katja Seim: What we look at in the paper is how Amazon’s distribution has changed over time. In online retail markets, just like in offline retail markets, the company typically uses one or multiple distribution centers, which in Amazon’s case would be called fulfillment centers. They come in multiple types. Some might be specialized and carry high-value items. Some might carry groceries. We are interested primarily in general merchandise fulfillment centers, which Amazon has placed all over the country. We’re interested in how this diffusion of distribution might have helped Amazon in trying to reach more customers and whether it has helped on the cost side.
Knowledge@Wharton: Amazon’s fulfillment network has grown dramatically over the years, and that comes with an important trade-off that you talk about in your paper.
Seim: Yes, we were interested in looking at how they choose where to put fulfillment centers. You might think that there are general demand conditions. I might want to be near a big city with a lot of demand, especially if those people value things like same-day distribution. But in the early days of online retail — not so much today — where you put your distribution center network also determined whether you would charge your consumers sales tax. We wanted to look at to what extent differences in sales tax across local markets drove where Amazon might have placed distribution centers, at least initially. Today, that’s less of an issue. Initially, you might think that I have a trade-off in that I want to be close to a big population center, and that helps me save on distribution costs, helps me maybe in attracting more consumers if they value the convenience of getting a package quickly. But it might also have downsides if sales tax in those locations is high, so the consumer ends up paying a higher price for getting products because Amazon has chosen to have a distribution center in the location.
Part of that was driven purely by the fact that online retailers are treated differently from offline retailers from a legal perspective in how they have to report and collect sales tax from their consumers. Until recent developments that challenged those laws, most of them had to collect sales tax from their consumers only in locations where they had a physical presence, and a distribution center would qualify as a physical presence. In locations where they don’t have a physical presence, the consumer is supposed to pay use taxes instead of sales taxes directly as part of their annual tax filing to the state. As you might imagine, the vast majority of consumers don’t do that. As a result, if Amazon entered a distribution center into a new state that has high sales taxes, then it is effectively as though consumers experienced a price increase.
“Given that Amazon is so much larger than most of its online competitors, it also gives us a sense of the economies of density that come with size.”
Knowledge@Wharton: How did you analyze the impact of what happened as Amazon was growing this fulfillment center network?
Seim: We used data on consumer purchases at Amazon and other online retailers, as well as spending at offline retailers, to try to look at how demand evolved over time and how that demand related to the fulfillment center network that’s near you when you make a purchase. That allowed us to look at how a demand responds to the presence of a fulfillment center. Do we see a spike in demand in an area where we have same-day shopping as it is introduced? If I know demand and I know how many packages are shipped from a particular fulfillment center to the consumer, I can get a sense of what the shipping costs might be for doing that. I can compare that to what shipping costs might have been had they chosen a different location.
In choosing a different location, you would then say that might have the same kind of demand effects, being closer to a different set of consumers who may look different from the ones you are serving now. Similarly, it might have different shipping implications in that you’re now serving a different set of consumers from a farther distance. It’s these comparisons of moving a distribution center to a different location, relative to what they actually chose to do, that give us a sense of how valuable that location must have been for them to have chosen it.
That’s how we try to get a handle on what might be cost savings from having a less centralized distribution network. Given that Amazon is so much larger than most of its online competitors, it also gives us a sense of the economies of density that come with size. We understand that pretty well in brick-and-mortar retail markets — think about Walmart compared to other local competitors — but we didn’t think there was that much evidence on the importance of those types of scale advantages in the online market.
I think there has been increasing evidence about concentration and growth by Amazon. Most of that has been attributed to quality effects that Amazon brings in terms of same-day shipping and buying assortment. But we wanted to look at whether, because of their size, they also have been able to generate cost savings that give it competitive advantage.
Knowledge@Wharton: Amazon’s product really is convenience and shipping speed because they don’t make the products that they sell. What did you find was the impact of this fulfillment center growth on Amazon’s profits and on customers?
“Statistics show that everybody goes to Amazon first to look at the products, and they don’t necessarily get diverted to a smaller online retailer.”
Seim: We were really interested in the cost side more so than consumer benefits. So, we tried to project out what Amazon’s shipping costs might have been had they not chosen to expand their shipping center network but instead stuck with where they had their initial fulfillment centers. We did a back-of-the-envelope calculation of what shipping costs would have been from serving all of the new demand from the original set of locations compared to where they are today. Then we used that to think about how much might their margin have increased purely because of the distribution center growth and the cost savings that come with that.
We find pretty significant increases, and profitability that is just purely due to the shipping cost savings. We thought that was interesting as a complement to the direction of the press and some academic studies that have mostly tried to focus on these demand-side benefits to being with a retailer like Amazon.
Knowledge@Wharton: Amazon grew significantly in this area over time, but do you feel they will reach a point of diminishing returns?
Seim: That’s a good question. It’s something we can’t answer very well with the data we have access to because our data just tells us very high-level statistics on the distribution center network. I think you’d need to be with Amazon to get more detailed data on that. But there is a big fixed cost to operating this distribution center network, and at some point, rolling out additional distribution centers seems to be unlikely to generate the same kinds of economies of density, or at least sufficient ones to offset the fixed costs that come with it.
… Amazon’s recent efforts at expanding its distribution center network have mostly been geared toward entering new metropolitan areas to be able to offer same-day shipping. You might think about some of the effects of the Whole Foods acquisition in a similar context. Those are expansions of the fulfillment center network that aren’t necessarily geared at purely cost savings on distribution, but they are geared at making a different product for the consumer. There, I think you’d really want to think about both the consumer benefits and possibly the distribution side jointly, and we don’t quite have enough information to be able to do that in our data.
Knowledge@Wharton: Amazon has also found other ways to expand its distribution network that don’t have the high overhead costs that come with a fulfillment center – for example, with the Amazon Locker system that allows customers to pick up their orders from kiosks that are placed inside 7-Elevens or other locations.
Seim: In some ways, that suggests the fixed costs have to be substantial because, on the return side, getting it back through the distribution center and then back through to the original manufacturer, I think that’s quite costly for them. That suggests that there might be other ways that they can partner with retailers to directly target high-demand locations where you can streamline that more easily than with a big fulfillment center. Like IKEA on the Penn Campus.
Knowledge@Wharton: Maybe in the future, customers can go to Whole Foods to return an Amazon purchase.
“It’s these comparisons of moving a distribution center to a different location, relative to what they actually chose to do, that give us a sense of how valuable that location must have been.”
Seim: That seems like a likely development for sure, especially given the match between the demographics of the Whole Foods shopper and the demographics of a high-frequency Amazon Prime member. That seems like a very useful use of space for them.
Knowledge@Wharton: Amazon is such a giant in e-commerce. Are there lessons for other retailers in terms of distribution and supply chains?
Seim: I think so. In particular, it suggests the big other competitor they have in online is Walmart. One reason why Walmart has been relatively successful is because they have a similar, already established distribution center network that allows them to get to consumers quickly. They could try to also integrate stores and online better. I think a lot of what we find would apply to other large chains that try to enter the online market. I think it’s maybe less applicable, but what was interesting to us was to think about what that might mean for smaller online retailers. A lot of the original Walmart debate was about what it does to localities and mom-and-pop stores. In many ways, these recent developments mirror some of those patterns in online markets now. I wouldn’t say necessarily that there are many lessons to be drawn for the smaller retailer, but I think it’s just one explanation for why the smaller retailers are struggling to compete against a company like Amazon.
Knowledge@Wharton: You mentioned in the paper that e-commerce was thought to be the space for the little guy, but it looks like it might not be.
Seim: Right. We initially were motivated by it because in the early days of online commerce, a lot of people pointed to the benefits of much easier comparison shopping. You could check prices in a price engine. You would think that would make markets much more competitive, but I think people increasingly are locked in. Statistics show that everybody goes to Amazon first to look at the products, and they don’t necessarily get diverted to a smaller online retailer. I think a lot of that is just driven by size and assortment, which is just hard to compete with as a small company.
Knowledge@Wharton: What’s next with this research?
Seim: We’re interested in looking at how that interacts a little more with offline markets. In particular, we’ve started looking at what the labor market implications are of a fulfillment center opening in a particular location. There’s pretty good data that looks at what happens to employment in a particular area as a fulfillment center enters. We wanted to think about how that might trade off with the demise of offline retail markets and brick-and-mortar stores, where we see employment falling pretty quickly and pretty dramatically. We want to compare these two and see whether there are also positive effects of online markets increasing so quickly.