The promise of one- and same-day shipping has forced companies to rethink a classic academic problem in the world of supply chain: Where does a company put its inventory? As companies work to live up to the promise of snappy speed, the result is increased demand for warehouse space close to major metropolitan areas and higher inventory spending for many companies.

"This is kind of the age-old dilemma in supply chain: right product, right place, right time, right quantity," Joe Dunlap, the managing director of Supply Chain Advisory at CBRE, said in an interview with Supply Chain Dive.​ "And I think it's been exacerbated the last seven or eight years as retail has become hypersensitive to e-commerce sales and omnichannel."

The difference between one- and two-day shipping is more than just 24 hours. The location of inventory doesn't matter as much for two-day shipping: it can fly just about anywhere in the lower 48 states within that time. But the location of inventory is imperative to get the item delivered within a single day.

"Customers are looking to get their product quicker, so the distance between consumers and the product needs to continue to shrink because of that, which really creates an issue for a lot of people's supply chain," Scott Sureddin, the CEO of DHL Supply Chain North America, said in an interview with Supply Chain Dive.

Strictly e-commerce companies will deal with the puzzle differently than omnichannel operations that have to stock stores and fulfill online orders.

"The biggest challenge is being able to manage those two channels pulling from the same inventory," Angela Jones, an assistant professor of information system and supply chain management at Howard University, told Supply Chain Dive in an interview.

Regardless, any retailer looking to increase speed is trying to figure out what items should be available for one-day shipping and how to get it closer to whoever clicks "add to cart."

A classic academic problem

A well-known curve in the supply chain world shows the number of facilities on one axis and the cost on the other. As you increase the number of facilities the inventory cost increases, but the transportation cost decreases​ up to a point when it then increases again. The trick is figuring out the optimal point — the right number of locations — at which cost decreases the most.

The problem isn't just about the location and network of distribution centers. Retailers are trying to figure out where it's best to fulfill orders from whether it's a DC or retail store.

Figuring out the right location starts with network analysis, which begins with an understanding of the baseline, Jones said. What are the current inventory levels? What is the current demand? Where are the nodes (stores, DC, etc.) in the network? The analysis can get more granular looking at transportation information, transfers between facilities and if the company paid for freight or if it was included in the cost paid to the supplier. The inventory information is often included in an item master file that includes dimension information on all of the SKUs.

Companies also need to understand who is buying their product: "Where is customer demand?" Jones said. The location of inventory, perhaps unsurprisingly, relies heavily on the location of this demand.

DHL is able to see the movements of its customers and can use its internal tools to provide visibility from a manufacturer to its customers, Sureddin said.

Dunlap and his team at CBRE work with clients on network analyses to figure out where storage locations should be for a business. The process begins and ends with math, he said.

"If I model the baseline given where the locations are today and I'm able to demonstrate models calculating pretty darn close to the same financial cost that's in the [profit and loss statement] I've demonstrated I have a working model," Dunlap​ said.

A model of the supply chain's current state makes it possible to model changes and see how those impact the network. Modeling hypothetical scenarios is done by seeding the model. So if you're trying to figure out the best location for inventory then you might give it the top 100 markets or 10 store locations, something to narrow it down slightly, and the model will be able to provide an output that allows for comparison of variables across scenarios, Dunlap​ said.

"You begin to drill down into and validate if they're pragmatic, if they're realistic, how can I really implement," he said.

The team then creates a roadmap to lay out how the selected scenario would actually play out.

So how many locations do you need? A large company fulfilling orders across the U.S. might not need more than 16 to make one-day a reality. Dunlap used the U.S. population as a surrogate for demand to figure out the optimal number of storage locations for a company with a national footprint and determined returns diminished after 16 locations. But not every company will need 16 locations — it will depend on where its demand is located.

How Amazon made it happen

"You can just fly to a customer in two days," Salal Humair​, a senior principal scientist in Amazon's inventory planning and control group within the supply chain optimization technologies team, told Supply Chain Dive in an interview. This is a reality that made two-day shipping a much easier feat than one-day shipping.

At Amazon, the inventory planning and control group within the supply chain optimization technologies team is responsible for optimizing inventory placement by researching the company's network and customer orders. The group considers what products to buy, where to place them, what products to remove from the network and how to keep the network balanced, Humair​ said. And how Amazon deals with inventory questions changed significantly when it decided to offer one-day shipping to its Prime members, he said.

One-day changed the game

The ability to fly anywhere in two-days allowed Amazon to treat its nationwide network as one pool of inventory when fulfilling customer orders. The decision to offer one-day shipping made things more complicated. 

"Inventory really has to be very close to the customers now," Humair​​ said of the switch to one-day. 

But Amazon didn't want to wait to build out more locations, so it started with what it had.

"And that required largely an investment in capital in terms of inventory," he said. "And you might have heard, like the middle of last year, that Amazon invested quite a bit in additional inventory to make Prime one-day by default."

Inventory investment had to be done strategically, though. Amazon had to determine which items would offer one-day shipping and which would not. The decision required Amazon to look at demand and what would sell, which can be hard at a local level.

"The demand at a local level is much, much more variable than demand at the national level," Humair​​ said.​

The items it chooses for one-day also have to be complementary to each other to try to increase the likelihood that if a customer orders multiple items, they will all fall in the one-day offering. Using the classic grocery store example: If you make bread one-day shipping then you want to include milk, too, Humair​​ said.​

The algorithm

Getting the mix right isn't just about making sales. Amazon considers its inventory a risk and lists it as such in its financial filings.

"We endeavor to accurately predict these trends and avoid overstocking or understocking products we manufacture and/or sell," Amazon wrote in its most recent quarterly report. "Demand for products, however, can change significantly between the time inventory or components are ordered and the date of sale."

To get right, Amazon could look to optimize a number of things including cost or speed, but neither of those is the focus of its in-house algorithm, Humair said.

---

"This is a problem that has never been formulated before."

Salal Humair

Senior Principal Scientist in Amazon's Inventory Planning and Control Group

---

"So the North Star for our stuff is the long term value that we create for our customers," he said.

The focus on the customer sounds like a squishy concept that's more of a nice thing to say than an actual company focus, but Humair said his team has been able to find statistical measures for customer value. One thing customers really care about is seeing items in stock. So it has the "customer in-stock value" as part of the model used for optimization inventory and its location.

Location, location, location

After Amazon decided on a mix, it had to figure out where the inventory would go. Software systems that run on mathematical models Humair​​ and his colleagues created help make the decision. But after a customer order is placed, it has to do another optimization problem to figure out the best location to fulfill the order from.

"This is a problem that has never been formulated before," Humair​​ said.​ 

The inventory planning and control group had to treat the problem like any other original research happening at a university, said Humair, who, before his time at Amazon had published multiple academic papers on inventory planning. Amazon's in-house algorithm focuses on optimizing safety stock or the inventory carried above the average demand. This stock is distributed in anticipation of demand and replenished after the demand is realized, he said.

The transition to one-day shipping has raised questions inside Amazon about how it will manage such a network moving forward.

"When you have a network of the type that we are now talking about, there are legitimate discussions about whether we should have some warehouses purely for resupplying our FCs, which are closer to the customers, they don't serve customer demand," he said.

Amazon has also had discussions about how to store items in facilities and how to best manage their third party networks. Amazon can work with a third-party seller, for example, to store some of its inventory, but some items could be slow-moving.

"There are questions about when you have to place inventory closer to the customers, how do you design the network to make sure that the fastest moving things that customers need are in the location close to them," he said.

So while Amazon started its one-day operation with a large investment in inventory, the next steps are continuing to build out its facility footprint taking into consideration a layered design with fulfillment centers and other locations that send inventory to the FCs.

Getting closer, getting smaller

CBRE has seen a push toward population centers play out in the data. Companies are buying smaller locations, closer to large cities, which has resulted in increased demand for smaller storage facilities less than 200,000 to 150,000 square feet.

"In other words, people want smaller facilities that are closer to those population centers," Dunlap​ said.

The length of last-mile trips has decreased as a result. The average length-of-haul for dry van truckloads has dropped 296 miles, or 37%, since 2000, according to a 2019 report by the American Transportation Research Institute.

Still, not every e-commerce company is rushing out to buy real estate.

"A lot of people find it cost-prohibitive and challenging to manage so many locations, particularly small locations," Dunlap​ said​.

But companies can partner with 3PLs to help get inventory in multiple locations around the country. Companies that can't handle each picking can outsource their e-commerce business to a 3PL while using their existing network to handle retail and wholesale operations.

3PLs are the largest segment buying and building warehousing and fulfillment centers. "It's growing and increasing at a higher rate, as opposed to other segments as a result," he said. Though the latest numbers from CBRE show the food and beverage sector is the fastest-growing segment of industrial warehousing as grocery companies update their supply chains to accommodate online sales.

Retailers can also consider another alternative: maybe it just doesn't need to make it to the customer that quickly.

"I don't think it's necessarily a given that they need that fastest delivery for everything," Jones said.