M&SOM Review

As Internet retailing grows, so does the complexity of the supply chains that support it. Initially many organizations utilized a handful of regional fulfillment centers (FCs) to serve customers’ online orders. Each FC was dedicated to a region, and operated independently of the other FCs. By ignoring interactions among FCs, firms could employ traditional inventory policies designed for decentralized distribution systems, policies that are half a century old.

As the online economy has grown, retailers have implemented two novel strategies for order fulfillment.   First, retailers have enabled online orders to be shipped from their brick and mortar stores, as a way of getting more leverage from their stores’ locations and inventory. Now, every Walmart, Best Buy, Target, Home Depot, and (just recently) Whole Foods acts as a potential “miniature” FC. Second, each FC is no longer dedicated to a specific geographical region. A Macy’s retail brick-and-mortar store in Florida can ship a pair of shoes to an online customer in Seattle. The customer in Seattle does not care from where his order ships as long as 1) it arrives on time and 2) his shipping costs depend only on the delivery time he requests, not on the actual costs Macy’s incurs.

Suddenly, what was once easy to manage with traditional inventory policies is difficult. The network has been both dramatically expanded and instantly interconnected in a fairly complicated way. When the Seattle customer orders his pair of shoes, the fulfillment decision depends on the customer’s requested delivery speed, as well as the inventory in every store and every FC.  On managing the inventory, the firm must account for the interactions between the stores and the FCs; in particular, the inventory management needs to recognize that when an inventory site stocks out, its nominal demand will “spill over” to another site that has the inventory (see Figure 1).


Figure 1. Illustration of “Spillover”

Notes. When one FC is out of stock, it cannot serve the customers nearest to it. In this case the FC on the left ships to both customers, depleting inventory twice as fast as compared to serving its nearest customers only. One customer is nearby and can be served cheaply with a truck. The customer on the right, however, is distant and her order must be shipped at a high cost via airplane in order to meet her delivery deadline.

These stockouts are important because any inventory management policy attempts to have “the right” amount of inventory in each FC or store; to do this, the firm must estimate how the inventory in each building will be depleted over any period of time.  Furthermore, the retailer must make these fulfillment and inventory replenishment decisions in a “messy” environment. Things like the laws of physics (only so many pallets can fit inside each building), the laws of human nature (not everyone shows up to work every day), and the law of Murphy (what can go wrong will) disrupt the network, which is another way of saying the inventory is not (and never will be) exactly where it should be.

In our recent article (Acimovic and Graves 2017), we investigate inventory management in an online retail environment. We show that 1) traditional inventory replenishment policies applied to Internet retailing create dynamics that can be costly; 2) inventory is more imbalanced across the system than one would naturally expect, presumably due to the aforementioned laws; and 3) our new inventory policies lead to significant savings (with respect to outbound shipping costs), especially in the face of operational realities.

We partner with a pure online retailer that replenishes its inventory at set intervals (perhaps weekly). At each order occasion, the retailer must decide the amount to order from a vendor and how to allocate that inventory to the FCs. Formerly, our partner retailer employed a traditional inventory replenishment policy, whereby each FC had its own “order-up-to level.” At each order epoch, each FC would order an amount that brought its inventory up to this level. We find that such a policy results in a phenomenon we call whiplash. When a system starts out of balance, an FC will stock out in one period and its demand will spillover to another FC; the stocked-out FC will then order too much while the other FC orders too little. In the next period, the table is turned: the other FC stocks out and demand spills over to the first FC. This cycle can repeat indefinitely where an FC will order “too much” in one period then “too little” in the next, and so on, resulting in costly out-of-region shipping. We find evidence of this whiplash dynamic on replenishment data from our industrial partner.

This whiplash phenomenon is triggered by an inventory imbalance. How prevalent is this in actual systems? Imbalance arises because “stuff happens” in real systems: a vendor ships the wrong amount or to the wrong FC; an FC has to shut down or scale back due to weather or labor shortage; customer orders get rerouted to avoid order splitting; regional demand spikes occur; and so on. Randomness in demand is the one operational reality that inventory analysts and academics most often account for, while the others are left for the practitioners to deal with in the real world. By analyzing data from our industrial partner, we find that a 10% increase in the level of inventory imbalance can be attributed to these operational realities. 

Based on these empirical findings, we propose a new inventory management policy that mitigates whiplash and spillover, is robust to inventory imbalance, and accounts for network effects. The policy works like this: Each FC has a new order-up-to level that accounts for network effects and potential spillover. At a replenishment epoch, the retailer estimates how much inventory will be on hand in each FC at the time the replenishment order will arrive. This estimate of the inventory remaining in each FC takes into account what stockouts and spillover may occur over the vendor lead time. The firm then orders the difference of the order-up-to level and the on-hand inventory estimate for each FC. When simulated on a realistic network, this policy resulted in 4% to 6% reduction in spillover and 1% reduction in outbound shipping costs when compared to the traditional policy. In 2016 Amazon spent 13% of its international net sales on outbound shipping costs, equating to $16 billion dollars (does not include Amazon Web Services net sales) (Amazon.com, Inc. 2017). A 1% reduction in these delivery costs is a savings of $160 million.

In the new online retailing environment, traditional inventory policies can be ineffective because they ignore important network effects. New policies must be developed and tested that account for these network effects as well as operational realities. This research is a first step in that direction.



Acimovic J, Graves SC (2017) Mitigating spillover in online retailing via replenishment. Manufacturing Service Operations Management 19(3):419–436.

Amazon.com, Inc. (2017) Form 10-k for 2016. Accessed July 3, 2017, http://www.sec.gov/edgar.shtml.

Let us first read a story written and posted on the thekrazycouponlady.com:

Like many Krazy Coupon Ladies, I consider myself a smart and savvy shopper, but I'll be the first to admit that sometimes my emotions get the best of me. I saw a beautiful cashmere sweater at one of my favorite shops last week for 30% off; it even came in petite sizes—it was too perfect. I thought I had an amazing deal. It was still more than I would normally spend, but I went against my gut and bought it anyway. The next day I got an email that they were now all 40% off—just my luck! Thankfully, while I was online I came across their price adjustment policy and got an additional $10 credited back to my card. Since I had such good luck with this, I thought I'd share a few other fabulous price adjustment policies so hopefully you can save more money and stress less this season! (The Krazy Coupon Lady 2016; emphasis added)

View Full Post »

As every savvy consumer knows, successful products are often quickly copied by lower-cost producers (copycats). This can be observed in a wide variety of manufactured items, from pharmaceuticals to electronics. These copycat firms piggyback on the success and creativity of others by quickly introducing their own facsimile of a much more expensive product. To combat the copycats, firms may turn to legal means. However, lawsuits are often difficult as the copycats skate the fine line between stealing designs and inspiration. In some cases, government protection in the form of patents can be helpful (for example, audio equipment manufacturer Klipsch has sued competitor Monoprice for patent infringement, and fashion firm Burberry includes legal action as one of its strategic initiatives). In other cases, a manufacturer may choose to compete on price, hoping that consumers will stay loyal in exchange for price concessions (for example, research has shown that two years after the introduction of a generic version of pharmaceuticals the cost of treatment falls by an average of 35.1%). In still other situations, the designer of an original product may be left to basically ignore the copycat for lack of effective options.

View Full Post »

From April 2017, large UK companies are required to report their supplier payment practices and performance twice a year, according to a new regulation of the Department for Business, Energy and Industrial Strategy (https://www.gov.uk/government). Firms must report detailed contractual terms such as the time they take to pay their invoices. And the United Kingdom is not the only country aiming to regulate firms’ payment behavior. France, for example, has passed a law that prevents French trucking firms from extending payment terms more than 30 days.

View Full Post »

Electric vehicles (EVs) have been considered an integral part of smart city development and offer a promising solution to environmental issues in transportation. EVs produce no tailpipe emissions and offer significant improvements on well-to-wheel energy efficiency and emissions levels over their gasoline counterparts, especially when powered by clean sources of electricity (e.g., solar and wind power). The diversity of power sources also make EVs less sensitive to fossil fuel depletion and to supply uncertainty of crude oil. Despite their potential, however, mass consumer adoption of EVs have been hampered by several major hurdles, including their short driving ranges (coupled with inadequate public charging facilities), their high upfront purchase costs, and their potentially high depreciation rates due to rapid technology development.

View Full Post »

In the past decade, about 95% of the renewable power generation resources added in the United States was wind and solar power. Generating energy from these intermittent renewable resources requires no fuel and produces no pollution. Because of these obvious economic and environmental benefits, these renewable resources have often been prioritized over fossil generation; that is, electric grid operators aim to use as much wind and solar power as possible. There was nothing wrong with prioritizing intermittent generation in its early stage of development, but as intermittent resources increasingly penetrate the power supply, the grid operator’s task of balancing supply with demand has become more and more challenging. What are the ways and associated costs of buffering the variabilities introduced by renewable resources? Are there any benefits of curtailing (i.e., reducing the output of) intermittent generation? How does renewable energy and its curtailment affect the power market competition?

View Full Post »

For nearly three decades, the predominant trend in the U.S. manufacturing industry has been offshoring, which refers to manufacturers moving or building factories offshore (most prominently in China) to produce goods for the onshore (U.S.) market. The top driver of this trend has been lower labor costs in emerging economies. Recently, however, this labor arbitrage has been gradually tapering off as wages in China grew 10–20% annually. As such, a growing number of U.S.-based companies started to consider bringing factories back to the United States—dubbed reshoring—and some have taken actions. However, the adoption of reshoring is slower than many have hoped, generating much debate about whether reshoring is viable and scalable.

View Full Post »

The educational service since the early days of the School of Athens was built on the idea of a service provider (i.e., the instructor) delivering the content and orchestrating a discussion among participants. The service quality, measured by learning outcomes, depends critically on the quality of discussion and the instructor’s ability to engage participants in the learning process (i.e., the Socratic method). The main issue with this method is that it is expensive: college student loan debt in the United States, reaching $1.2 trillion, is the second largest debt after mortgage debt.

View Full Post »

From management consulting to legal teams and flight crews, product development or scientific research teams, fluid teams are the norm rather than the exception in many knowledge-intensive work settings. Such teams are assembled for a specific purpose and operate for a limited time, after which they dissolve and some team members may work together again as part of another team. Fluid teams are particularly prominent in dynamic operational environments of healthcare. Imagine a surgery unit of a busy hospital with continuous assembly and dissolution of teams.

View Full Post »