Beyond the Speed-Price Trade-Off
In the early days of online retailing, e-commerce companies fulfilled consumer demand from a small number of large-scale warehouses that carried similar catalogs of items. Retailers stocked inventory for low-volume products in as few locations as possible while maintaining service levels that met customer expectations. It was a way to keep inventory costs low and take advantage of the economies of scale that large fulfillment centers provide. Since consumers were willing to wait for deliveries, proximity and speed were less important than cost savings.
But the online retail market has changed. Today’s shoppers want more than low prices — they also want the products they order delivered quickly. To achieve same-day delivery, retailers are experimenting with new business and operations models, including using third parties (such as local city-specific delivery services), crowdsourcing (such as paying individuals by the task to shop for and deliver groceries), self-service (such as setting up physical lockers where customers retrieve their packages), and even unmanned aerial vehicles, or drones (which could deliver packages in less than 30 minutes in some locations). At the same time, many retailers, including Amazon.com, Nordstrom, and Macy’s, have recently redesigned their distribution networks. A growing number of omnichannel retailers are using their physical store networks to fulfill online orders (for example, they may be shipping online customers’ orders from physical stores or allowing online customers to pick up their packages at physical stores), while online-only retailers are adding warehouses, particularly near major urban markets.
The trade-off between cost and response time has traditionally been one of the primary factors companies consider when designing their supply chain networks. Historically, if you wanted something right away, you expected to pay significantly more to account for the costs retailers incurred to maintain local inventory or provide high-speed shipping. Lately, however, the terms of competition have changed.
In response to increasing consumer demand for fast deliveries at no extra cost, more companies are implementing IT solutions that enable access to real-time sales data and inventory data across the whole enterprise. Real-time sales and inventory information, coupled with advanced analytics (such as recently developed network-wide fulfillment algorithms), enable networks to accommodate fluctuations and changes in the business environment quickly, a quality we call distribution agility. The result: Retailers can treat their whole distribution footprint as a single entity as opposed to a group of individual depots.
We recently studied the impact of proximity and agility in supply chain network design, examining how online retailers can benefit by restructuring their distribution networks to move beyond scale-based network design. In addition, we studied ways in which companies with agility-enhanced networks can leverage centrally controlled systems through better fulfillment and replenishment algorithms. Through this research, we found that scale and responsiveness don’t need to be in direct conflict with each other.
In fact, the ability to deliver on-the-fly fulfillment and network-wide replenishment means that retailers can offer faster delivery times without driving up costs much and can even improve their resiliency against risks of disruptions.
Incorporating agility into the distribution system is a three-step process that involves rethinking network design, planning for information centralization, and building inventory and pricing into order-fulfillment decisions.
1. Rethink network design. Implementing distribution agility begins with redesigning your physical distribution network and the information network that supports it. In both centrally managed systems that can respond immediately to new information (so-called agile systems) and traditional systems, the network design has important implications for both cost and performance as they relate to customers. Consider Amazon.com Inc. Since 2013, growing demand for rapid delivery has led the online retailer to open 43 small-scale delivery stations and 53 hubs in the United States to augment a distribution network of 101 fulfillment centers and 29 sorting centers. Traditionally, expanding a network this way undermines the scale economics. But real-time stock visibility across the network and intelligent product replenishment and fulfillment significantly mitigate the cost of this trade-off.
When physical stock is distributed, information about supply and demand needs to be centralized. Having a real-time information system that incorporates data on sales by time and location, inventory availability, and replenishment schedules helps a retailer satisfy and predict demand on the fly. Some companies are already making use of this capability. One example is Cainiao Logistics, a China-based company that is majority-owned by Chinese online retailer Alibaba Group Holding Ltd. Cainiao is an online retail distribution leader in that country, executing more than 50 million deliveries per day. Its real-time big-data algorithms determine optimal fulfillment routes by calculating inventory availability at different warehouses and assessing traffic and weather conditions. The algorithms also provide couriers with forecasts of demand spikes. By leveraging massive data sets from disparate sellers, selling platforms, and third-party logistics companies, Cainiao’s centralized decision process increases efficiency throughout the selling network.
2. Plan for information centralization. As the size and complexity of the network grows, the likelihood of stockouts at any particular node is likely to increase, leading to dynamics that drive costs higher. To minimize the downsides, the data analysis system should be able to make forecasts and manage inventory.
Forecasts should include a variety of elements: hourly demand, information about the time sensitivity of orders (how fast customers want items delivered), traffic conditions, labor requirements and likely absenteeism, product return information, and customer price sensitivities. Inventory management can be complicated, partly due to the fact that there is a high degree of uncertainty over the lead time (the time between when a retailer places a replenishment order with a vendor and when the order arrives at a warehouse). Not only is demand random, but a stockout during the period when an item is on order at one fulfillment center can overload demand at a second fulfillment center, causing a domino effect of stockouts across facilities. Algorithms that examine the network as a whole can mitigate these problems by predicting the cascading stockout patterns that might occur across the network and ordering accordingly.
3. Build inventory and pricing into order-fulfillment decisions. The final step is to enable flexible operations that consider inventory and pricing across the entire organization at the same time. For example, should you send the last game console in a given warehouse to a nearby customer who requested a five-day delivery window? Or, recognizing the local unit might be needed for a customer who requested one-day delivery, should you fill the five-day-delivery order with a console from a more remote warehouse at a slightly higher cost?
An agility-enhanced distribution system should be able to make such decisions quickly and estimate the value of a specific item in a specific warehouse at a given moment in time, given current system-wide inventory levels.
There’s another lever managers can use when a store has too much inventory. Traditionally, when inventory exceeds demand, retailers are under pressure to mark down prices to maximize revenue. However, in agility-enhanced networks where pricing and fulfillment decisions are made jointly, online orders can be fulfilled from overstocked stores, reducing the need for markdowns.
An Outmoded Conflict
Traditionally, retailers could not offer both low prices and fast delivery: No free lunch was ever served quickly. The core benefit of embedding agility into a retail distribution network is that it mitigates the traditional conflict between responsiveness and efficiency. At the same time, it enables retailers to keep prices low while meeting consumers’ preferences for faster deliveries.
In the broader supply chain context, the concept of agility could extend beyond retail, for instance, to manufacturing. The benefits of this approach are visible at toy maker Lego Group, which recently completed a large-scale redesign of its supply chain. With most of its manufacturing base concentrated in Europe, Lego, based in Billund, Denmark, often faced challenges fulfilling variable holiday demand in North America and Asia because it had long manufacturing and shipping lead times. To overcome this bottleneck, the company added new plants in China, Mexico, and Hungary to be closer to key markets. To take full advantage of the improved market proximity, Lego made further efforts to enhance its distribution agility. By improving its forecasting analytics, Lego was able to dynamically allocate manufacturing capacity between different products and parts, which is analogous to dynamically allocating customer orders to warehouses in a retail setting.
A secondary benefit of embedding agility into the distribution network is that it makes networks more resilient. Scale-based networks can have serious difficulty recovering from disruptions. The failure of any node or link can effectively shut down a substantial portion of network capacity. Agility provides protection against risks from both market factors (such as labor strikes and increases in the costs of materials) and nonmarket factors (such as weather).
We have found that, whereas scale-based strategies can only adapt to fluctuations by adjusting the distribution network’s design, agility-enhanced strategies can make adjustments at the planning and operations levels. Increased network density provides a natural hedge against disruptions in that it limits the impact to a smaller service area. With agility-enhanced networks, there’s less need for major adjustments even when inventory cost factors fluctuate.