Although supply chain management is widely recognized now as a core component of global competitiveness, demand chain management is often overlooked. This equally important capability means having a good grasp of customer demands—and having a responsive system to meet those demands in a timely and cost-effective manner.

Understanding demand requires a comprehensive knowledge of who your customers are, what products/services they like, how they make their purchases, how often they order, and what constraints they place on the purchasing process. Demand knowledge can then be used to drive a replenishment system that generates the right inputs from suppliers and reliably delivers the products and services to customers.

Demand chain management, at first glance, might seem basic and fundamental. But in reality, comparatively few companies have mastered the concept. In fact, the competitive differential between the best-in-class practitioners and the rest of the pack can be huge.

The key to success for many demand-chain leaders is the smart use of information. These companies make use of a rich set of well-chosen and timely data to drive their replenishment processes. We call this "data-rich" demand chain management.

The result is not only a highly efficient demand chain (one characterized by low operational cost, low inventory, short cycle time, and high customer service) but also increased sales, better market penetration, higher profits, and ultimately, superior shareholder returns.

This article examines two demand chain management leaders, both giant retailers but ones that operate in two very different markets. One is Longs Drug Stores, whose demand chain is so efficient that the company has been dubbed the "hyper-efficient pharmacist." The other is the legendary Seven-Eleven Japan, whose superb demand chain has led to a stock performance that mirrors that of Wall Street darlings like Dell and Intel, even during the Japanese economy's slowdown.

Longs Drug Stores is a major U.S. drug chain with more than 400 retail outlets and $3.7 billion in annual sales. Since its founding more than 60 years ago, the company has emphasized high-quality customer service. However, a corporate target of "never-empty shelves" often translated into lots of dusty inventory.

Today, drug chains like Longs cannot ignore the high cost of inventory. Price pressures from HMOs, insurance companies, Medicaid, and Medicare—combined with ever-higher costs of opening, stocking, and operating stores—have been squeezing margins and increasing capital requirements. At the same time, higher drug prices, the influx of new designer drugs, and the aging of baby boomers have forced the industry to invest more and more capital to satisfy an expanding and diversified demand. The end result for both drug chains and wholesalers has been a diminishing return on invested capital (ROIC).

Longs President Steven Roath several years ago recognized the need to address the inventory situation. Roath felt that the collective wisdom of his pharmacists and buyers was no longer sufficient to manage inventory in a fast-changing environment. Though these individuals certainly were knowledgeable about the broad seasonal patterns of demand for major drugs, such as flu and allergy products, they lacked the data to manage the hundreds of thousands of SKUs sold by the company effectively.

In Roath's mind, a comprehensive information system was the only way to go. Sure, the company had systems in place—specifically, a commercial software package for purchasing at its distribution centers (DCs) and an in-house designed inventory program at its stores. But these systems did not appear to make full use of the available data. Plus, they required heavy manual maintenance. Perhaps the most annoying part was that they were not totally automated. Pharmacists often had to be pulled away from customers during the day to respond to questions generated by the systems.

As part of an effort to "use more science in our demand chain management processes," Roath, in early 1997, invested in a technology supplied by a startup company, NONSTOP Solutions.¹ The NONSTOP technology uses state-of-the-art methodologies to optimize demand chain activities: forecasting, inventory control, transportation, materials handling, and warehousing. The key is that the solution is data-driven, taking away the burden of guesswork for 99 percent of the products.

Today, Longs's performance is the envy of its competitors. Product availability levels at the stores are at 99 percent. Inventory at the DCs has dropped by 65 percent from 1997 levels, while the corresponding store inventory is down 38 percent. These reductions amounted to an overall reduction in drug products inventory of 48 percent. (See Exhibit 1 on the preceding page.) The value at cost of this inventory reduction was $90 million. So, in effect, the reduction helped release $90 million of capital tied up in inventory assets. During a period in which working capital was a particularly valuable asset, this relief enabled Roath to acquire 20 stores of the Drug Emporium chain in Ohio and 38 Rite Aid stores on the West Coast in 1998.

Performance on inventory turns has been similarly impressive. At a time when the industry average for pharmaceutical chains is 5.8, Longs's inventory turn rate averages 9.4—or 62 percent better. (See Exhibit 2.) Longs now is the most efficient drug chain by a significant margin. The company has identified $36 million in annual savings that resulted from the use of scientific methods to run its data-driven demand chain. Importantly, all of this went to the bottom line. In 1999, Longs achieved close to a 20-percent increase in pharmacy sales and a 10-percent increase in net income overall. The company recorded these impressive numbers despite the decline in Medicaid/Medicare reimbursements and increasing pressure on margins facing all drug retailers in the era of managed care.

When NONSTOP took over the management of Longs' demand chain in late 1997, it started by analyzing two years' worth of demand data at the stores and at the distribution centers. This was a painstaking process. In addition to the promotional events taking place during these two years, a number of other special circumstances would sometimes lead to sales spikes and dips. To truly understand the pattern of demands, the effect of such events needs to be removed from the data. Historical sales data usually can be retrieved from a company's database. But identifying promotional and other special events typically requires extra effort. Furthermore, these events are idiosyncratic, both in incidence and regularity, which only adds to the challenge.

It was important to analyze at least two years' worth of demand data because some drugs are highly seasonal. These products may be affected by weather and temperature fluctuations (flu and allergy remedies, for example) or by holidays (some drugs may see spikes after Thanksgiving or Halloween). To isolate the effects of such seasonal effects, you need to analyze a minimum of two years' worth of demand data, preferably more.

Longs and NONSTOP also conducted a critical analysis focusing on day-of-the-week sales. Most drug stores in the United States also sell grocery products, health-and-beauty aids, and other consumer goods. Consequently, customer traffic follows the general retail pattern of increasing on the weekends when most customers shop.

The pattern, however, is not uniform across all stores. Because of the underlying customer demographics, some stores see relatively stronger weekend sales than others. Thus, to replenish in quantities that match demand, Longs needed a good understanding of demand variation over the week at each store. It also needed to be able to capture historical daily demand data while projecting future demand on a daily basis for replenishment purposes. Most commercial software systems use weekly demand as the unit of analysis, both for historical data capture and in demand prediction. The NONSTOP system, by contrast, uses daily demand as the unit of analysis. Fortunately, Longs' database included the most recent daily demand data.

As new sales data are collected at the stores, Longs sends the data to NONSTOP via a secure intranet. NONSTOP first analyzes the data to determine if any special events or promotions could cause spikes or valleys. Next, the revised data are used to update a proprietary forecasting algorithm. Such fine-tuning on a continuous basis enables the system to adapt to changing customer needs or preferences accurately. The system then generates demand projections, capturing the day-of-week, seasonality, and other effects.

Recognizing that forecasts will never be 100 percent accurate, the system attempts to use forecast errors intelligently. First, as the new data arrive, it tracks forecast errors against the performance of the forecasting system in order to identify possible out-of-control conditions early on. Correcting any such conditions could lead to an overhaul of the seeds used in the underlying algorithms. Second, NONSTOP and Longs use the forecast errors to help drive safety stock and replenishment decisions. A product that is easy to forecast—that is, one with low and relatively predictable forecast errors—does not need much safety stock. By contrast, a product with high and fluctuating forecast errors will require much more safety stock. Under the NONSTOP system, safety stock is not determined arbitrarily or subjectively by buyers or pharmacists, as had been the case at Longs and is still the case at some drugstore chains. Instead, it is based on the characteristics of the forecast errors of the individual SKUs.

Once identified, the projected demands and associated safety stocks determine the stores' replenishment needs. These needs could generate withdrawals from the distribution centers to the stores. At the DC level, they could generate procurement orders from the suppliers (which, in the case of Longs, can be wholesalers like Bergen Brunswig or drug manufacturers like Eli Lilly). In addition to the replenishment analysis, the NONSTOP system takes one more important step. It determines whether it may be more economical to ship products in quantities larger than specified in the replenishment analysis. Factors considered here are the costs and means of transportation as well as the costs of receipts, putaways, and handling at the warehouses. The system also determines whether certain SKUs coming from the same supplier would be put on the same truck, or whether products from multiple suppliers in the same vicinity could be co-loaded on a single truck. This additional level of analysis is needed to generate the actual desired withdrawals from the DC to the stores and the orders from the DCs to the suppliers.

NONSTOP also has a dedicated team to help Longs determine the optimal order and delivery frequencies, the optimal sizes of each product stocked at each store, and the most profitable service level for each product. It also works with vendors to help them fill Longs' orders more accurately. In addition, every two weeks, Longs and NONSTOP create a tailored top-10 action list for each store delineating the most important activities that might improve store performance.

The key to the success of Longs' operation is that the multiple activities involved in managing a demand chain—forecasting, inventory control, ordering, warehousing, and transportation planning—are not carried out independently. Instead, the company takes a holistic approach that strives to optimize all of the underlying costs, balancing the tradeoffs between inventory and transportation efficiencies.

Seven-Eleven Japan (SEJ) is that country's largest convenience store chain. Under the strong leadership of Chairman and CEO Toshifumi Suzuki, SEJ has been a business success from the time the company opened its first store in downtown Tokyo in 1974. For the last fiscal year, which ended in February 2000, SEJ's total sales were ¥1,963 billion (about $18 billion), with record operating revenue of ¥327 billion ($3 billion) and net income of ¥68.2 billion ($620 million). This marked the seventh consecutive year that SEJ has posted the highest operating income in the Japanese retail industry.

Seven-Eleven Japan enjoys significantly higher sales per square foot or per store than its competitors do. Average inventory turnover time is between 7.0 and 8.4 days, which means that a store basically turns its inventory about every week. This level of performance has not gone unrewarded. A ¥100 investment in SEJ stock in 1980 would have been worth ¥300,000 in 2000. (Exhibits 3 and 4, respectively, show the company's significant increase in shareholder value and give key performance figures.)

SEJ has more than 8,200 stores in Japan, a number that is growing by 400 to 500 every year. The average store is only about 1,200 square feet in size, or about half the size of a sister Seven-Eleven in the United States. With the size limitation, the typical SEJ store can carry only 3,000 SKUs (compared with, say, a large supermarket, which can have well over 100,000 SKUs). Sales can be classified into processed foods such as drinks, noodles, bread, and snacks (32.9 percent); fast foods like rice balls, box lunches, and hamburgers (31.6 percent); fresh foods such as milk and dairy products (12.0 percent); and nonfood items, for example magazines, ladies' stockings, and batteries (25.3 percent). SEJ is the country's number one retail outlet for fast foods, batteries, and ladies' stockings, and the number two outlet for magazines and paperbacks.

Effective demand chain management is a key factor behind SEJ's success. The company has created a solid information system that provides timely and comprehensive signals about market demand. It also has developed an intelligent process to turn such data into useful information for product replenishment and new product creation. In addition, the company has built an extremely agile logistics system that supports product replenishment to the stores. In a very real sense, SEJ has sharp eyes to see what the market likes, a smart brain to quickly develop plans to react to the market preferences, and a fast arm to deliver the plan.

In 1991, SEJ began using an Integrated Service Digital Network (ISDN) to link its retail stores with the central headquarters. The two-way communication gives franchisees direct access to the host computer and the central database containing POS (point-of-sale) data and analyses. In place since 1998, this advanced system integrates store-level information with supply chain-wide data by using new client applications, satellite communications, and Internet connections.

When a customer comes to the checkout counter with a basket of items, the clerk first keys in the person's gender and age (estimated) on a separate keypad. The clerk then scans the bar codes of purchased items. These sales data are passed on to headquarters via the ISDN. At the same time, the data are processed by an in-store computer system that controls all equipment and peripherals in the store. The in-store computer enables both the store manager and SEJ headquarters to update and analyze POS data simultaneously. Using the real-time on-site information, store managers can analyze the hourly sales trends and stockout rates of all SKUs by customer groups. Headquarters aggregates the data by region, product, and time, and makes that information available to all stores and suppliers by the following morning. (Exhibit 5 presents an overview of SEJ's information system.)

Every Monday, Chairman Suzuki presides over a business meeting in Tokyo attended by 100 corporate managers. In the morning, they review the stores' performance for the previous week; in the afternoon, they develop strategies for the upcoming week. On Tuesday morning, the strategic guidelines are presented to SEJ's operation field counselors (OFCs), who have gathered in Tokyo. Each OFC is in charge of about eight stores. That afternoon, these field managers assemble in regional meetings to map out tactics for executing the strategies. A key part of this exercise is to consider local factors such as weather, road construction, advertising programs, and activities such as sporting events. They also assess any local trends in consumer tastes. On Tuesday night, the OFCs fly back to their regions. The next morning, they visit their respective stores to deliver the messages developed at headquarters and help them implement the tactics recommended for the week.

The POS data and the feedback from the weekly meetings are summarized in a list of recommended SKUs and tactics for actions. The store manager walks along the store aisles carrying a hand-held device that can be used to check stock levels and sales trends as well as place orders to headquarters via the ISDN. Aggregated orders then are transmitted from headquarters to the manufacturers, wholesalers, and distribution centers. Orders for fast food and fresh food items are placed three times a day, magazines once a day, and processed food items three times a week.

SEJ also uses sales trends in deciding whether to keep or drop an item. Typically, a new product reaches its sales peak within a week or two and begins to decline several weeks later. When per-store sales drop to a certain level, the product is deleted from the recommended list. The life span of most products is shrinking over time, and new products are being introduced and older products dropped at a faster rate. Of the 3,000 SKUs carried by each store, about half are replaced every year. POS data also are used to forecast future consumer trends and to assist manufacturers in new-product development. In the early 1990s, for example, sales of semi-prepared fresh noodles were going up at the expense of dry ramen (thin-coiled noodles). Spotting this trend early on, SEJ developed a new category of fresh noodles jointly with the manufacturer, Nisshin.

Seven-Eleven Japan also effectively uses the data to adjust store layout multiple times a day. For example, a store may detect a sales pattern in the sizes of milk containers sold at different hours of the day. The store manager can rearrange the milk products accordingly so that customers can easily pick up their favorite items when they visit the store.

The lesson here is that to have the right product replenished in the right quantity at the right time, you need comprehensive data on sales, purchase patterns, and customer profiles as well as on store characteristics and any local constraints. SEJ makes an intensive effort to collect these data not just through advanced technology but also through the human effort of headquarters planners, operation field counselors, and store managers. The process may seem laborious: weekly management meetings, OFCs flying to Tokyo every week and then visiting all of the stores, store managers checking on individual SKUs in order to make replenishments and merchandising decisions. But the results speak for themselves.

In a conventional Japanese distribution system, each manufacturer has its own designated wholesalers, which exclusively distribute its products. But Seven-Eleven Japan requires frequent deliveries of small lot sizes because of the stores' limited storage space. For this reason, the company created a "Joint Delivery Program" whereby product groups in the same temperature zone are cross-docked at a single supplier distribution center (SDC) and delivered in truckloads to groups of stores in different geographical regions. The SDC is run jointly by suppliers. This consolidation and cross-docking operation has decreased the average number of deliveries to each store from 70 per day in 1974, to 12 in 1990, and down to 10 per day at present.

Store delivery routes and times are painstakingly planned and executed. Drivers are expected to make the delivery within 10 minutes of the established timetable. On average, they spend one and one-half minutes at a store. If a delivery is late by more than 30 minutes, the transportation company pays the store a penalty equivalent to the gross margin of the product being delivered. The transportation company maintains radio communications with drivers, and SEJ headquarters keeps a log on all delivery activities. When an emergency arises, drivers, the store manager, and headquarters communicate with each other to take corrective action.

To handle the uncertainties of traffic situations in Japan, SEJ forces its logistics partner to diversify its vehicle portfolio to include trucks, motorcycles, ships, and even helicopters. On the day of the Kobe earthquake, SEJ pressed into service seven helicopters and 125 motorcycles to rush delivery of 64,000 rice balls to earthquake victims in that city (an area in which SEJ has no stores, by the way). Traffic on the Kobe highway that day had slowed to two miles per hour.

Given the importance of frequent replenishments to prevent stockouts, SEJ's investments in the logistics system are understandable. Without such an agile operation, the promise of smart replenishment could never become reality.

SEJ's unrelenting attention to data-rich demand chain management has paid big dividends over the years. To illustrate, the average daily sales per store were ¥366,000 ($3,300) in 1977; in 2000 they were ¥681,000 ($6,200). During the same period, average inventory turnover time improved from 25.5 to 8.4 days, while the average gross margin rose from 24 to 30 percent. It's important to note that the benefits of SEJ's system extended throughout the entire supply chain to franchisees, suppliers, and logistics service providers.

SEJ's logistics system and data-rich processes have been emulated by competitors like FamilyMart, Lawson, AM/PM, and Sunkus. As a result, the whole convenience store category in Japan has enjoyed a level of success and growth unmatched in other countries.

SEJ and Longs share a number of common experiences and points of view. Both have enjoyed great financial successes as a result of their smart demand chain operations. Both also recognized that data are the key to this success and have developed elaborate information systems to capture POS data to drive replenishment in the demand chains. The comprehensiveness of the data collected at both companies is particularly impressive. At Longs, every store collects and stores data daily, allowing the company to capture and act on day-of-the-week fluctuations. The retailer also keeps data for an extended period of time to detect trends and seasonal patterns. At SEJ, point-of-sale data are captured hourly, thereby enabling detailed analysis of demand patterns. In addition, basic customer demographics (age and gender) are collected, along with local events and other region-specific data.

Longs and SEJ also use the data to create knowledge. Replenishments are based on sound analyses of the underlying data, which results in smart decisions. The success of this approach is reflected in the operating results of both companies.

Yet while the two companies share many common traits, there are important differences based on their respective environments and physical characteristics. These underlying characteristics have forced SEJ and Longs to focus on different levers in managing their demand chains. (Exhibit 6 compares the environments and levers for the two demand chains.)

In the case of SEJ, the company must go to extraordinary lengths to ensure that its stores are stocked with the right product to avoid stockouts and eliminate nonperforming items. To maximize the sales opportunities at each location, SEJ must carefully tailor the merchandise selection to the tastes and needs of the local market. Additionally, given the limited number of SKUs, the company's supply base is not as widespread as those of other retail chains.

The costs of stockouts are particularly high in an operation such as SEJ's. Because of this, the company has developed a sophisticated logistics system that can bring products to the store at a high frequency and in small quantities. The proximity of the supply bases enables SEJ to consolidate shipments from the multiple suppliers and cross-dock products to the stores. This approach incurs high logistics costs relative to other retail chains (such as Longs). But because product availability is so important, SEJ believes that the additional costs are justified.

Thanks to the small number of SKUs and localized product offerings, SEJ can afford to use intensive human inputs and analyses in developing replenishment plans and making product merchandising decisions. The need to have just the right product in the right quantity justifies the labor-intensive processes and the involvement of multiple parties in the decision-making. And because of the extensiveness of the data used in these activities, SEJ's demand chain can be considered to be data rich.

Longs, on the other hand, operates in a different environment. Its stores are much larger than Seven-Eleven Japan's and carry many more SKUs. Plus, they tend to be more uniform in terms of product offerings. Having fewer space constraints gives Longs a little more flexibility in carrying inventory.

Given the greater geographical dispersion of the stores (from Hawaii to the Midwest) and the supply bases, transportation costs play a relatively greater role at Longs than at SEJ. Therefore, it's essential to make replenishment decisions that arrive at the proper balancing of inventory, transportation, and product handling costs. Longs requires a more automated replenishment system (such as that offered by NONSTOP) that uses scientifically based methodologies to perform the right statistical analyses of demand patterns while balancing inventory, transportation, and handling costs. Often this results in moving merchandise in bulk or in batches—as opposed to the frequent, small-order quantities that move to the SEJ stores.

As the Longs and SEJ examples clearly demonstrate, demand chain management is both art and science. Although both companies have adopted certain common principles, each has applied specific techniques appropriate to its respective environment and operating characteristics. Perhaps the most compelling lesson learned from these leaders is that smart demand chain management goes far beyond cost reduction. It also brings great opportunities for market and sales expansion (in the case of Longs) and for significant shareholder value appreciation (in the case of SEJ). Put another way, smart data-rich demand chain management is a powerful vehicle for value creation.

Footnotes

¹ The authors serve on the scientific board of advisors to NONSTOP Solutions.