As consumer expectations grow, manufacturers try to satisfy every conceivable market segment. For example, in 2004 Colgate-Palmolive sold 35 varieties of toothpaste (combinations of long protection, toothpaste/mouthwash blends, herbal whitening, baking soda and peroxide, plaque and gingivitis prevention, kids’ products, fresh breath, tartar control, and cavity protection). Procter & Gamble offers dozens of varieties of its Crest toothpaste: In 2004 it offered more than nine flavors and three consistencies (gel, paste, and striped combinations) in several formulations.
The multiplicity of products means that each one is selling at a relatively small quantity with the disaggregate demand for each being very volatile. Rather than pooling the demand risk, the multitude of products aggravates it, requiring manufacturers and retailers to keep large safety stocks in order to provide the high fill rates that customers demand. At the same time, global competitive pressures mean that companies have to reduce costs by manufacturing at low labor-cost locations, leading to long forecast horizons. They have to forecast demand accurately to avoid lost sales when there is not enough product, and avoid obsolescence and discounting when there is too much product.
Postponement or mass customization strategy allows manufacturers to satisfy many customer segments, yet it takes advantage of economies of scale in manufacturing and offshore outsourcing in order to reduce costs.1 It is based on redesigning products and manufacturing processes so that a core component, common to a group of product varieties, is manufactured first. The focus in manufacturing the base component is on costs. Thus, it can be manufactured offshore and, if warranted, in long production runs to spread the fixed costs of producing a batch over a large number of items. The finished product is then manufactured or customized from the base product later, according to customer orders, rather than based on a forecast.
Such strategies are being used by many supply chains to provide high service levels for many product varieties at a lower cost than that involved in providing the same number of varieties directly from the factory. But they also provide an inherent flexibility that creates resilience in case of disruptions, since the base product can be tailored to changing demand for products and locations.
In 1997, Marks & Spencer’s annual profits exceeded £1billion for the first time in its history. Continuing the growth trajectory, brisk spring sales at Marks & Spencer made for a great start for 1998. The U.K.-based high street retailer was fulfilling its 1997 promise to shareholders “to become the world’s leading volume retailer with a global brand and global recognition.”2 The profitable company was expanding globally and at home, causing journalists and academics to present Marks & Spencer as a role model for British industry.3
Facing increasing competition from more fashion-oriented yet lower-cost retailers such as Zara, The Gap, and H&M, Marks & Spencer took three strategic steps:
It pressed its suppliers to procure material and finished garments in the Far East and Africa, thereby lowering its costs of goods.
It increased the range of items it carried.
It began shifting from its more conservative, timeless styles to more fashion-oriented styles and colors.
In early 1998, when M&S was deciding on its orders for the autumn/winter season of 1998, grays and blacks dominated the catwalks in Paris and Milan and seemed to be the hot colors of the year. Accordingly, the company ramped up production of the new apparel in those colors, using the slogan “Gray Is the New Black” in anticipation of the forthcoming holiday season. But with its three strategic steps, M&S had exposed itself to three forecasting challenges:
First, it now operated a long supply chain, stretching from the Asian factories where its products were made to its U.K. distribution centers, and from there to its European stores; the result was longer lead times over which it had to forecast.
Second, it started carrying more items and therefore a smaller amount of each; the result was more volatile demand for each of the items.
Finally, with these challenges in place, it entered the much more mercurial market for fashion goods, rather than the steady market for functional clothing for which it was known.
As 1998 progressed, the first signs of the impending storm appeared: Retail sales fell somewhat during the summer, yet M&S executives remained hopeful.
Then the holiday season failed to deliver on the spring season’s promise of a banner year. “We all thought that sales would recover in September and October, and in fact they have gone further south,” said Sir Richard Greenbury, executive chairmen of Marks & Spencer at the time.4 Declining sales created a three-month supply of unsold apparel. The company suffered £150 million in losses: £90 million because of pre-Christmas sales failures and a further £60 million lost from clearing the excess stock.
Marks & Spencer not only had three months of extra inventory, it also had the wrong inventory because of the company’s gamble on gray. Andrew Stone, managing director of Marks & Spencer’s British retailing division, acknowledged, “The customers don’t want 12 million square feet of gray and black. They want a changing range through the season.”5 Worse, Marks & Spencer had canceled new orders when it saw the softening sales. Although the tactic seemed sensible at the time, the company’s 1999 annual report described the impact, noting, “This damaged the balance of ranges because, as popular goods sold out, we lacked the usual injection of fresh merchandise and still had to clear unsold goods.”6 Stores full of unpopular, stale merchandise further hurt sales and caused pundits to charge that the former British icon was out of touch.7
Marks & Spencer’s stock on the London exchange slid from a high of 650p in the last quarter of 1997 to less than 175p in the last quarter of 2000.
It took a long time for Marks & Spencer to realize the forecasting difficulties embedded in its new combination of Asian sourcing, increased number of styles, and, with these in place, its entry into the fashion market. The fashion apparel industry is notorious for supply/demand mismatches resulting from the fickle tastes of the young consumers who make up the bulk of the customer base. A study by the Industry Forum found that nearly onethird of orders in the fashion garments industry have forecast errors of more than 25 percent and that color choices suffer from even worse forecast errors.8
At the time of year when retailers have to place orders for merchandise, they don’t know the next season’s demand. Consequently, they have to estimate what their customers may want and order the merchandise to meet this estimate. Furthermore, forecasting total demand is not sufficient; retailers must forecast the combinations of styles, colors, and sizes that will be sought in each store. If they order too many green blouses and not enough red ones, they may disappoint the customers looking for the red ones and be forced to sell the green ones at a discount or even at a loss. All too often, retailers will have either too much or too little inventory.
The problem is not limited to retailers. Suppliers of computer parts, agricultural ingredients, automotive components, and almost every other product are subject to the same market forces. Customers require the products faster than supply chains can respond.
To avoid disappointing customers, companies can order large amounts of everything, increasing the chances of meeting demand but at the cost of being burdened with excess inventory. With their razor-thin margins, most retailers cannot afford such a strategy.
A related approach is variable pricing. Automobile dealers charge extra for “hot” models while car companies give rebates on unpopular ones; airlines sell seats to business travelers at a premium while discounting seats for leisure travelers; and retailers charge full price for trendy items while discounting unwanted merchandise. In each of these cases, companies use price reductions “after the fact” (i.e., after the product has been built, shipped, and offered in the marketplace) to sell items that customers would not buy at a full price. Pricing is an important weapon in the struggle to balance supply and demand, but it does not create flexibility and resilience.
Chapter 6 explained the challenges of forecasting and outlined several supply chain principles that make companies less prone to forecasting errors. A combination of two of the principles mentioned there underlies the postponement strategy described in this chapter:
• Aggregate forecasts are more accurate than disaggregate forecasts (this is the principle of risk pooling).
• Forecasts over short time horizons are more accurate than forecasts over long ones.
Shifting demand patterns, in particular color preferences, are not a new phenomenon in fashion retailing. In the mid-1980s, Benetton, the Italian clothing manufacturer and retailer, faced that very problem. While it had relatively accurate demand forecasts for the total sales of each of its styles and sizes, Benetton continuously missed the forecast of which colors would catch the fancy of its fashion-conscious customers. It ran out of popular colors, lost sales, and disappointed customers, while having to slash prices on out-of-favor colors. To remedy the situation, Benetton changed its manufacturing and distribution process.
Most clothing makers dye the yarn first, weave it into fabric, and then cut and sew the fabric to create the finished garments. Under this traditional manufacturing system, the manufacturer must decide how much to make in each color six to nine months in advance of replenishing the stores.
In contrast, Benetton redesigned its manufacturing process to make some of its clothes—those with difficult-to-forecast demand for colors—in an undyed, generic state called greige. Benetton dyed a test batch of each new garment and sent it to a set of carefully chosen stores where its sales were monitored closely in order to discern consumers’ preference for colors. With this information in hand, it was easier to forecast which colors would sell well during the rest of the season. Benetton then quickly dyed the greige garments and shipped the items to the retail outlets. It could dye and deliver the latest fashion apparel to all of its stores in only five weeks.9
Such postponement of the dying operation increased Benetton’s manufacturing costs per garment by about 10 percent. On the other hand, it increased sales by reducing stock-outs of popular colors, and it decreased the cost of overstocking and the associated costs of discounts and merchandise liquidations. (Discounts can be well more than 50 percent of an item’s list price; some retailers, such as Boston’s Filene’s Basement, will give their merchandise to charity when it fails to sell after successive rounds of discounting.)
Traditional garment manufacturing forces retailers like Marks & Spencer to forecast winter sales of each color before summer even starts; postponement lets Benetton make the color decision much closer to the selling season.
The postponement strategy had two major benefits for Benetton, both of which had increased its forecast accuracy and allowed it to manage the demand/supply imbalance:
• When deciding on the number and sizes of (griege) items that it needed to manufacture, Benetton had to forecast only the total number of items and sizes regardless of color. This forecast is more accurate than a forecast by color because it is an aggregate estimate.
• When deciding on how many items to dye each particular color, Benetton already had data of actual sales (from the test batches). In addition, the selling season with the current trends was already at hand. The result was that the mix of colors could be determined more accurately.
Products that can benefit from a postponement-based supply chain design are those characterized by a combination of uncertain demand, short customer lead times, high inventory carrying costs, and modular product design.
The first characteristic of such products is that they are facing uncertain demand for many product variants. The opening paragraph of this chapter described the varieties of toothpastes offered by two leading manufacturers. In other markets, the variety can be even more impressive. Nestlé’s Purina offers eleven cat food brands (as distinct from its three cat treats brands), one of which—Fancy Feast—includes 50 varieties in 10 textures. Even if Purina could forecast the total demand for cat food, it is difficult to imagine that it could have the right mix of products at every store at all times.
The second characteristic is short customer lead time. If customers were willing to place their orders ahead of time and wait until their choice could be made to order, or at least shipped from a central warehouse, then postponement would have little benefit. Unfortunately, for the most part, retail customers will not wait, requiring retailers to carry inventory.
The third characteristic that makes a product a candidate for postponement is high inventory carrying costs. Products that are inexpensive to carry can be stocked with little penalty, giving customers instant buying gratification.
Last, postponement has to be technologically feasible. The company must have some means to convert a generic base product into any of the harder-to-forecast product variants. Some types of postponement—like postponing packaging or distribution decisions—can be exercised on most products. But form postponement requires a modular rather than integrated product design, so that the production process can be separated into two phases: manufacturing the base product first, and customizing it at a later time, given a more accurate demand forecast or an actual customer order.
The manufacturing of many small, inexpensive consumer products (such as most items sold in convenience stores) are typically not postponed, even though the variety of products is very large and customers are not willing to wait for special orders. The primary reason is that their inventory carrying costs are not too high. In addition, most of these products are not modular and would be hard to manufacture in stages. But sometimes the product packaging is delayed: Gillette manufactures its razor blades based on a long-term forecast of the total demand for blades, but their packaging is postponed in accordance with specific country and retail-chain orders and forecasts.
In most cases, when companies redesign products for postponement, the total manufacturing costs will go up, reflecting the two-stage manufacturing processes. This cost can be counterbalanced by lower inventory carrying costs, lower discounting, improved availability (higher sales), and better customer satisfaction. Postponement, when implemented with vulnerability in mind, can also increase supply chain resilience.
While the examples outlined in the following sections are not yet motivated by security or vulnerability considerations—because few firms have yet integrated such considerations into their product design—they are given here to demonstrate the basic notion of postponement. As new products requiring new supply chains are developed, and as the likelihood of high-impact disruptions grow, more companies are likely to use such postponement methods to increase their flexibility and resilience.
Hewlett-Packard manufactures its popular Deskjet and Deskwriter printers in its Vancouver and Singapore plants and distributes them to the United States, Europe, and Asia. Selling printers in Europe means following each country’s requirements for printer configurations: different decals, a country-specific power plug, and language-specific manuals. In the past, Hewlett-Packard forecasted demand for each European country and then manufactured the appropriate numbers of printers for each country. Unfortunately, the vagaries of forecast errors meant that HP might have had, for example, not enough printers for Denmark yet too many printers for Slovenia, without an easy way to convert Slovenian printers into Danish ones. Six printer models and 23 different country configurations meant that HP had 138 versions of the finished printers. The result was frequent shortages.
To increase product availability without increasing the retailers’ inventory carrying costs, HP changed its processes, switching to a pan-European forecast and shipping generic printers to its European distribution center in Holland. As the printers arrive in Holland, an easily accessible side panel in the shipping carton lets HP quickly configure printers for each country once HP knows the local demand. This postponed customization operation turns a box with one of the six generic printer models into one of the 138 country-specific printers.
Of course, HP must still forecast the pan-European demand, but this aggregated forecast is more accurate than country-level forecasts. The result is a lower level of printer inventory required to achieve high service levels to the European customers.
Extending the same supply chain design worldwide, HP outfitted its five other regional distribution centers around the world for postponement. Using postponement, HP reduced inventories by 18 percent while maintaining service levels.10 Overall, HewlettPackard cut printer supply costs by 25 percent.11
Building on its success, HP also extended the postponement concept to its packaging operation. In addition to configuring the printers, HP began shipping printers to the distribution centers in bulk and packaging them for distribution to retailers only once the orders were at hand. Postponing the packaging afforded HP several additional advantages:
Shipping costs from the plants to the distribution centers dropped by millions of dollars. Because retail printer packaging includes a lot of cushioning, sending printers in bulk increased the shipments’ density. That meant that HP could ship 250 percent more printers in a container-load from its plants to its distribution centers.
Storage of the bulk-packed printers required 60 percent less space in the distribution centers.
Dealing with unpacked generic printers simplified the basic postponement work at the distribution centers.
HP was able to offer its customers language-specific packaging rather than multilanguage packages. In addition, the printer cartons were “fresh” when arriving at the customer location, with no soiling or scuffing from the extra handling on the long voyage across the Atlantic.
Dell’s assembly operation uses several dozen parts to assemble a large number of possible computer configurations. Table 12.1 depicts the number of options available on a Dell Dimension 4600C desktop computer.12 These options can be put together in almost 100 million combinations, not including the service and accessory options.
Table 12.1: Number of Dell Options
Intel Pentium 4
Storage devices and media
Keeping all, or even a relatively small fraction, of these combinations in stock on retailers’ shelves would mean forecasting what customers may want. With this large number of optional combinations, such a forecast would be challenging. Thus, facing a manufacturing process subject to uncertain demand (new components are coming to the market continuously, making old ones obsolete as customers demand the latest options) and high inventory carrying costs, Dell realized that the modular structure of desktop personal computers allows for postponement at the manufacturing phase. With efficient transportation, Dell could build and deliver computers to users within a few days. Indeed, Dell’s success made it clear that most customers were willing to wait a few days in return for getting exactly the computer they wanted. The result is that Dell can operate with no finished goods inventory.13
Dell does not extend its “build-to-order” postponement strategy to many of its laptop computers. For example, its Inspiron 1000 laptop came in 2004 in three preconfigured combinations and allowed little customization. Many of Dell’s lower-end laptops are pre-built by Taiwanese contract manufacturers. These laptops allow only a limited number of options. The reason is that laptops, in general, are much more integrated, less modular products than desktop PCs and therefore not as easy to customize as desktop computers. In addition, transportation and inventory carrying costs on laptops are lower since they take less space, and these inventory carrying costs are particularly low on the low-end laptop product lines. (Dell offers more configuration options on its high-end laptop, such as the Inspiron 9200.)
Sherwin-Williams house paint comes in a number of varieties (such as interior/exterior, matte/semi-gloss/gloss, and seven different performance features) and an array of colors (more than 1,000 hues in the palette). Sherwin-Williams could never hope to forecast sales of each color, or to stock bulky cans of all the possible premixed colors in each of the 2,500 retail locations that sell its products. Thus, the company postpones coloring the paint. Instead of carrying cans in every color, Sherwin-Williams retailers carry cans of white base paint and colorant dispensers. A database of recipes and mixing machinery lets the retailers blend the color in the store exactly to each customer’s requirements. The strategy is similar to Benetton’s delaying of dying its apparel, except that Sherwin-Williams pushes the customization point all the way out to the retailer, when the consumer is ready to buy.
Because of the low cost and compact size of colorants (a can of paint is only 2 percent colorant), postponement lets SherwinWilliams dramatically reduce the costs of carrying inventory in its warehouses. It also reduces the transportation costs to its customers’ retail outlets while letting the retailers reduce the floor space devoted to inventory. The result is low inventory carrying costs for both Sherwin-Williams and the retailers, and at the same time a virtually unlimited choice of colors for customers. As an added bonus, retailers can use the white base paint and colorants to customize colors to match samples brought in by customers—for example, creating a paint color that matches that of a treasured curtain, or a pre-existing wall color in the house.
Demand for the base paint (the component with the relatively high inventory carrying cost resulting from its bulkiness and relatively short shelf life) in each retail store can be forecast accurately because it represents the aggregate demand for all the colors. The demand for the various colorants can also be forecast with some accuracy because a modest number of colorants can be combined to make all the hues; and in any case their inventory carrying costs are low.14
Birkenstock Shoes started manufacturing in 1774 when Johann Adam Birkenstock was registered as “subject and shoemaker” in the church archives of the small German village of Langenberg. In 1897 Johann’s grandson, Konrad Birkenstock, developed the first shoe with a contoured insole, which together with his 1902 innovation—the flexible arch support—formed the basis for Birkenstock Shoes. The shoes were first exported to the United States in 1966 and became immensely popular in the 1970s. Today the company sells hundreds of styles, each in dozens of sizes and colors all over the world.
Such variety of styles, sizes, and colors implies that each model is sold in relatively small quantities, subject to significant variability. One of the main forecasting challenges is allocating the right mix of shoes to the right U.S. stores.
In 2004, UPS offered Birkenstock a postponement option as part of its international shipping service called Trade Direct. To cut down the distribution time from manufacturing until the shoes arrive at the U.S. stores, UPS took over the entire U.S. distribution of the shoes, allowing Birkenstock to eliminate several interim handling and repackaging steps.
More important, the arrangement allows Birkenstock to delay the point when store allocation decisions have to be made. Instead of making the allocations in the German factories, Birkenstock could now wait until the ocean shipments were received in Newark, New Jersey. This allowed Birkenstock to take advantage of the most recent retailer order updates, just days before delivery. “With Trade Direct, we’ve cut the time it takes to get shoes to stores in half,” said Gene Kunde, chief operating officer of Birkenstock Footprint Sandals Inc. “Our spring fashion merchandise shipped 100 percent on time—for the first time in history.”15
Supply chains that need to respond quickly to unpredictable customer demand, but cannot afford to build large inventories, can benefit from postponement-based design. Postponement provides flexibility in case the demand for one product variation is unexpectedly high while the demand for another is lower than expected. Such variations may, of course, be the result of a disruption. A generic product, prior to customization, can be redeployed or customized quickly to satisfy unexpected demand for a certain variant of the product or demand in an unexpected location. The reason is that undyed, undifferentiated, uncustomized products are fungible while fully finished ones are not.
Postponement also adds flexibility and resilience not only in responding to demand changes downstream in the supply chain, at the customer level, but also upstream. Specifically, one of the benefits of postponement is that the base product can be made in large quantities regardless of the number of final variants. This means that the materials can be made in more than a single plant or by a single supplier, creating both redundancy and flexibility in capacity and skills. For example, HP makes its generic “base” printers both in its Vancouver and its Singapore plants, allowing these plants to “cover” for each other.
Supply chains designed for postponement can also be helpful in cases of anticipated potential disruptions, such as deteriorating labor relationships at a critical supplier; seasonal material shortages resulting from expected high demand; or as a result of growing international tensions. In such cases, inventory of the base product can be accumulated without enduring the large carrying costs that would result from accumulating all the product variants.