A little of everything can go a long way

One of the consequences of product proliferation is the growing number of consumer goods with a low rate of sales per retail outlet. This phenomenon is evident in a wide range of sectors: apparel, spare parts, appliances, electronics, and even some food categories, to name a few. It has become so widespread, in fact, that an executive vice president of a British retail chain that offers a daily assortment of tens of thousands of stock-keeping units (SKUs) recently noted, "In terms of sales velocity, our products fall into three categories: slow movers, very slow movers, and ... deadly slow movers."

As that comment suggests, low-volume SKUs have become a serious problem for retailers, in part because traditional ways of dealing with slow sellers often are ineffective. One reason why is that most sales and inventory practices in consumer-products industries are based on high sales volumes. Another is that convention holds that local store managers are in the best position to determine which slow-selling SKUs to carry, at what levels, and how frequently they should be restocked. Yet it is impossible to accurately determine what local customers want. Indeed, most SKUs are sold at a rate well below one unit per week, even in large retail stores; this equates to perhaps one out of the thousands of potential buyers entering the shop, or to one person out of many hundreds living in the vicinity, actually purchasing that item. Any attempt to identify in advance where those few buyers—who are "abnormal" from a statistical point of view— will shop will be in vain.

Article Figures

[Figure 1] Numerical distribution and average sales per SKUEnlarge this image

[Figure 2] Sales pareto chartEnlarge this image

[Figure 3] Inventories at point of saleEnlarge this image

[Figure 4] Numerical distribution and average sales per SKUEnlarge this image

[Figure 5] Definition of the daily shortage rateEnlarge this image

[Figure 6] Daily shortage rateEnlarge this image

[Figure 7] Daily shortage rate and replenishmentEnlarge this image

[Figure 8] Definition of the sales uplift rateEnlarge this image

[Figure 9] Relative weight of the pilot zone and sales upliftEnlarge this image

[Figure 10] Effect on the shortage rate during the pilotEnlarge this image

[Figure 11] Relative weight of the phase 1 zone and sales uplift (weekly)Enlarge this image

[Figure 12] Sales patterns for phase 1 regions versus the rest of the countryEnlarge this image

[Figure 13] Sales patterns for regions outside phase 1 zoneEnlarge this image

[Figure 14] Sales uplift in different countries and sales channelsEnlarge this image

[Figure 15] The paradigm shift in assortment and replenishmentEnlarge this image

Through illustrations and a detailed case example, this article will argue that a more effective strategy for managing low-volume SKUs would be: a) every store should in principle carry the full range of products, even if in very small quantities; and b) stocking decisions should be made centrally, rather than at the store level.

This strategy may seem counterintuitive, but its effectiveness has been validated in several real-life tests involving a leading consumer electronics company. As this article will demonstrate, after implementing this strategy, the company immediately experienced an improvement in sales of low-volume SKUs. Importantly, it achieved this benefit without increasing inventories or costs.

What to stock, and where?
Managing the assortment of low-volume items at the point of sale (PoS) raises questions in two areas. The first concerns customization of the product range:

• Where should products be offered for sale? (Will there be more demand for a specific product at PoS A or PoS B?)
• What level of sales makes a product viable at a particular sales outlet? (One unit every third day? One every 27th day?)

The second area is the allocation of decision- making responsibility among local (point of sale, region) and central (country, company) entities:

• Is the "local" judgment regarding the sales potential of any SKU more legitimate than the judgment of a "central" decision maker? Does local knowledge about customers really ensure a more accurate assessment of the sales potential of specific items? This is an important question, given that sales success may sometimes depend more on a powerful marketing plan and enticing product displays than on a response to consumers' predicted needs.
• What is the effect of local autonomy on overall sales activity? Does the degree of freedom enjoyed by the stores relative to stocking decisions lead to a proliferation of SKUs, thereby limiting the emergence of best sellers? Or does it result in missed sales if attractive products are not offered to local customers?

A look at a selection of consumer goods found at large national retailers will help to shed light on the answers to these questions.

Example #1: small electrical appliances. The first example is a world-renowned manufacturer of small electrical appliances, which is selling some 50 products (SKUs) to consumers through a leading U.S. supermarket chain. An analysis of the numerical distribution and average sales for those products, shown in Figure 1, reveals three main findings:

1. Each product has average sales of only 0.8 units per week in retail outlets where it is available on the shelves.
2. A very significant portion of the product range— 40 percent—is deployed in less than half of the retail outlets.
3. The two products circled in Figure 1 are offered for sale in a small subset of the retail outlets, yet they have achieved—where they are displayed—unit sales that are similar to products that are available in most stores. This raises the question of whether they should be offered for sale at a larger number of stores. In fact, this question could also be raised for the products in the bottom-left quadrant, which when "given a chance" sell as much as the ones in the bottom-right quadrant.

The sales Pareto chart (Figure 2) provides another insight into this company's approach to managing low-volume items. Note that the two products circled in Figure 1 (which have a high sales rate in outlets where the product is present, but in small quantities) appear in the "C" class of the Pareto chart. A "vicious circle" that prevents higher sales of the two products has materialized. First, the company has determined in advance that these products do not deserve to be offered everywhere. As a result, they are offered in only 14 percent of the sales outlets. Accordingly, the products' low numerical distribution outweighs their strong local sales; therefore they appear in the C-class category (products that do not deserve to be displayed everywhere), closing the loop and perpetuating the "vicious circle" situation.

This company's approach to managing low-volume items reflects the three standard arguments against deploying the full product range at all points of sale: (1) full-range deployment increases inventories; (2) buyer behavior varies among customer catchment areas (the zones in which people are likely to buy from a given shop because of its proximity to their residence or workplace); and (3) shelf space is insufficient to handle so many different products.

But these objections are not valid, as evidenced by the appliance maker's own data. To begin with, Figure 3, an analysis of inventory per SKU per outlet, shows that in many cases, it is possible to offer the entire product range without increasing inventory. In many shops, in fact, it would be possible to do so while reducing inventory. For example, Outlet 100 has many units in stock but does not offer the full range of products. Of particular note is the fact that it has 30 units of several SKUs—sufficient for many months (and possibly years) of sales. In contrast, Outlet 120 has fewer total units in stock but offers the whole range of SKUs.

As for the second objection, buyer behavior does vary from one geography to another, but it is difficult to accurately judge in advance that a given product will sell in one retail outlet and not in another. It's especially difficult to do so when the average weekly sale consists of one-tenth of a unit at an outlet that brings in hundreds or thousands of shoppers a day, all of them with different profiles. Any stocking decision will be further complicated if the sale of a particular product was an exception for the area's demographics.

Thus, while it is possible to ascertain in retrospect that a product that was believed to be valid for all retail outlets actually sold very little (the bottomright quadrant in Figure 1), at these low sales levels it is virtually impossible to determine beforehand that a product should be deployed throughout the country—the United States, in this case—but only in some cities in each region. In the example described here, thousands of supermarkets are replenished from some 50 regional warehouses located all across the United States. The two products circled in Figure 1 (strong sales when present on the shelves but offered in only 14 percent of sales outlets) are present in more than 80 percent of those warehouses. In other words, both the stocking levels in the warehouses and the sales figures are independent of region (U.S. East Coast versus West Coast, for example). It is very unlikely that this indicates actual consumption patterns at a finer level of detail. More likely, this distribution pattern represents lost sales opportunities due to misguided judgment.

The third objection—that shelf space is insufficient—is easily dismissed. It is preferable to rethink the use of shelf space in order to manage a larger number of SKUs. Experience shows that it is possible to increase the assortment on display, often by more than 20 percent, while still arranging the shelves in an orderly and attractive way. In the case of the supermarket chain, some shops (for example, PoS 120 in Figure 3) have been able to display the appliance maker's whole product range. Moreover, the products circled in Figure 1, which were displayed in about 14 percent of the stores, were deployed more often in smaller outlets than in the larger ones.

Example #2: perfumes and cosmetics. Another example of the consequences of limiting distribution of small quantities is that of a world-famous manufacturer of perfumes and cosmetics. This manufacturer selectively distributes several hundred products to dozens of department stores owned by a top Spanish retailer.

An analysis of the manufacturer's numerical distribution and daily sales for a selection of SKUs, shown in Figure 4, yields two interesting findings. First, the top-selling products sell only 0.3 units per day per sales counter, or very little relative to the number of customers in a department store. Second, the top sellers are not available for sale in 100 percent of the retailer's stores.

A closer look at Figure 4 reveals two additional observations. When a sale is made, 90 percent of the time it involves only one unit. Indeed, less than 0.1 percent of sales are for three or more units. Moreover, an analysis of inventories and sales in the manufacturer's global distribution network showed that even when products are on display and are offered to customers, very few sales are made. Out of the huge number of possible SKU combinations for this manufacturer's products, 97 percent do not result in a sale on any given day.

The experiences of these two companies exemplify the general state of affairs when it comes to product distribution. In these cases and many others, assortment management is carried out with the objective of avoiding undue increases in inventories. In practice, that approach is very likely to result in some lost sales even though inventories may still be too high. Provocative as it may sound, the normal fate of a product on a shelf on a given day at a given point of sale is to not get sold. Sales in nonfood categories occur so seldom that they may appear to be "accidents" in the life of the product on the shelf.

Detailed case study: Consumer electronics
The third example is a global manufacturer of consumer electronics that was experiencing low unit sales at points of sale. This company distributes its products and services in 30 countries through about 2,000 of its own shops and 5,000-plus independent distributors. It has industrial operations in five countries. Its fashionable products have a short lifecycle because frequent product launches make previous models obsolete. It is therefore difficult to accurately forecast sales per product per individual store. Complicating matters are long procurement cycles and a relative product shortage on the supplier market.

In the country discussed in the following case example, the company is the leader in its market. It has 600 proprietary sales outlets and operates one national warehouse that typically delivers on the second day after an order is placed. Its supply chain has achieved a very effective service level of 99 percent, with an average of only 10 days of stock at the points of sale. Decisions about local assortment and replenishment are made by a geographically dispersed, 80person team that reports to the sales department.

An important concern for this company and others with low-volume products is the shortage rate, which is rarely measured. Shortage is not simply the lack of a product; it is difficult to assess because it can:

• Have no effect; for example, when it is possible to substitute another product or the customer is willing to await the next delivery of the requested item; or
• Result in a lost sale if the customer urgently needs the product, and possibly result in a lost customer if shortages occur too frequently.

Figure 5 depicts a shortage indicator developed for the consumer electronics company. The indicator shows the daily unit sales and the inventory remaining for each SKU at the close of business each day. This indicator was chosen by the manufacturer because its definition of a shortage is simple and unambiguous. It is also conservative because it does not take into account all out-of-stocks. Indeed, it avoids such arguments as "there was no product X in the morning at PoS 1, but we do not need to worry about it because it does not meet the needs of the store's service area and would not sell at PoS 1 anyway."

In this example, the indicator reflects the fact that at PoS 1 there is a shortage of product B, and at PoS 2 there is a shortage of product C. It also shows an absence of product C at PoS 1 and of product D at PoS 2; the indicator does not include them in the shortage column even though their absence the next morning may be a problem. The practical implication is that, if there is no replenishment overnight, 3/13 of yesterday's business cannot be done today under the same conditions (the same products and locations)— a possibility that certainly merits attention.

Although the overall inventory situation for this company is considered to be excellent, according to this indicator there appear to be shortages on the PoS shelves over the course of a sample 12-day period. (See Figure 6.) In light of this result, it is interesting to detail what happens on the following day, as shown in Figure 7.

Since in two-thirds of the cases, on average, the products that experienced a shortage are not replenished the next morning, it is reasonable to conclude that two-thirds of yesterday's sales cannot be repeated the next day. This does not prove that sales will suffer, but at the very least, what sales occur the next day will not be made with the same products at the same point of sale for two-thirds of X percent of yesterday's business.

Testing a different approach
The consumer electronics company's point-of-sale experience and information served as the background for proposing a counterintuitive hypothesis:

In an environment where service levels are almost perfect and inventories are low, some sales will be lost because product availability—although not measured—is imperfect. Deploying a wider range of SKUs at the points of sale would therefore be beneficial. Moreover, centralizing control of and accountability for deployment and replenishment would prove more effective than the current mode of operation, which is based on local autonomy. Furthermore, responsibility for assortment and replenishment should be shifted from sales to the supply chain function, which is very well suited to making those decisions.

When the hypothesis was presented to the consumer electronics company with the suggestion that it adopt these recommendations, the proposal was challenged on all grounds. First, opponents argued that shortages do not lead to lost sales (except very marginally) because the width and depth of the product range is such that it is always possible to substitute a similar product. In addition, they said, most sales are assisted (that is, a salesperson is advising the potential buyer), and it is part of the sales force's responsibility to ensure the substitution.

There also was strong opposition to deploying a wider range of products at local outlets because opponents feared that inventories would increase. The proposal to centralize stocking decisions, meanwhile, encountered the objection that central decision makers could not know the particular characteristics of a customer catchment area, and therefore they could not make better decisions about assortment than the local sales staff could. Finally, the proposal that the supply chain function drive assortment and replenishment met with this objection: "Supply chain managers hardly ever meet with clients and prospects, therefore they cannot know what customers want and need."

After several weeks of discussion to overcome internal resistance, the company decided to run a pilot program to test the hypothesis. It selected a test region with 20 outlets, primarily on the basis of proximity and the fact that the project was acceptable to local management. A two-person team was assigned the task of deciding the range and quantities to be replenished at each outlet. A set of daily measurements was established, including the extent of the effect of the new replenishment method on sales.

Measuring change in sales is even more difficult than measuring shortages because it requires first establishing a baseline. This is extremely difficult to do in a fast-paced environment characterized by new products, promotions, and sales patterns that change from one day to the next. To resolve this difficulty, it was decided to measure total sales for all SKUs in the pilot area relative to total sales for all SKUs in a reference area that surrounded it, and where stocking practices did not change.

This sales uplift rate was chosen as an indicator in part because its definition is simple and unambiguous. In addition, it neutralizes all parameters that affect absolute sales (new products, promotions, holidays, and so forth), provided that these parameters are identical in both the pilot area and in the reference area. Furthermore, it cancels out any effects related to customers returning to the point of sale in the days following a shortage (because the measurement is made over time) and to "cannibalization" among products (because an increase in total sales will not reflect competition or substitutions among different SKUs in the assortment). Consequently, if the moving average increased by some percent immediately after the new assortment and replenishment practices were implemented, and the higher sales rate continued in a sustainable fashion, then the logical conclusion would be that the new practices were responsible for increasing sales by that percentage.

Figure 8 depicts the indicator and shows the sales increase as "A percent."

Because the reference area was 2.5 times larger in sales volume, the pilot area's sales were deemed to "weigh" about 40 percent in comparison to the reference area. An increase in relative weight (and therefore of sales) of around 10 percent occurred, with sales in all of the weeks during the pilot being above the average seen prior to the pilot. (See Figure 9.) The results were subsequently tested by conducting similar comparisons of the pilot area to many other reference regions during varying periods. Each follow-up comparison produced the same result: an increase in sales of approximately 10 percent.

The pilot produced an additional important benefit. From the first day of implementation of the new assortment and replenishment practices, the shortage rate was drastically reduced compared to the threeweek average prior to the pilot's start. Figure 10 quantifies this improvement.

Companywide benefits
The initial success of the pilot test led to a decision to roll out the new stocking practices to all sales outlets in phases. Phase 1 included 120 sales outlets (six regions of 20 outlets). As shown in Figure 11, these outlets experienced an increase in sales of 11 percent almost immediately.

To verify that the sales improvement was not due to a sharp increase in one region masking stagnation or decline in others, an analysis was conducted for each of the six regions involved in Phase 1. Figure 12 shows those results.

With the exception of a single region, the relative weight of each region increased in comparison to the rest of the country, significantly and sustainably. It was decided not to investigate the sales decline in Region 1 because the increase in five other areas was sufficiently conclusive. In addition, it was understood that since the new practices could not reduce sales, the reasons for the decrease in Region 1 must have been peculiar to this region and external to the test. Most likely, the decrease in sales would have been even greater under the previous assortment and replenishment rules.

Figure 13 illustrates the results of an analysis of weekly sales in regions that were not included in Phase 1. This analysis found that the relative weight of each region declined in nearly all of the weeks measured, thus confirming that the sales increase during Phase 1 was not due to a sudden, drastic drop in sales in another area. Further confirmation came during Phase 2, comprising a new set of 120 PoS, which achieved a sales increase of 7 percent.

Following these initial positive results, the consumer electronics company continued its deployment of the new assortment and replenishment practices at all of its sales outlets in the territory. A subsequent analysis found that sales across the country rose with no increase in inventories or costs. In addition, the workload required to handle assortment definition and daily replenishment for each point of sale has been greatly reduced. This responsibility is now assumed by a team of five people—far fewer than the 80 salespeople who had been spending part of their time on those tasks.

After completing the deployment of the new strategy in one country, the consumer electronics manufacturer company expanded it to other countries, each time achieving similarly positive results.

The decision process regarding the assortment deployment and the algorithms for determining the quantities to be replenished are not described in this article and will be published separately. However, the fundamental principles that led to the results described here can be summarized as follows:

• Simple rules for managing the assortment per PoS, tending toward a more comprehensive deployment of the range
• Simple calculation methods that do not include any integral or exponential, which are often encountered in the management of supply chains
• A team that is dedicated to these issues

This approach to assortment and replenishment is applicable to most product types and channels. In the past few years, in fact, it has been adopted by other companies, which have increased sales and reduced inventory in a wide variety of business conditions: diverse product categories and assortment sizes (from dozens to thousands); in direct and indirect channels; in different countries; during normal and promotional periods; whether sales are assisted by a sales force or not; and regardless of whether the change was driven by the manufacturer or the retailer. Figure 14 provides a sampling of their results.

Gaining acceptance for a new paradigm
The approach described in this article is intended to apply when the sales per SKU per outlet are small—that is, counted in units per week—which corresponds to the vast majority of products, even in hypermarkets or department stores. In such cases, trying to accurately predict the sales level per product per PoS will be futile, and therefore assortment management calls for a paradigm shift in thinking. Figure 15 compares the usual and the new ways of thinking about this subject.

Although the benefits of this new paradigm—in the form of significantly increased sales with no additional inventory or costs—have been conclusively demonstrated time after time, companies that want to adopt it may still encounter internal opposition. At the consumer electronics company, for example, centralized control of local stock deployment raised concerns about governance among the PoS managers. Because they were responsible for their stores' income statements and felt that they understood their local markets best, they believed that control over assortment and inventory belonged at their level.

However, those arguments do not justify full local autonomy over the assortment. Many other issues that have at least as much impact on profitability fall outside the control of both the PoS managers and the local sales force. These include, among others: product cost and pricing (which determine the margin); marketing plans (which affect traffic); and the global referencing of products (even if local managers are allowed to decide not to place some products on their shelves, they are almost never allowed to stock products that are unknown to the group).

Moreover, it does not make sense for the PoS managers, who already have their hands full with other responsibilities, to oversee product assortment and replenishment. Their first priority is to be sales force leaders and local human resource managers. They must deal with dissatisfied customers, ensure that all security and safety procedures are followed, and carry out many other business responsibilities. It is difficult enough to combine all of those capabilities in one person. It will be even harder to find someone who offers all of those skills and has the ability to accurately identify whether a product that will produce only small, erratic sales deserves placement on the shelf. Furthermore, in distribution networks of significant size, it would be virtually impossible to find dozens or hundreds of professionals with all of those qualifications.

However, central governance does not represent a "victory" of the central over the local. It only reflects the fact that questions regarding assortment and replenishment cannot be accurately answered at the point of sale, and in any case do not require a response that is specific to the PoS. It is better for local managers to dedicate as much time and energy as possible to the core missions of the point of sale: merchandising and building relationships with customers ... and let the supply chain function do what it does best.

Editor's note: This article is based on an academic paper published in Revue française de gestion, vol. 34, no. 186 (2008): 117-132.