Increase Profitability by Managing Local Market Assortments

On July 1, 2008, Starbucks Corporation announced the closing of 600 U.S. stores due to a substantial downturn in store traffic. On the same day, Home Depot's stock price hit its lowest point in more than five years after a leading analyst firm downgraded the company's stock, citing an "unprecedented decline" in the U.S. housing market. The previous week, JCPenney CEO Mike Ullman announced plans to cut store openings and renovations in 2009, consistent with continuing expectations of a challenging year for the American consumer.

In the current economic downturn, retailers have been hit particularly hard: the Standard & Poor's Retail Index has declined by more than 30 percent over the past year. Yet leading retailers almost stubbornly continue to innovate and explore new opportunities that will position them to compete even more strongly as the economy recovers. Innovations in pricing, sourcing, inventory management and assortment management are being funded with scarce capital resources. And for good reason: History shows that sector leaders can win disproportionate advantage in times of adversity.

Most retail executives would agree that continuing success is predicated on profitable top-line growth. Once-great retailers Montgomery Ward and the F.W. Woolworth Company—two of the largest retailers in the United States and world, respectively—saw sales decline as competitors drew customers away, even while the market as a whole saw solid growth. Reasons for their decline—likely to be repeated by some of today's industry leaders—include unanswered competitor success in pricing, logistics, merchandising, access, presentation, service and overall shopping experience. These reasons, and others, follow a central theme: consistently failing to meet customers' changing product and service expectations.

Rx for organizational stagnation

A localized assortment strategy is perhaps the most effective way to fend off this sort of organizational stagnation. Local market assortments improve sales and margin performance by enabling merchandising decisions that attract and retain high-value customers, recognizing that these customers form heterogeneous groups that vary depending on product, place and time.

As early as 2003, Best Buy popularized the term "customer centricity" through an initiative to realign stores with their most profitable customers. Similarly, Wal-Mart's "Store of the Community" initiative was developed in 2006 to meet the shopping needs of local customers, with stores specifically tailored to reflect the demographic makeup and customer preferences of their respective communities. The benefits are real, and more and more major retailers are taking note. In May 2008, The Wall Street Journal reported that Macy's CEO Terry Lundgren had given merchants across the country the authority to tailor about 15 percent of store assortments to reflect local market preferences. Ross Stores, a leading off-price fashion retail chain, has developed a similar micro-merchandising focus (Ross Stores 2007 Annual Report, p. 4).

All Retailing is Local

Retailing is a local activity in as far as each transaction is a successful encounter with an individual customer, but the fact remains that national chains dominate all segments of retail. It is ironic that their success, driven in large part by the advantages of scale on the supply side, is limited by the increasing complexity in defining the customers they already serve or are trying to reach. Every effort to meet customers' expectations for value is hamstrung by the complexity in accurately identifying those customers in the first place, and then interacting with them appropriately. Adding to the irony, efforts to serve those customers—if they are a heterogeneous group—might lead to a corresponding segmentation of supply, undoing (at least in part) the very economies of scale that are key to the value those customers seek. Nonetheless, the tradeoffs cannot be understood until those customer segments are understood.

The Retail Segmentation Challenge

The retail segmentation problem is multidimensional: customer preferences vary not only by who they are, but also where they shop and when, and what they are shopping for, as illustrated in Figure 1.

While common-sense customer segment classifications seem a reasonable place to start, they tend to be too broad, highly subjective, and quickly lose relevance unless continually validated using local intelligence. Consider, for example, the description of a group as "Hispanic" or "Latino" (and some would draw important distinctions between these descriptions themselves). Can people of Mexican origin be grouped with those of Cuban origin? Do age, income and marital status matter? What about home ownership? And would a merchant based in New York or Chicago be able to correctly account for these differences? While efforts to segment customers into behavioral and socioeconomic groups continue to evolve, one fact is clear at a glance: the complexity of mass markets is growing. San Diego-based demographics company Nielsen Claritas now identifies 66 distinct consumer clusters in the United States. Changing economic fortunes, migration and immigration, assimilation patterns, aging, new technologies and a host of other factors have contributed to the number of segments and the variation within them.

Consider next the notion of seasonality and place, or geography. In most cases, seasons are mapped to months on a calendar, recognizing important holidays, religious, cultural and sporting events, school calendars, and so on. Place might indicate geography in a broad sense, say "northern" or "southern," or it might describe a market on a finer geographic scale, for example, whether the market is urban or rural, and perhaps the nature of the competitive environment. Finally, place, at its most granular level, is the store shelf.

Naturally, segments will vary by product category, perhaps even by specific product. Every retailer uses some hierarchical structure to categorize products—for example, by divisions that include departments, which in turn contain product classes and so on—but any notion of product classification biases the structure of the resulting segments to some degree. Nevertheless, it is a common and reasonable practice to adopt a product classification structure that reflects some compromise between a customer's internal notions of classification, and how product is managed by the retail organization.

The purpose of segmentation is, of course, to partition a chosen universe of, say, customer, product, location and time combinations, into meaningful segments. To be useful, segment profiles should accurately reflect member profiles. Moreover, each segment should differ sufficiently from all other segments to warrant partitioning. And finally (and this is by no means easy to do) each segment should have a common-sense interpretation that is both meaningful and pertinent to the business purpose.

Characterizing Customer Segments

Most commonly, retailers use anonymous point-of-sale data as a proxy for the target customer, simply inferring alignment with target customer segments based on category or product sales and margin performance. Simple store grading is the most basic such scheme: Assortments are built around a central core offering, with extensions to the core justified by above-average sales or margin performance as shown in Figure 2.

Another more sophisticated approach based on this idea is illustrated in Figure 3. As the merchant considers each product—and the selections across a category as a whole—he relies on profile matching to infer target customer product preferences based on historical segment sales for similar products.

An even more advanced approach to providing merchants with an understanding of a market entails building profiles for each customer segment. In broad terms, this means stepping back to first answer the question: Whom do we serve (or wish to serve) to grow market share profitably? The next question logically proceeds from the first: How do we align the organization to serve these targeted customers with the right products and services, ensuring access to them at the right time and price?

Now the challenge lies in identifying who current or prospective customers are, and then once identified, whether they are understood in terms that will help the merchant decide what each segment is likely to buy. The result is shown in Figure 4. Examining customer loyalty program data combined with syndicated market research data provides key insights that tie market and customer profiles together in useful ways. This is how a planner at Best Buy knows the proportion of, for example, tech enthusiasts to suburban moms, by market and by store. Â

The field continues to evolve, and new approaches to developing consumer insights focus on automating direct customer feedback to make product recommendations, as Amazon and Netflix do. Leveraging the growth and increasing relevancy of social networking might add new richness to localization solutions with little added algorithmic complexity.

The Data Challenge

There are, of course, many challenges to successful segmentation. The most significant of these almost always pertain to data. Availability, relevance, freshness, quality and completeness are all important attributes of good data; the "garbage in, garbage out" principle applies with certainty.

Retail data volumes are so vast that it is desirable and often even necessary to summarize or classify data in many different ways. Similarly, selling conditions can be grouped according to patterns—such as seasonality—enabling application in even abstract, irregular conditions such as significant public events. However, these classifications might be based on incongruent merchant and customer perceptions of product and market similarities and differences. The merchant's only defense is to continually validate assumptions using a wide range of data sources within and outside the organization.

Once the data has been cleansed and collated, segmentation is itself the litmus test for relevance. Although this might seem strange initially, it should be no surprise that many attributes of products, customers, stores and time play little or no role in what makes a segment distinct, unique or even relevant. Also, the discovery of the most important distinguishing features of a segment is no guarantee that the segment is describable in any useful way. Moreover, "over-fitting" the data—tracking noise rather than information—is a common problem. The solution is simply to avoid questionable detail. Getting the segments right underpins the effectiveness and manageability of the subsequent merchandising process.

Finally, it is worth noting that historical data provide intelligence and clarity—in hindsight. Game-changing events can overwhelm detectable trends, and there is no substitute for human intelligence in recognizing sea change and anticipating its effects. Just as Netflix changed how customers rent movies, and Amazon has redefined how customers access content, new retail paradigms will emerge to challenge current notions of value.

Few retailers would argue against the value of customer intimacy, no matter what the market conditions. This is evident today, even with the challenges of the current economic downturn. America's most admired retailers continue to invest in initiatives to understand and cater to their most profitable customers through localized assortment strategies. Simple store grading processes are being enriched by target market and customer attributes, often using traits that are managed locally.

More advanced techniques rely on customer loyalty program data, augmented by market research data, to drive assortments targeted not only to maintain and grow historically important customer segments, but also to identify and cultivate new segments. And finally, the growing relevance of online social networking sites has raised the intriguing prospect of collaborative micro-merchandising: customers engaging directly with merchants to identify new value propositions in consumer product and service offerings.

— by Ijaz Parpia and Sai Buddhavarapu