PDF File: 

CRM6_wp_merl.pdf

Accenture

By utilizing techniques from attribute science, retailers can understand why consumers make particular purchasing choices anddevelop the best product assortment to match those purchase drivers, thereby generating stronger sales and profits.

Success in retail has always hinged on knowing one’s customer. Whether it’s the local butcher catering to the little old lady down the street or the big-box merchant meeting the needs of an entire neighborhood, retailers continually strive to gain the greatest insights into what their customers want and provide a product mix that best satisfies those preferences.

In recent years, merchants have taken ever-greater strides to that end. Loyalty cards have enabled retailers to collect shopper-specific data as never before, and technology solutions such as data warehouses and data mining software have made it much easier for retailers to make sense of that data – to understand, for instance, which products or brands customers repeatedly purchase and those which engender little loyalty.

Yet despite these and other advancements, retailers still find themselves facing a major challenge. Ever-increasing availability of store locations and types – all competing for a share of consumers’ requirements – continues to drive greater competition, as well as foster the blurring of traditional channels and consumers’ buying attitudes, habits and patterns. Consumers have more options than ever in choosing where to shop, and can tap multiple channels for their needs. In such a highly competitive and mature environment, it is more important than ever for retailers to know who their customers are.Without such an understanding, retailers and their manufacturing partners, even those with well-known, longstanding brands, may find themselves on the losing end of the customer proposition and ultimately face a dramatic shrinking in size or cessation of business altogether.

An approach to consumer insights called “attribute science” can help retailers more effectively understand what drives consumers’ purchasing decisions. More specifically, using attribute science, merchants can predict how consumers would react to various product assortments that are different from those that they have been presented with previously, and subsequently identify the optimum mix of products from the universe of goods that could be sold most profitably at retail. In the process, companies can stimulate greater top-line growth while improving margins by reducing or eliminating unproductive items in their product portfolio.

The Dilemma of Choice

One of retailers’ primary responsibilities is deciding which of the hundreds of thousands of available consumer products to present to customers of particular stores. This is truly a daunting task, especially given the speed with which consumer preferences change and new products are developed.

Most retailers, lacking in-depth knowledge of what customers really want, are forced to essentially take educated guesses when filtering the universe of available products down to a manageable number for their stores. And because their insights into consumers are limited, merchants tend to overcompensate in their selection process by trying to offer customers everything they can in the hope that customers find something they like. The problems inherent in this approach are obvious. At any given time, 20 to 40 percent of a retailer’s assortment is either redundant or simply undesirable to the store’s customers. Unfortunately, typical approaches to analyzing customer purchase data are unable to precisely identify which items are included in that 20 to 40 percent, meaning valuable store shelf space is wasted on unproductive items. Complicating matters is the fact that many consumers buy products simply because the products are there. What would happen if a particular product was removed? Most retailers would have no way of knowing (see Figure 1).

Consider a simple example of a supermarket’s cola sales. The retailer knows that it probably could present a more optimal mix of items in the cola category to meet customer demands. It regularly studies POS data to determine how each product in its cola category sells, and continually tweaks its assortment to reflect perceived consumer demand. But the retailer doesn’t know how customers would respond if it were to, say, stop selling cola in 12-ounce cans. Would 12-ounce-can customers simply switch their purchases to 2-liter bottles? Buy 12-ounce cans of a different type of soda? Purchase orange juice instead? Stop shopping at that particular store entirely? To find out, the grocer theoretically could make the change and then monitor the results. But that would mean risking sales and customer goodwill, something that no retailer is in a position to do, given the industry’s razor-thin margins. Besides, even if a company were willing to take such a chance to experiment, it would be impossible to do so for every conceivable item mix for every possible SKU in its store. If it tried, the merchant would find itself in a continual state of trial and error, which is no way to run a business.

Applying Attributes to Assortments

While retailers have become adept at knowing which products sell (and when and in what volume), they typically don’t understand why. And that’s at the root of the assortment challenge. If a merchant could identify the specific factors that drive consumers’ purchase decisions, and tailor store assortments accordingly, it could build stronger customer loyalty, reduce or eliminate unproductive items or categories, and increase sales, profits or both.

For example, consider two brands of pancake syrup, Brand A and Brand B. Both brands sell at about the same volume, yet Brand A has 30 percent higher margin than Brand B. Because of their comparable sales volume, the retailer assumes that both brands are important to consumers, so it believes it must carry both. Yet because the merchant doesn’t know why people buy both brands, it can’t know that 90 percent of both brands’ sales are strictly due to chance, as neither Brand A nor Brand B carries any special meaning to a large majority of shoppers. Armed with such knowledge, the retailer could feel confident in dropping the lowmargin Brand B and still retain 90 percent of its sales in the category. What it loses in sales from Brand B-loyal customers it more than makes up for in the elimination of the considerable costs associated with carrying a second brand of syrup.

Key to such an understanding is attributes, the specific characteristics of products and how they are merchandised that elicit some kind of emotional, behavioral or attitudinal response from consumers (see Figure 2). An attribute can be a measure of a product’s performance (a laundry detergent’s ability to get clothes cleaner than others), it can evoke a certain feeling (a cereal that brings back happy childhood memories) or it can even relate to how a product is displayed in a store (it’s more attractive to consumers in an end-cap than on the shelf).

Traditional market and consumer research has shed deep insight into how individuals react to specific attributes. In fact, such information has been available for quite some time. However, few retailers have capitalized on attribute science to optimize their assortments, primarily because they lacked the capability to manipulate attribute data in a way that would help them make assortment decisions.

Today, though, several leading merchants have employed a new approach to attributes, underpinned by an innovative technology solution, which enables them to effectively model how changes in attribute mixes will affect consumers’ purchase decisions and, ultimately, the sales and profitability of product categories. Using this approach, companies have been able to boost sales of a particular category, remove items with minimal revenue loss, and increase store traffic and overall market basket size of a particular customer segment.

The approach begins with an identification of the customers being targeted, whether it’s a specific demographic, geographic area or simply the entire pool of customers patronizing a particular store, and the product category being modeled. This segmentation can be done at a relatively simple level, such as urban or rural, large or small or high or low intensity. These basic segmentations can generate significant preliminary value. Ultimately, segmentation by shopper behavior supplied by named loyalty card data is the most difficult but highest added-value segmentation.

Next is assigning the known attributes of the product (let’s continue with the cola example) to representative items in the category that customers of the store have purchased for some time. By analyzing sales history of each item in the category, a company can determine the relative importance of each attribute (brand, flavor, package size, etc.). This analysis results in a picture of customers’ historic responses to the attributes associated with the cola products carried in that store (say, Brand A cola with caffeine in 12-ounce cans outsells all other combinations by a 2-1 ration). Additional analysis of this data enables the retailer to derive a specific dollar value for each attribute. For example, if an item is responsible for $10 of sales per day and its brand attribute is 75 percent of its importance to consumers’ purchase decision, then $7.50 of that product’s volume per day is directly associated with its brand. A number of techniques are available for weighting attributes, including team judgment at the basic level, usage and attitude research at the next level, and econometrically derived coefficients at the most advanced level.

The final step involves using a search algorithm to model how the deletion and addition of specific products, including those that could be sold at that store, but aren’t currently, will affect sales of the category as a whole, based on the uniqueness of each product’s attributes relative to others in the category and dollar value associated with each of those attributes. Models may show that a new product with a unique attribute that’s highly valued by the customer segment being studied would likely bring additional volume to the category if added to the assortment. Similarly, the deletion of an item that carries the last remaining representation of a particular attribute would likely result in a loss of sales for the category. But if all attributes of a particular item are represented in all the remaining items in the category, modeling would reveal that a redundant item could be removed with no negative impact on sales because its volume would be assumed by the rest of the products in the category.

Let’s look at how this might play out in the cola category. A retailer has two brands of cola, each with and without caffeine, offered in five different package options (12-ounce cans, 2-liter bottles, etc.). If the model revealed that the package size attribute was essentially irrelevant to customers of that store, the retailer could assume that removing one packaging option, say, 2-liter bottles, would have little to no impact on overall sales of cola because the valued attributes of cola flavor, brands and caffeine or no caffeine were still available. However, if packaging were important, and one option, 12-ounce cans, were removed, the chances are good that shoppers seeking 12-ounce cans of cola would satisfy their needs at another store that does carry that packaging attribute, resulting in an overall loss of cola sales for the retailer.

Attribute Modeling in Action

The approach described is far from theoretical. In fact, it has been used by a number of innovative retailers to improve their assortments and, ultimately, business performance.

One such company is a retailer of office products, which faced increasing space constraints. The company wanted to make better use of its high-traffic, in-line store space by opening it up for category expansions, emerging offerings and high-impact promotions. It was also looking to improve the sales and margin of existing categories and determine the category/space combinations that would yield the highest performance.

Through attribute modeling, the retailer was able to pinpoint and remove 97 of 547 redundant SKUs from its office supplies category (five of which were in the top 25 in terms of category volume), as well as identify specific product attributes and new SKUs that would provide incremental sales and margin. Ultimately, the retailer was able to free up four feet of high-value space in more than 1,000 stores – space that the company reallocated to the fast-growing, highly profitable technology segment. The result was a 5.75 percent increase in sales and an 8.35 percent jump in contribution to margin.

Few would dispute the fact that there’s a tremendous need for all companies to be customer-centric. This is especially true for retailers. Retail supply chain costs continue to escalate, which makes efficiency a vital concern for retail executives. Strong competitors – and more of them – are everywhere, making product proliferation and channel blurring a huge challenge in the fight for consumers’ attention and dollars. On top of all that is the fact that overall consumer demand is growing very slowly, if at all.

In such an environment, retailers can no longer afford to overassort in an attempt to keep sales strong, because the accompanying cost implications make that approach untenable. Conversely, merchants can’t risk costly assortment mistakes by blindly tinkering with their product portfolios in an attempt to pare their assortments to a manageable size. There’s simply no room for error.

By adopting an approach that enables retailers to understand why consumers buy what they do, and subsequently presenting the optimal assortment mix to match up with those purchase drivers, merchants will be able to rise above the challenges that make retailing in the 21st century so difficult. Although such an approach won’t necessarily provide mass merchants with the perfect customer knowledge enjoyed by the local butcher shop, it will go a long way toward helping retailers more precisely meet consumers’ needs, and ultimately, generating stronger sales and profits.