«HARVESTING THE MARKET BASKET ARE YOU ABLE TO MINE THE TREASURE TROVE OF INTELLIGENCE ABOUT CONSUMER PREFERENCES AND BUYING DECISIONS BURIED IN THE ...»
HARVESTING THE MARKET BASKET
ARE YOU ABLE TO MINE THE TREASURE TROVE OF INTELLIGENCE ABOUT
CONSUMER PREFERENCES AND BUYING DECISIONS BURIED IN THE SHOPPING CART?
80 BROAD STREET, 31ST FLOOR, NEW YORK, NY 10004 • 212 405 1010 WWW.1010DATA.COM
Harvesting the Market Basket
“God is in the details”
Ludwig Mies van der Rohe, German born American architect, 1886-1969 One of the well known and frequently quoted adages in the American lexicon these days reminds us, “(t)he devil is in the details”. Generally attributed to the 19th century French novelist Gustave Flaubert, this expression suggests that even the grandest project depends on the success of the smallest components. The more positive version, above, from van der Rohe holds that success comes to those who can parlay the details into a grand design. Either way you look at it, competitive advantage is in the details. For retailing that means understanding everything possible about the sales transaction; such things as: What time of day did the customer shop? How long did it take to check out? Was their loyalty card used? Who was the cashier? How long did it take to tend? How many items--by type--were in the basket? What was the relationship among the purchased items? How do other baskets compare? The massive universe of available consumer data can be gold for merchandisers, if they are able to perform analyses quickly, easily, and inexpensively.
And that is what 1010data has developed - a powerful yet straightforward way to conduct powerful Market Basket Analysis. Let us explain...
Market Basket Analysis: A Definition The introduction of the barcode scanner ushered in an era of metrics-based decision making for retailers. It was the first data analysis revolution in retail – providing a complete and reliable source of objective sales data. Since then, the POS T-log (transaction log) has been heralded as a “goldmine” of information, but in truth, it’s not been systematically tapped into.
Traditional approaches to data warehousing and business intelligence have generally been adequately successful in meeting the needs of SKU-by-store level analysis for most retailers. Data collected from POS systems today is usually summarized and then loaded into a data warehouse. Most of us are familiar with the software and hardware technologies that have been used over the years to produce dashboards and reports to facilitate decision making using this level of data. It has been a relatively good thing, although expensive and cumbersome for most retailers. The challenge now is that the solution has not gone far enough nor fast enough for the ever data-hungry user.
Let’s define Market Basket Analysis. Simply, it is a process to understand,
• What items are being purchased by customers in their individual baskets, and over time across trips?
• What are all the metrics associated with the basket? (i.e. sales, margin, time, etc)
• What are the relationships of the items in the sale? (i.e. likely to be purchased together, or affinity) However, Market Basket Analysis is often interpreted more broadly to refer to the analysis of detailed “transactional” data, or data at the “atomic” level. Such datasets would include the most granular level of data – the “line item detail” – the contents of every basket. Hence the term “Market Basket” data since the presence of such data is what most visibly sets this kind of system apart. But according to this definition, the focus is not only on the interactions of items within the basket. Rather, it is the effort to understand all of the characteristics of a retailer’s operation that are impossible to be gleaned from summarized data, or is not commonly found in systems that store just summarized data.
A white paper from 1010data © 2008
-2Harvesting the Market Basket
At a minimum, a Market Basket Analysis system will store the following data values for each transaction:
The next most desirable attribute is typically some measure of profit for each UPC/SKU in the basket. But this only scratches the surface of analytical possibilities given the wealth of data typically collected at the POS. Every attribute
of each transaction can be loaded into the data warehouse - including, but not limited to:
• Exact time of transaction
• Duration of transaction
• Time spent on individual actions in transaction
• Net-price at item level, accounting for coupons
• Item Weight
• Tender details (cash, credit, debit, food stamps, etc)
• Reward points - awards and redemptions
• Was UPC manually entered?
• Was weight manually entered, or from scale?
This information concerning the basket can then be supplemented with additional sources of data such as:
• Demographic characteristics of individual consumer
• Geo-demographics surrounding stores
• Inventory data
• Promotional details
• Coupon billing (vendor, department, or corporate funding, etc.)
• Historical weather observations These datasets represent a vast amount of data when viewed at the transactional level. Those retailers that are able to harness its power to make more informed decisions will be successful.
Why you need it… Now more than ever, successful retailing requires a comprehensive understanding of the consumer. Today’s customer has changed and will continue to change. The changing economy, the continual onslaught of new technology, the war on terrorism and the green movement have all contributed to changing the way customers shop. Consumers are smarter, more nimble, more selective and less loyal than ever before. Accordingly, a retailer’s repository of data must switch gears from simply reporting what sold, to provide insight into who bought, when and for what reason.
A white paper from 1010data © 2008
-3Harvesting the Market Basket Retailing today also requires extremely efficient execution. Store operations, merchandising, marketing and advertising must all perform consistently with little room for error. Due to continually increasing pressure on margins in many segments of the retail industry, only the best are surviving. This all requires that retailers develop a firm understanding of their operations and maintain the ability to delve into the operational data to ask any question.
Traditional data warehousing systems were designed to handle large amounts of summarized information to address a predefined set of questions. In order to be nimble and adaptable, retailers need the ability to continually measure and probe at all aspects of their business with new questions.
What you can do with it… Following are a sampling of real-world scenarios that demonstrate the power of market basket analysis.
More Effective Pricing and Promotion Example 1 – Product Profitability A retailer sells laundry detergent at an aggressive every-day low price. Historically, they’ve been happy with the results. Sell-through is high and sales look good. They’re confident that the sacrificed margin is justified as it must be driving traffic to the stores and generating incremental sales of other items. However, upon looking more closely at the baskets that contain the laundry detergent, they begin to realize that those baskets tend to be single-SKU, or otherwise small baskets. In reality, the pricing strategy was not at all efficient, but this would have been impossible to determine without gaining visibility into the market basket.
Using this insight, the retailer decided to raise the every-day price. They expect sales of laundry detergent to drop, and they may even lose some customers. But those customers were not profitable in any case, and the improved margin on the future detergent sales will result in profits being net-positive.
Example 2 – Promotional effectiveness A mass merchandising retailer is looking to increase sales in a particular product category (A) they view as strategic. Their first approach was to engage with the principal supplier of product in this category to jointly fund a promotional program. Unfortunately, the supplier was not willing to participate. They then looked at a product affinity analysis, which shows the other categories, brands, or items most tightly associated with this category. They discover that Category B has a strong relationship with it. Buyers of Category B products tend to also buy items from Category A, more-so than the general population, and it so happens that they have a great relationship with the suppliers of Category B products! They were able to secure promotional funding from this supplier, and the “tag-along” effect generated a lift in sales in Category A as predicted.
Example 3 – Promotion expense control on key items Grocery retailers regularly fund promotions of key items to generate predictable traffic for the store. One such retailer is struggling with the burden of funding their Premium Orange Juice promotion, where they’ve traditionally offered a $2 price point, a savings of $1.50. In the past, the supplier has participated in the funding, but they’ve been forced to cut back as a result of rising manufacturing and supply chain costs. The obvious solution is to reduce the discount amount, but that would result in a very visible price hike in the eyes of the consumers, resulting in churn to other chains in the very competitive grocery market. Desperately seeking another solution, the retailer examined the quantities purchased in individual baskets. They A white paper from 1010data © 2008
-4Harvesting the Market Basket discovered that 80% of customers purchased 1-3 units, but the promotional expense incurred by those purchasing 4 or more units was very significant. In a single week, there were 150,000 units sold as the fourth or greater unit in a basket – resulting in $225,000 of additional expense. A “limit” of 3 was placed on the number of items that may be purchased at the sale price. Customer penetration was substantially the same as the previous promotions, but the expense was controlled, without raising the price point for the consumer.
Their success with this particular product motivated the retailer to implement limits on many other promotions. And with market basket insight, they are able to easily determine the optimal limit for each product.
More Intelligent Merchandising and Marketing
Example 4 – Actuals more insightful than averages A discount retailer always knew that their average basket was $8. This metric was easily measured and tracked using summarized sales numbers (by computing dollar sales divided by transactions). However, when they dug deeper into their detailed data, they discovered that there were very few $8 baskets! In reality, there was a large representation of two very different types of shopping trips – those less than $4, and greater than $15. The “average” just served to obscure a good understanding of the nature of the typical shopping trips. Armed with a detailed understanding of which types of trips are prevalent in which stores, they are able to better strategize on merchandising and marketing.
Example 5 – Single SKU location analytics and profitability Upon further analysis of the $4-or-less baskets, the retailer determined that there are certain products that are very often found in single-SKU baskets. Using the rationale that customers must be coming into the store specifically for these items, they were moved further into the store, forcing the customers to pass through many other product categories resulting in increased basket sizes from “impulse” items and other merchandise.
Example 6 – Planogram insights A large discount retailer noticed that sales of a particular category were more variable than expected from store to store. A closer examination revealed that the difference in category sales were somewhat correlated to a difference in store layout. The chain maintained stores with different types of planograms, and one of those planograms appeared to benefit this particular category. To fully understand what in the planogram could be causing this, they performed a product affinity analysis for the category, which shows the products often purchased together with the target product. It turns out that in the store with higher category sales, many of those sales occurred in baskets that also contained another associated product, which was physically located in the same vicinity.
Example 7 – Assortment effectiveness During a recent assortment planning effort, category managers were re-allocating space to make room for new product introductions in a particular category. At the time, the bulk of shelf space was assigned to Brands A, B and C – each receiving roughly the same amount of space. The traditional approach is to examine the sales volume of each brand to determine the best candidate for a reduction in shelf space or to be removed entirely from the assortment. However, that doesn’t tell the whole story. By looking at the detailed activity of shoppers over time, they determined that shoppers who buy Brand A are loyal to that premium brand – they rarely switch to another. On the other hand, Brands B and C experience a high degree of “switching” between themselves, and with other brands. Consequently, they decided to reduce the space A white paper from 1010data © 2008
-5Harvesting the Market Basket assigned to B and C, with the logic that reduced visibility of those brands, or an increased risk of out-ofstocks, would not have a negative impact as would be the case with Brand A.
Similarly, such insight is extremely valuable to the product manufacturers/vendors, who stand to develop a better understanding of the brands they are really competing against.