Expert Column: The Science of Retail

Grocery retailers have lots of historical transaction data. Lots of it.  Tens of millions or even billions of data rows provide a historical record of every item that’s been purchased, when, from which store and in what combinations. Retailers with loyalty programs can augment this transaction data with information on who is buying, both on an individual and an aggregate basis.

Retailers continue to collect this data and many have made good use of it, segmenting and targeting customers and rewarding loyal behavior with discounts and offers. Still, many sense that there’s still untapped potential. They’re right. With the costs of data storage plummeting and the capabilities of analytical tools on the rise, this data’s value is set to skyrocket.

Properly used, insights from these vast data storehouses can scientifically inform retailers’ decision-making in critical strategic, tactical and operational areas, including category management, shelf space allocation and new product introductions. For the supply chain, the data can be used to optimize replenishment schedules at the store and DC levels, even for multi-echelon networks containing specialty DCs. It can help set automated business rules governing acceptable in-stock levels for categories, products, and even individual SKUs. Of greatest interest to the bottom line, retailers can more accurately determine how much safety stock to hold, and where it’s most optimal to hold it –- all while keeping overall inventories as lean as possible.

By embedding an advanced science engine and space optimization technology in a new generation of grocery and supermarket business solutions, retailers are turning these mountains of aggregated transaction data into a true-life picture of customer demand.  The result yields more accurate answers to basic questions:

• Which items must I always keep in stock? These are the products that if customers can’t find them, they will walk away without buying anything in the category – or they may just walk out of the store, period. For some items, if this scenario recurs too often the retailer will lose not just this sale but all future sales.
• Which items exhibit high levels of demand transference? For example, if the large-size laundry detergent container isn’t in stock, many customers will transfer their demand to smaller containers of the same brand, or buy a large-size container of an equivalent competing brand.
• Which items are “must-haves” for my most important/profitable customers? Grocers may be tempted to drop some slow-moving items when category review time comes around, but if the product is routinely purchased by high-spending, high-value customers, it should stay on the shelf. Even though the product’s profit margin is low, its contribution to the store’s overall profitability is high.

Many layers of complexity underlie these basic questions. Customers will exhibit different shopping patterns based on different trip objectives, from the weekly pantry stock-up to the last-minute fill-in before a party. Factors such as the store’s location (suburban/urban/rural) and the demographics of its customers all play their part. Some customers will only buy organic products no matter how high the price is, while others are almost solely motivated by price. Even within the same customer segment, behavior can differ markedly from product category to product category: the toilet paper bargain-hunter may pay top dollar for locally raised, grass-fed beef.

Analytics Solutions

New retail analytics solutions take these complexities into account, providing both the analytical insights and the structural tools needed to incorporate insights into day-to-day business decisions.  For example, an embedded Consumer Decision Tree (CDT) tool will precisely measure a product’s demand profile, allowing retailers to set more relevant service levels.  Must-have products may require a 98 percent in-stock level, but others can fall to 75 percent without causing undue harm to sales or customer relationships. And after retailers learn where they should place their inventory for maximum benefit, replenishment optimization solutions ensure automated replenishment scheduling leverages these insights on a store-by-store basis.

Many of these issues are basic to all types of retailing. The prototypical “corner grocer” of 100 years ago could have acted on them based on his own observations of customer behavior – when stores carried far fewer items and retailers had far fewer customers. What’s needed today is the ability to reproduce this level of intimacy, but on an industrial scale. The latest generation of retail science makes this not only possible, but practical.