Retailers’ dirty little secret

Inventory discrepancies can cost retailers money in lost sales.

You might say that Adam Mersereau’s research into retail inventory management started when he went to pick up a rug at Pottery Barn.

Before making the half-hour drive to the store, he called to make sure the rug was in stock. The clerk typed the SKU number into the computer, which showed three rugs were available. But once Mersereau got to the store, there were no rugs. The clerk had checked the computer’s inventory data, but no one had checked the racks.

The discrepancy between a store’s computerized inventory and what the store actually has in stock is astoundingly large. One study showed that the computer inventory at a prominent retailer was wrong 65 percent of the time.

Mersereau buckled down to help retailers with their dirty little secret: Store managers don’t know what they have on their shelves.

If you’re a retailer, it’s your job to know what’s going on in your store,” said Mersereau, an operations professor at UNC Kenan-Flagler. “It’s understandable that retailers don’t publicize the problem, and academics have largely ignored it, too.”

Inventory discrepancies can cost retailers money in lost sales or the added expense of keeping too much inventory in stock. A big box store, department store or grocery store might have tens of thousands of items for sale. To keep up with inventory, retailers rely on computers to tell them when to reorder an item. Managers input how many of each item they order, and the computerized cash register subtracts each item sold.

But many factors can throw off that simple process:

• Items lost through shoplifting won’t show up on the computer
• Damage takes a toll on the count. If a customer drops a jar of mayonnaise, the person who cleans it up might not enter the loss in the computer.
• Misplacement contributes to the problem. If a child grabs a bag of cookies and her dad doesn’t notice it until the detergent aisle, he might appropriate it and set it on the nearest shelf. The computer still shows the item for sale, but because it is not where cookie shoppers would look for it, it is not in stock for all practical purposes.
• Minor errors accumulate over time. When a customer buys a 12-pack of Coke and a 12-pack of Diet Coke, the cashier might scan the Coke’s barcode twice if both items cost the same. But the computer inventory then registers one fewer Coke than is on the shelf and does not note that the Diet Coke stock has been depleted by one.

Store managers brushed off the inventory discrepancies as an annoyance, mainly because there wasn’t anything they could do about it cost-effectively.

But in some instances, the discrepancy cost them in lost sales. If the computer registers that there are 10 garden rakes in stock but in reality there are none, customers who come in to buy a garden order more because it thinks there are still 10 in stock. The store has lost those sales through inventory “freezing.”

While the research literature is rife with instructions about how store managers should place orders and policies they should use to keep inventory on the shelves, nearly all of it assumed that the store managers knew what they had in stock.

Mersereau used his expertise in statistics and optimization to attack the problem. He develops algorithms that can be turned into software programs to fix problems. “Our research shows how inventory management changes when store managers don’t know what’s on their shelves,” he said.

Working with data from a national retail chain, Mersereau used Bayesian updating to show a probability distribution of the item being in stock rather than showing a single number. For instance, rather than say there are five widgets in stock, the Bayesian inventory record might say there is a 50 percent chance of having five widgets on the shelf, a 20 percent chance of having four and a 30 percent chance of having six.

He designed a system intelligent enough to shift those probabilities up or down daily to factor in sales and new shipments. The probabilities evolve over time by monitoring sales activity and adjusting. For instance, if a store normally sells 10 frozen pizzas every Friday but this Friday sells none, the system gives more weight to the probability that there are no pizzas in stock.

By integrating the information conveyed by replenishments and sales observations, the retailer can better understand what’s on the shelf. The store manager can then better match supply with demand, which translates into both better availability for customers and fewer gluts of excess merchandise. “This has the potential to increase both sales and margins,” Mersereau said.

Mersereau’s system makes two decisions every day: how much to order and what to count. By focusing their counting efforts on the highest impact items, store workers can make more efficient use of their time.

“This information is useful to store managers because it saves time and results in more accurate inventory,” Mersereau said.

Retailers’ solutions to inventory management problems fall into three categories.

1. Prevention: Find and fix the root cause. For instance, if the root cause of inventory discrepancies is shoplifting, beef up security.
2. Correction: Fix a problem after it has happened, for example, through auditing policies.
3. Integration: Admit there is a problem, and use inventory planning and decision tools robust enough to account for inaccuracy.

Mersereau’s program focuses chiefly on integration and correction, but he is quick to point out that all three categories are important. “Our system does not eliminate the retailer’s incentive to remove
the causes of errors,” he said.

“The research raises awareness of record inaccuracy and provides tools for retailers to balance cost with customer service even when their inventory records aren’t perfect,” he said. “By recognizing the problem and admitting they don’t know exactly what they have on the shelf, store managers can recoup a lot of the
cost of these inventory errors.”