Even with today’s sophisticated inventory management systems, retailers are woefully unaware of just how low on-shelf availability is for many of the products they carry. Our research, carried out in collaboration with a major consumer goods manufacturer, suggests the problem is significantly worse and costlier than many retailers assume.
The culprit is “phantom inventory”—goods that show up in management systems as available but in fact are hidden from view because they’ve been misplaced, often tucked away in a backroom and forgotten.
Phantom inventory is extremely costly and remains a serious hurdle to efforts by retailers to keep their stocks lean while ensuring that the right goods are available on shelves for consumers.
The impact runs from lost sales when customers can’t find what they’re looking for to higher operating costs owing to time wasted when employees have to respond to queries about missing items—an even bigger problem during peak shopping periods when retailers attract most of their sales.
Most measurements of known stock-out levels give a misleading impression of how a store or product is performing. For instance, our research showed that for a category of laundry detergents sold by a large retailer, lost sales were almost five times greater than previously assumed owing to unobserved stock-outs.
Inaccurate inventory readings also have a ripple impact across supply chains, leading to problematic demand forecasts because systems may show products as in stock but unsold when in fact they haven’t made it to store shelves at all. Such faulty readings mount if the problems are repeated across many stores, triggering flawed sales reports that affect forecasts, production planning, measurement of store performance and automatic replenishment.
Many things can cause phantom inventory.
Inventory records won’t register that certain items may have been stolen, for instance. Scanning errors at the checkout counter give false readings of what product has been sold. Inventory sent to the store for a special promotion is often marooned in the back room because store managers did not run the promotion or stocked the products in a haphazard way. Product misplaced by customers and unexpected peaks in demand can also result in inventory management systems giving a false picture of stock availability.
Retailers deploy various measures to prevent the problems, but the flaws persist and can eat into retailers’ profit margins.
One solution that has received scant attention but proved effective when implemented is to develop special analytics using machine learning technology. The methods re-create the demand patterns for individual products, and incorporate the inventory uncertainty for each stock-keeping unit into forecasts and plans.
This method uses existing data—important because emerging solutions usually require retailers to invest in getting additional information – and quantifies lost sales. In addition, the analytics are fast, and can be scaled to enterprise-size applications.
When used in concert with existing solutions, this approach improves forecast accuracy and increases sales by dramatically reducing the number of stock-out events. Our research showed that the improvements in inventory planning alone could boost sales of that laundry detergent category by almost 5% while reducing inventory in the system.
Retailers and manufactures do not have to be victims of the phantom inventory menace. By paying special attention to retail operations and employing modern data analytics in combination with existing measures, phantom inventory can be a ghost of Christmas past.
This post first appeared as a Guest Voices column in the Wall Street Journal’s Logistics Report. The article was written by Fredrik Eng Larsson, assistant professor at the Stockholm Business School (email@example.com), Daniel W. Steeneck, an assistant professor at the Air Force Institute of Technology (firstname.lastname@example.org), and James B. Rice Jr., Deputy Director at the MIT Center for Transportation & Logistics (jrice@MIT.edu).