Can AI help retailers turn inventory into sales?

AI-assisted inventory rebalancing can cut transfer-planning time by as much as 90%.

Retailers are turning to artificial intelligence (AI) to close the gap between recorded inventory and products customers can actually buy, potentially lifting sales by as much as 1%, according to a September report by Kearney.

“Whether the product is in the backroom, misplaced, incorrectly reflected in the system, reserved against another order, or unavailable for fulfillment, the outcome is the same: the customer can’t buy it,” according to the report.

A 1- to 2-percentage-point improvement in product availability could generate a 0.5% to 1% increase in sales, depending on the product category and how easily shoppers can find substitutes, Kearney said.

Products recorded as available may be misplaced, damaged, reserved, lost, or otherwise inaccessible. This creates what Kearney calls “phantom inventory,” where stock appears in a retailer’s system but cannot be sold.

The problem can also distort future orders. When products cannot be found, sales decline and forecasting systems may interpret the drop as weaker demand, prompting retailers to order less stock.

The effects differ by product. Footwear retailers may have the wrong size or colour, whilst beauty stores may lack popular shades despite having sufficient overall stock. Electronics retailers may have a device but not a required accessory, whilst pharmacies may lack the required strength or pack size.

Artificial intelligence (AI) can detect such problems, identify likely causes, and help retailers decide where to act.

Forecasting systems can adjust demand estimates as purchasing patterns, promotions, weather, and other factors change. Kearney estimates AI can improve forecast accuracy by 10% to 20% in categories where demand fluctuates significantly.

Better forecasting and restocking could reduce instances when products are unavailable by 5% to 15% without increasing inventory, it added.

AI can also detect delayed deliveries, ordering problems, and potential shortages. Automated detection and prioritisation could reduce the amount of time planners spend reviewing problems by as much as 50%, whilst AI-assisted inventory rebalancing could cut transfer-planning time by as much as 90%, Kearney said.

The technology introduces its own risks. Computer vision systems can generate false alerts because of glare, packaging changes, blocked products or camera angles.

More alerts can also overwhelm store and supply-chain employees if problems are detected faster than workers can resolve them.

Retailers could also send too much stock to stores, replacing lost sales with excess inventory and markdowns.

Kearney recommends verification, minimum confidence levels, and human oversight to reduce those risks.
Retailers should first use AI in product categories with predictable demand, frequent availability problems, and standardised products, according to the report.

Accurate product information, inventory records, store stock data, and shelf plans are also needed before retailers expand AI forecasting, allocation, and restocking.

“As AI-enabled forecasting, sensing, and replenishment mature, retailers that build this capability early will protect more sales, improve customer trust, and operate with greater control,” Kearney said.

Questions to ponder:

• Can AI improve product availability enough to produce measurable sales gains?
• How can retailers prevent AI alerts from overwhelming store and supply-chain teams?
• Could efforts to prevent shortages leave retailers with too much inventory?