Halloween has yet to arrive, but candy and costume companies are already preparing for the season in summer. Mars, whose brands include M&M's, Twix and Skittles, is seeing consumers begin buying Halloween products as early as July, forcing manufacturers and retailers to rethink the traditional selling window. Ramesh Kollepara, senior vice president of Mars' snacking business and global chief technology officer, said companies must forecast not only how much demand will emerge, but also when it will begin, where it will occur and whether consumers' preferred product mix is changing.
Earlier Buying Narrows the Halloween Replenishment Window
Halloween is a seasonal market with a firm deadline for candy and costume supply chains. Products must reach stores before the holiday, while excess inventory can leave retailers with waste after the season ends. Benjamin Bond, senior vice president of customer strategy and success at Simbe, said retailers typically have only about a 10-day window around peak demand to replenish shelves, put products on display and rebalance inventory.
Madhav Durbha, vice president of the industry strategy group for manufacturing at RELEX Solutions, said supply planners once relied heavily on manual parameter changes, making forecasting a process driven in part by trial and error. AI can now update those settings automatically, reducing the repeated work required to adjust forecasting algorithms.
Mars Matches Regional Demand With Inventory
Mars is using AI to track changes in Halloween candy demand and bring early purchasing trends, existing inventory, production capacity and customer requirements into a single planning system. Kollepara said the system links production and distribution around four questions: what consumers want to buy, what Mars should produce, where inventory should be placed and when consumers will need it.
Mars' internal research shows that consumers in the western United States have a stronger preference for specialty Halloween-themed candy, while those in the South favor gummies. The company can use that information to direct inventory by region instead of stocking every part of the country with the same product mix. That approach affects not only whether seasonal goods reach shelves on time, but also the cost of production, transportation and markdowns.
Walmart Brings Sales Data Into Ordering Decisions
Indira Uppuluri, Walmart's senior vice president of supply-chain technology, said the retailer's AI supply-chain system starts with demand forecasting. It combines the previous year's Halloween sales, newly launched candy products and social-media trends to determine how much to order from suppliers and when to place those orders.
In distribution, Walmart uses AI to review how goods are moving through its networks. The system examines the products loaded on each trailer and compares them with current store inventory, then flags which trailers need to be unloaded immediately and which can wait. During the holiday period, when goods move continuously in and out of facilities, the process is intended to reduce stockouts at some stores while limiting unnecessary congestion in warehouses.
Prebuilt Pallets and Tags Speed Store Replenishment
The supply-chain work does not end when products reach a distribution center. Walmart must still move goods from store backrooms to sales floors and adjust replenishment as demand changes. Once candy and costumes arrive at a distribution center, automated equipment sorts them onto different pallets according to each store's shelf locations, allowing store employees to unload and stock products more quickly.
Walmart assembles seasonal pallets in advance and holds them at distribution centers, avoiding the need to occupy limited backroom space at stores. When a store needs the inventory, a prepared pallet can be dispatched directly. The pallets also carry Internet of Things and RFID tags, helping Walmart confirm the specific store to which each shipment should go.
Uppuluri said the system continues to look for signs of excess or insufficient inventory throughout the selling season and adjusts plans as goods move through the network. If unsold candy begins to accumulate, for example, the system may prompt suppliers to pause shipments. For the companies involved, the measure of whether the algorithms are working remains straightforward: whether consumers can find the candy and costumes they want when they need them.