Top Demand Forecasting and Inventory Optimization Software 2026

Purchasing at a distributor is one of the hardest jobs in the business. Too much stock ties up working capital. Too little sends customers to the competitor down the road. Lantern, the Purchasing AI for Distributors, guides what, when and how much distributors buy. The scale is huge. One Lantern customer carries 50,000 products across 22 branches: 1.1 million stocking decisions for a small team of buyers. Demand is less predictable, lead times keep moving, and catalogs keep growing, while spreadsheets and fixed ERP rules have not kept up. Lantern now guides more than $800 million of inventory across 700 locations for over 30 distributors in HVAC/R, industrial, electrical and plumbing. NAW named Lantern a top innovator in distribution in 2026. Three Co-founders, 200 Conversations Lantern was founded by Matt Rojas, Conor Leen and Jay Shitole. It began as a Stanford research project in which Rojas and Leen interviewed 200 distribution executives, who named forecasting as their biggest pain point. Rojas, now CEO, grew up around the industry through his father. Shitole, now CTO, spent five years leading Walmart’s Applied AI Lab, building demand forecasting models at massive scale. A Custom Model for Every Distributor Rather than picking from off-the-shelf models, Lantern builds a custom machine learning model for each distributor, trained on its own sales, inventory, purchasing, lead time, customer and product data. “Every business we work with gets their own unique model, because every business we work with is a bit different,” says Conor Leen, Co-founder and President. Each week, Lantern forecasts demand for every item and location, drawing on external signals such as weather and carrying forward carrying history from linked or replaced products. It forecasts supplier lead times so stocking levels adapt before delays become shortages and classifies products by margin, hits or revenue, adjusting safety stock as items change groups. From Forecast to Purchase Order Lantern turns those forecasts into orders. Buyers open suggested purchase orders, grouped by supplier and branch, with supplier minimums, free-freight targets and buy packs already factored in. Before buying more, Lantern checks whether another branch has stock to spare and recommends a transfer instead across distribution centers, hub-and-spoke networks and independent branches. Lantern is a first draft, not an autonomous agent. Buyers see the numbers behind each recommendation and can override it when they know something the data does not. In practice, they accept 98 out of 100, then send orders to the ERP with one click. Business Impact Lantern judges itself on commercial results: fill rate, excess inventory, turns and GMROI. On average, customers reduce inventory by 30 percent, lift sales by 5 percent, and save 20 buyer hours a week. R&E Supply moved from annual to weekly stocking updates and cut inventory 34 percent while maintaining sales, freeing cash to fund a new location. CAPP grew sales 9 percent while holding less inventory, and Lesman cut inventory 25 percent in year one while sales beat budget. Live in Weeks Customers go live in six to eight weeks, set up free of charge by the same team that builds the product, with integrations for Prophet 21, Eclipse, Infor, SAP, Microsoft, Acumatica and DDI. Distributors can test Lantern first: it trains on two and a half years of their data, buys for the final six months, and compares the outcome with what their team actually did. By pairing custom machine learning with buyer judgment and measurable purchasing results, Lantern has earned recognition as a Top Demand Forecasting and Inventory Optimization Software 2026.