How does AI handle demand forecasting for new products with no sales history?

The Core Challenge of Forecasting Without History

Demand forecasting fundamentally relies on learning from the past, which creates an obvious problem for new products: there simply isn’t a sales history to learn from. This is one of the hardest forecasting scenarios manufacturers face, since the usual foundation for AI models — historical demand data for the specific product in question — doesn’t exist yet. Getting new product forecasts wrong carries real consequences too, since launch-phase production and inventory decisions are often made with long lead times and limited ability to course-correct quickly.

Borrowing From Similar Products

The most common AI approach to this problem is analogous forecasting, sometimes called “new product forecasting by analogy.” Rather than relying on the new product’s own sales history, the model identifies existing products with similar characteristics — comparable category, price point, target customer, features, or launch context — and uses their historical demand patterns as a reference point for what the new product’s demand curve might look like.

Machine learning models can go beyond simple manual analog selection by systematically analyzing large catalogs of past product launches, identifying which attributes tend to correlate most strongly with certain demand patterns, and weighting multiple analogous products together rather than relying on a single comparison. This can produce a more nuanced starting estimate than a human planner picking one or two “similar” products by intuition alone.

Rapidly Updating as Real Data Arrives

Because analog-based estimates are inherently uncertain, AI forecasting systems are typically designed to update quickly once the new product actually launches and real sales data starts coming in. Early sales signals — whether from initial retail orders, e-commerce activity, or distributor sell-through — are fed back into the model, which recalibrates its forecast to reflect the product’s actual observed demand pattern rather than continuing to rely solely on its initial analog-based estimate. This rapid recalibration is one of the more valuable aspects of AI-driven forecasting for new products, since it shortens the window during which planning decisions rest on the least certain data.

The Continued Role of Human Judgment

Even with sophisticated analog modeling, forecasting a genuinely new product remains one of the areas where human input continues to play a meaningful role. Sales teams, marketing, and category managers often bring qualitative context — competitive positioning, anticipated marketing spend, or early customer feedback — that isn’t captured in historical analog data. Many manufacturers deliberately blend this human judgment with AI-generated forecasts for new launches, rather than relying purely on either input alone, precisely because the underlying uncertainty is so much higher than with established products.

Bottom Line

Since new products lack their own sales history, AI forecasting relies primarily on analogous forecasting — borrowing demand patterns from similar existing products based on shared attributes — and then rapidly recalibrates once real sales data starts arriving after launch. Because uncertainty is inherently higher in this scenario, human judgment from sales and marketing teams commonly supplements AI-driven forecasts for new product introductions.

Go deeper

• Can AI Inventory Forecasting Actually Prevent a Small Store From Overselling?
• Can AI Help You Find Profitable Dropshipping Products, or Is That Overhyped?
• Do You Have to Disclose AI-Generated Designs to Customers Buying Physical Products?