Overview

Amazon offers FBA sellers a built-in demand forecasting tool that projects future customer demand for individual products up to 40 weeks ahead. The tool uses a probabilistic model tailored to each ASIN, drawing on historical order data, active promotions, and seasonal patterns to generate its estimates. Understanding how to read and use these forecasts can help sellers make smarter restocking decisions and avoid costly stockouts or overstock situations.

Key Points / What Sellers Need to Know

• Product-specific forecasts — Each ASIN gets its own unique forecast model rather than a one-size-fits-all projection, meaning the estimates reflect the actual sales behavior of that particular product.
• Up to 40 weeks of visibility — Sellers can view projected demand nearly ten months into the future, providing ample lead time for sourcing and logistics planning.
• Two prediction levels — The tool displays both a mean (expected) forecast and a P90 (optimistic) forecast, giving sellers a range rather than a single number to plan around.
• Tied to restock recommendations — The demand forecast feeds directly into Amazon's restock recommendation engine, so understanding the forecast helps sellers evaluate whether Amazon's suggested reorder quantities make sense.
• Historical comparison included — The tool overlays actual past order data and year-over-year comparisons when available, making it easy to spot trends and validate the forecast.

How the Forecast Model Works

Amazon's demand forecast relies on a probabilistic model, which means it doesn't just produce a single prediction — it generates a distribution of possible outcomes and communicates both the expected case and the uncertainty around it. The model pulls in several data inputs including your product's historical order volume, any deals or promotions you currently have running, and broader seasonal demand patterns relevant to your category. Because each product gets its own model, a winter-seasonal item and a year-round consumable will produce very different forecast shapes even if they have similar average sales. This approach is more sophisticated than simple moving averages and can adapt as new sales data comes in over time.