The National Retail Federation estimates US retailers lost over $1 trillion in revenue in 2025 from inventory distortion — stockouts and overstocks combined. Amazon itself has moved to AI-driven forecasting for exactly this reason: its new foundational forecasting model, which factors in regional demand signals down to ski goggle sales in Boulder during peak season, delivered a 10% improvement in long-term national forecasts and a 20% improvement in regional forecasts for popular items. For third-party sellers, getting this wrong isn’t just a cash-flow problem — on Amazon specifically, a stockout tanks search ranking and Buy Box eligibility, and recovery takes weeks, not days.
AI inventory forecasting for Amazon sellers uses machine learning to predict demand at the SKU level — factoring in sales velocity, seasonality, lead times, and increasingly PPC and pricing data — to recommend when and how much to reorder. It matters more on Amazon than on most other channels because the consequences of getting it wrong are structural, not just financial: a stockout drops your Best Seller Rank and Buy Box eligibility, and overstocking triggers escalating storage fees plus a lower Inventory Performance Index (IPI) score, which can cap how much space Amazon even lets you store. Amazon’s own free tools (Restock Recommendations, the Inventory Performance Dashboard) work for sellers under roughly $30K in monthly revenue; most sellers outgrow them by 20-30 SKUs and need a dedicated forecasting layer.
Every retail channel deals with the basic stockout-versus-overstock tradeoff, but Amazon compounds the consequences in ways a standalone e-commerce site or a physical retailer doesn’t. Run out of stock during a peak period and you lose Best Seller Rank, Buy Box eligibility, and review momentum — all of which take weeks to rebuild even after you’re back in stock. Overstock, and FBA’s escalating storage fee structure quietly erodes margin: aged inventory fees of $0.15 per unit begin accumulating after 271 days, and long-term storage fees jump to $6.90 per cubic foot after 365 days unsold.
Amazon has invested heavily in AI forecasting for its own operations for exactly this reason. The company’s newer foundational forecasting model — currently live in the US, Canada, Mexico, and Brazil — layers in regional and time-bound signals like weather on top of historical sales data, and Amazon reports the shift contributed to a 10% improvement in long-term national forecasts for major deal events and a 20% improvement in regional forecasts for popular items. Third-party sellers don’t get access to that system directly, but the same principle — richer signal beats sales history alone — is exactly what dedicated seller-side forecasting tools are now built around.
Amazon’s Inventory Performance Index (IPI) is a 0-1,000 score updated weekly that determines how much FBA storage space a seller gets and what fees apply. It’s built from four components: excess inventory percentage, sell-through rate, stranded inventory percentage, and in-stock rate — all four of which are directly shaped by forecasting accuracy, not separate from it. The minimum threshold to avoid storage restrictions sits at 400 in 2026, with some sources citing 450 as the safer target; falling below it means Amazon can limit shipment volume, restrict new inventory, and increase storage costs specifically for the sellers who most need flexibility to fix the problem.
Machine learning models trained on historical sales, seasonality, and trend data predict future demand at a granular, per-product level — well beyond simple moving averages.
Combining lead times, safety stock targets, and demand forecasts to recommend not just how much to reorder, but exactly when to place the order to avoid a coverage gap.
Better tools flag excess inventory risk at 60+ days of inventory on hand — before Amazon’s own threshold triggers at 90+ — giving sellers time to run a promotion or removal order proactively rather than reactively.
The most advanced tools coordinate forecasting with PPC and pricing strategy directly — automatically flagging when ad spend is driving traffic to a listing that’s about to run out of stock, since paying to accelerate demand into a stockout wastes both the ad spend and the sale.
| Policy | What It Means for Forecasting |
|---|---|
| Low-Inventory-Level Fee | Applies when 30-day and 90-day days-of-supply both fall below 28 days on eligible products — running too lean now costs money per unit on higher-velocity SKUs |
| Reduced storage capacity limits | Amazon cut standard capacity forecasting from 6 months to 5 months of projected sales in 2025 — less room for error in over-ordering |
| Aged inventory fees | $0.15/unit monthly begins at 271 days unsold; escalates further at 365 days — a slow-moving SKU becomes an active cost, not just dead capital |
| IPI-based restrictions | Below the 400 threshold, Amazon can cap storage volume and increase fees — the sellers most in need of flexibility get the least of it |
Amazon’s built-in Restock Recommendations and Inventory Performance Dashboard provide a reasonable starting point for new sellers doing under roughly $30,000 in monthly revenue. Beyond that, the free tools tend to favor overstocking, don’t account for PPC or pricing strategy, and lack the sophistication needed for complex, multi-SKU catalogs. Most sellers outgrow Amazon’s native tools somewhere between 20 and 30 SKUs, or around $50,000 in monthly revenue — the point where manual review of restock reports stops being a reasonable weekly task.
Amazon’s free tools tend to bias toward overstocking and don’t factor in your specific pricing or advertising strategy — useful as a starting point, but not a substitute for demand-specific forecasting once a catalog grows past the basics.
Because Amazon-specific stockout costs compound — lost rank, wasted PPC spend, ceded market share to competitors who stayed in stock — a purely reactive reorder process is consistently more expensive than proactive forecasting, even accounting for the cost of a forecasting tool.
By the time a seller notices reduced storage limits, the underlying excess or stranded inventory problem has usually existed for weeks. Weekly IPI component tracking catches the drift before it becomes a capacity restriction.
Standalone forecasting tools can predict demand accurately and still leave a seller manually reconciling that prediction against an active PPC campaign or a pricing change — the gap between an accurate forecast and a coordinated response is where a lot of real money gets lost.
High Dreams LLC is a Colorado-based digital growth agency specializing in Amazon store setup, listing optimization, and account management — helping sellers build inventory forecasting into the same strategy as pricing and advertising, instead of treating it as a separate, reactive task. The agency has shipped work for 150+ clients worldwide across Amazon, Walmart, Etsy, and eBay.
SKU-level forecasting built around real sales velocity, seasonality, and supplier lead times.
Excess and stranded inventory flagged early, before Amazon’s own thresholds trigger restrictions.
Ad spend and reorder timing managed together, so advertising never drives traffic into a stockout.
Services include e-commerce management across Amazon, Walmart, Etsy, and eBay, plus AI chatbots for customer inquiries and website development for sellers building beyond the marketplace.
Get a free consultation to review your current forecasting approach and IPI standing.
A stockout on Amazon doesn’t just cost the missed sale — it drops your Best Seller Rank and Buy Box eligibility, both of which take weeks to recover even after restocking, unlike most other retail channels.
IPI is a 0-1,000 score based on excess inventory, sell-through rate, stranded inventory, and in-stock rate — all four are directly shaped by forecasting accuracy. Falling below the 400 threshold in 2026 can trigger storage restrictions and higher fees.
They’re a reasonable starting point under roughly $30,000 in monthly revenue, but most sellers outgrow them by 20-30 SKUs since Amazon’s native tools favor overstocking and don’t account for pricing or PPC strategy.
A per-unit fee introduced in 2024 that applies when both 30-day and 90-day historical days of supply fall below 28 days on eligible FBA products, exempting items with fewer than 20 units sold in the last 7 days.
The National Retail Federation estimated US retailers lost over $1 trillion in revenue in 2025 from inventory distortion — the combination of stockouts and overstocking across the industry.
Sources: Retail Dive, “Amazon improves AI-powered demand forecasting” (citing Amazon’s own June 2026 announcement) · National Retail Federation, 2025 inventory distortion research · Seller Labs, “Amazon IPI Score 2026” · SupplyKick, “Amazon Inventory Forecasting: How to Plan FBA Stock” · Profasee, “Amazon Inventory Forecasting Tools [2026]” · Drivepoint, “Amazon Inventory Forecasting: Best Practices Guide” · Nova Analytics, “Best Amazon Inventory Management Tools 2026” · EcomCircles, “Amazon Inventory Management Tips” (2026) · StockDesk, “Amazon Inventory Forecasting: How FBA Sellers Avoid Stockouts.”
