Amazon Inventory Forecasting How to Predict Demand and Prevent FBA Stockouts

Running out used to cost you sales. Now it can cost you sales and a fee. Amazon's low-inventory-level fee can apply when both the short-term (30-day) and long-term (90-day) historical days-of-supply metrics fall below 28 days. Beginning in 2026, Amazon assesses the fee at the FNSKU level rather than the parent-ASIN level, making variation-level forecasting more important, and grocery is exempt.

Overstocking has its own penalties: monthly storage, aged inventory surcharges, and consumption of the capacity you will need for your next launch. Forecasting well means finding the band between the two, per SKU, with your actual lead times in the math. The broader fee picture sits in our guide to screening products against the full fee stack.

Running Too Lean	Running Too Deep
Lost sales while out of stock	Monthly storage costs
Low-inventory-level fee exposure	Aged inventory surcharges
Sponsored Products cannot serve	Capacity consumed that a faster SKU could use
Sales history gap on the ASIN	Cash tied up in slow-moving units

Neither side is free, which is why "just order more" is not a strategy.

Start with velocity, measured three ways and compared against each other rather than used as fixed rules:

• 7-day velocity shows current momentum, including whatever advertising is running.
• 28-day velocity can provide a useful intermediate planning window for SKUs without strong seasonality or event distortion.
• 90-day velocity shows the trend and smooths promotional spikes.

If 7-day velocity is materially above 28-day velocity, investigate the cause before forecasting from it. The increase may reflect a promotion or advertising push, but it can also signal genuine demand acceleration. If 7-day sits below 90-day, investigate whether demand is softening before using the longer-term average for the next reorder.

Three adjustments before the number is usable:

Strip out event distortion

A Prime Day or Q4 week is not a baseline. Keep event sales in a separate line so your normal-week velocity stays clean. Our guide to event planning covers forecasting the spikes themselves.

Account for stockout days

If an item was out of stock for nine days last month, its 28-day velocity understates real demand. Divide units sold by in-stock days, not calendar days.

Forecast at the child SKU level

Because the low-inventory-level fee is assessed per FNSKU, a parent-level forecast hides the variation that will actually incur it.

Amazon's own restock recommendations in Seller Central are a useful sanity check. They are built on Amazon's view of your velocity and lead time, which often does not match your real production and freight timeline, so treat them as a second opinion rather than the plan.

Reorder point = (average daily velocity x total lead time in days) + safety stock

Total lead time is the part most sellers underestimate. It is not your supplier's quoted production time. It is production, plus freight, plus customs clearance, plus prep, plus the time FBA takes to receive and make units available. Measure each component from your actual shipment history, using enough recent shipments to capture normal variability rather than relying only on the supplier's quoted production time.

One simple demand-variability method is: safety stock = (peak daily demand minus average daily demand) x average lead time. For SKUs with meaningful lead-time variability, use a model that accounts for both demand and lead-time variation rather than relying on this simplified method alone.

Illustrative example using the simplified method. A SKU sells 20 units a day on average, peaks at 32, and has a total lead time of 60 days:

• Lead time demand: 20 x 60 = 1,200 units
• Safety stock: (32 minus 20) x 60 = 720 units
• Reorder point: 1,920 units

When available inventory hits 1,920, the purchase order goes out. Not when it feels low, and not when a dashboard flags it.

Your reorder model, Amazon's historical days-of-supply metrics, and your capacity limit are three different numbers, and a plan has to satisfy all three. Before any purchase order is approved, look up all three:

Check Before the PO Goes Out	Where to Find It
Current inventory position, including units in transit	Your own records and FBA inventory
Current Amazon historical days of supply per FNSKU	Seller Central
Capacity available for the period	Capacity Monitor

CMO planning rule: never plan from a supplier's quoted lead time. Quotes cover production and, at best, the freight leg. They never include customs clearance, prep, or the time Amazon takes to receive and make units available, and those last three are where reorder plans quietly fail. We rebuild total lead time per supplier from actual receipt-to-available dates across recent shipments, then plan from that number. When a client's reorder point looks wrong, an understated lead time is the first place we check.

The cash constraint nobody models. The reorder point says 1,920 units and the bank balance says 800. That is the real planning problem for most brands, and the answer is not to pretend the shortfall does not exist. Rank SKUs by contribution per unit and velocity, fund the ones that earn the most per dollar of working capital first, and stagger shipments so inventory arrives in waves rather than in one order you cannot afford. Where a SKU has to run lean, price the low-inventory-level fee into the decision rather than discovering it later, and consider merchant fulfillment as a bridge for the tail of the cycle.

Your operational days-of-supply forecast and Amazon's low-inventory-level fee calculation are related, but they are not the same metric. Amazon calculates historical days of supply using inventory levels relative to actual historical sales across short-term and long-term windows, and the fee can apply when both fall below 28 days. It is charged on applicable FBA units shipped while that condition holds.

Use your reorder-point model to reduce the risk of inventory becoming chronically lean, but monitor Amazon's actual historical days-of-supply metrics in Seller Central rather than assuming your internal forecast proves fee eligibility or exemption. Confirm current rates and exemptions there too, since these change annually.

Forecasting demand correctly does not help if Amazon will not take the inventory.

FBA capacity is measured in volume by storage type and shown in the Capacity Monitor in Seller Central, with limits announced for the period ahead and estimates provided for the following two months. Volume is the binding constraint even though unit estimates are displayed alongside. If you need more space, Capacity Manager lets eligible sellers request it by specifying a reservation fee per cubic foot, and performance credits can offset up to 100% of that fee when the inventory sells through. Exceeding your limit incurs an overage fee, which is expensive enough to plan around.

IPI is one factor Amazon uses when determining capacity for established FBA sellers, along with factors such as sales performance and available fulfillment capacity. Amazon communicates the applicable threshold and capacity information in Seller Central, so check the Inventory Performance dashboard and Capacity Monitor rather than relying on a historical threshold. Addressing stranded inventory, improving sell-through, and reducing excess inventory can improve the factors Amazon uses to evaluate inventory performance.

Amazon Warehousing and Distribution is worth evaluating as an upstream buffer if capacity is your recurring constraint rather than demand.

Amazon does not publish a ranking penalty for stockouts, and anyone who quotes you a specific percentage drop is guessing. What is observable:

• The offer cannot win the Featured Offer while there is nothing to sell.
• Sponsored Products cannot serve for an out-of-stock product, removing paid traffic during the stockout period.
• Review accumulation stops, since there are no new orders.
• The ASIN records no sales while unavailable, creating a break in its recent sales history. Amazon does not publicly quantify how that gap affects organic ranking.

After inventory becomes available again, monitor organic traffic, Featured Offer status, advertising delivery, conversion, and sales rather than assuming performance will immediately return to the pre-stockout baseline.

Prevention is most of this article. If the shelf is already empty, the sequence matters more than the speed.

Decide whether merchant fulfillment bridges the gap

If you hold units outside Amazon, listing them as merchant fulfilled keeps the ASIN buyable and the sales history intact while the FBA shipment travels. Set a delivery promise you can actually hit, because a bridge that generates late shipments costs more than the gap did.

Do not switch the ads off entirely

Sponsored Products cannot serve while the product is unavailable, so spend stops on its own. What matters is the restart: bring campaigns back as inventory becomes available rather than leaving them paused, and expect to rebuild data rather than resume at the previous level.

Sequence the restock by what recovers fastest

If you can only expedite part of a shipment, prioritize the FNSKUs with the strongest velocity and the deepest history, not the ones with the largest order quantity. Air freight for a partial quantity often costs less than the combined loss of sales, fee exposure, and rebuilding period on a top seller.

Fix the input that caused it

A stockout is almost always a lead time that was understated or a reorder point that was never recalculated after velocity changed. Update both before the next cycle, or the same gap arrives on schedule.

• Review days of supply by child SKU, flagging anything trending lean
• Compare 7-day against 28-day velocity for direction, and investigate the cause of any gap
• Check the Capacity Monitor for the next period's limit against your inbound plan
• Check the Inventory Performance dashboard and clear stranded or aged inventory
• Confirm in-transit shipments against expected receive dates
• Update the reorder point for any SKU whose lead time or velocity has shifted

When we run an inventory review, we work the same list per SKU: we rebuild velocity from in-stock days, recalculate lead time from actual receipts, reset the reorder point, then check the plan against capacity and Amazon's historical days-of-supply metrics before any purchase order is approved.

How do I calculate a reorder point for Amazon FBA?

Multiply average daily velocity by total lead time in days, then add safety stock. Total lead time includes production, freight, customs, prep, and FBA receiving, measured from your own shipment history rather than a supplier quote.

How much safety stock should I hold?

Enough to cover the gap between peak and average demand across a full lead time, plus coverage for lead-time variability if your shipments arrive unpredictably. Holding materially more than your modeled variability requires increases storage, working-capital, and capacity costs, so weigh that extra protection against its carrying cost.

How do FBA capacity limits work?

Amazon sets limits by volume and storage type and shows them in the Capacity Monitor. IPI is one factor in the allocation, Capacity Manager lets eligible sellers request more space by specifying a reservation fee, and exceeding your limit triggers an overage fee. Check your limit before placing a large order, not after.

Does going out of stock hurt my Amazon ranking?

Amazon does not publish a ranking penalty, so treat specific claims skeptically. What is certain is that you cannot hold the Featured Offer, Sponsored Products cannot serve, and the ASIN records no sales during the gap. Monitor performance after restocking rather than assuming an automatic return to baseline.

Should I just follow Amazon's restock recommendations?

Use them as a cross-check. Amazon's estimate of your lead time is rarely your actual lead time, and its recommendation does not know about your supplier's minimum order quantity, your cash position, or the shipment already on the water.