Amazon Inventory Management in 2026: Capacity, Restock and What a Stockout Really Costs

By Archit Mathur · Published 2025-11-05 · Last updated 2026-08-15

Inventory writing on Amazon opens with a precise-sounding statistic about what stockouts cost: a percentage of sales lost, a number of days to recover, a dollar figure from a case study. Almost none of those numbers have a source. An earlier version of this page did the same thing, and they have been removed rather than softened.

What follows is the part that is actually checkable: how Amazon constrains your inventory today, how the fee structure squeezes from both directions, and how to size a reorder point without pretending to know things nobody knows.

Capacity limits, not storage limits

This is the single most common way a guide reveals its age. Effective 1 March 2023 (opens in a new tab), Amazon replaced the weekly restock limit and the quarterly storage limit with one monthly capacity limit per storage type, stated in cubic feet, on the grounds that a single number gives more predictability and control.

So there is no separate restock allowance to manage, and there is no unlimited tier to aspire to. There is one allowance (opens in a new tab), it is measured in volume rather than units, and it covers both what you hold and what you send in. You manage it in Capacity Manager, which is also where you request more.

What IPI actually controls now

The Inventory Performance Index still matters, but it is an input rather than the gate. Your monthly capacity allowance is derived from IPI together with your recent sales and forecast, your product type, and your history of using the capacity you were given.

That last input does more than sellers expect. Consistently taking up far less than your allowance is itself a signal, and so is repeatedly bumping against it. Amazon has also moved the IPI threshold that qualifies for standard treatment more than once, which is why quoting a specific number in a guide is unwise. Read yours in Capacity Manager.

The four things IPI responds to are unchanged and worth stating plainly: excess inventory, sell-through, stranded inventory, and in-stock rate. Of these, stranded inventory is the one that is free to fix and most often ignored. It is inventory you already own, already paid to store, and cannot sell because of a listing problem.

The fee structure now pushes from both sides

Until a couple of years ago, inventory strategy on Amazon had one failure mode: too much stock, sitting too long. That is no longer true, and it changes the shape of the optimum.

The practical consequence is that the old advice to run as lean as possible now has a price attached, and the old advice to keep deep cover has a different price attached. The optimum is a band rather than a floor, and it is narrower than it used to be.

The September send-in advice is mechanically wrong

A widely repeated tactic says to ship inventory in September to avoid the higher October-to-December storage rates. This does not work, particularly in a year where peak fulfilment fees (opens in a new tab) also carry the 3.5% surcharge on top.

Monthly storage is charged on what is in the network during each month. Units sent in September and still held in November are charged November's peak rate, exactly like units that arrived in October. Sending early does not exempt anything; it adds September and October storage on top.

There are two good reasons to send early anyway, and they are worth doing for their own sake: getting ahead of the Q4 receiving backlog, when check-in times stretch, and securing capacity before your allowance is competing with everyone else's peak inbound. Do it for those reasons and price the storage honestly rather than believing you have avoided it.

Building a reorder point that survives contact with reality

The arithmetic is not complicated. Reorder point equals average daily units multiplied by total lead time in days, plus safety stock. Where sellers go wrong is in both inputs.

Total lead time is not the supplier quote

The number your supplier gives you is production time. The number you need is production, plus booking and freight, plus customs clearance, plus transit to the fulfilment centre, plus Amazon receiving and check-in. That last stage is the one people leave out, and in Q4 it is also the one with the widest variance.

Measure your own, from purchase order to sellable, across your last several orders. Use the worst one, not the average, because the whole purpose of the calculation is to survive a bad cycle.

Safety stock scales with variability, not with sales

A formula circulating in Amazon guides sets safety stock as average daily units multiplied by the square root of lead time. That is not the standard formula and it produces the wrong answer in a specific way: it makes safety stock a function of how much you sell, when it should be a function of how unpredictably you sell.

Two products selling ten units a day need different buffers if one sells ten every day and the other sells nothing for a week and seventy on a Friday. The standard approach scales safety stock with the variability of demand and with the square root of lead time, and with a service-level multiplier reflecting how often you are willing to stock out. If that is more machinery than you want, the usable shortcut is to size the buffer against your worst recent week rather than your average one.

What a stockout actually costs you

Rather than quote a number that would be invented, here is the mechanism, which you can price against your own listing.

Going out of stock removes your offer, so the Featured Offer goes to someone else. Sales history stops accruing on a listing where recent sales feed discoverability. Advertising, if left running, spends against a page that cannot convert. And when you come back, you are competing against a competitor who has had uninterrupted sales history for the duration.

The cost is therefore not the lost margin on the units you did not sell. It is that plus the cost of rebuilding position, and the second part is larger and harder to see. How much larger depends entirely on how contested your listing is, which is why nobody can publish a general figure and why everyone who does has made it up.

Watching the two ends

Almost all of the above reduces to two moments worth knowing about on the day they happen: when a SKU crosses its reorder point, and when inventory becomes unsellable without disappearing: stranded, suppressed, or held in receiving.

Neither of those fires a notification. The restock report is a report you open. Stranded inventory sits in a view you have to visit. Both are visible and neither is announced, so the failure is usually a late decision rather than a bad one: the reorder point was crossed eleven days ago, and the reorder that would have been comfortable then is now an air freight conversation.

Frequently asked questions

Sources

1. 01
   Simplifying FBA capacity limits for your business (opens in a new tab) — Amazon Seller Forums
2. 02
   FBA capacity limits (opens in a new tab) — Amazon Seller Central
3. 03
   Monthly inventory storage fees (opens in a new tab) — Amazon Seller Central
4. 04
   Low-inventory-level fee (opens in a new tab) — Amazon Seller Central
5. 05
   Holiday 2026: Same fees, same eligibility, earlier deadlines (opens in a new tab) — Amazon Seller Forums