Peak Inventory: September's Last Fixes

Ocean sourcing for Black Friday is already late. Manufactured goods commonly run 60 to 120 day true lead times from order to dock, and Amazon's Black Friday inbound cutoffs fall four to seven weeks from today. The early-scheduling discounts on Amazon deal fees, $50 off per deal, closed on 5 August for the October Prime event and 5 September for Black Friday and Cyber Monday. If your lead time sits anywhere in that band, the ordering decision was made weeks ago, whether or not anyone made it on purpose.

So this is not an article about ordering more. Three decisions are still open in mid-September, and they decide your peak margin: where the inventory you already own goes, whether any of it is worth flying, and whether the forecast the plan rests on is honest.

Deadlines in the next 30 days

Every date below comes from Amazon's Seller Central peak readiness page and from carrier announcements, checked on 10 September 2026. Two have already passed.

Just outside the window: the Black Friday and Cyber Monday deal submission deadline in Seller Central, first set for 20 October, has since been extended to 13 November (confirm it in your own account; Amazon has moved this year's deal deadlines before), and that event's FBA inbound cutoffs are 21 October for minimal splits and 28 October for Amazon-optimised splits. Amazon has since set the October Prime event for 6 and 7 October 2026. The Black Friday and Cyber Monday window is not yet announced; last year Amazon ran it as one window, 20 November to 1 December, and its press release archive is the place to watch. Amazon later extended the October event's deal deadline from 8 to 22 September, a change that appeared in-account before any public page, which is why your own account is the source to plan against. If you ship orders yourself, Amazon has not published the 2026 holiday delivery cutoffs either; watch the News and Announcements category of the Seller Forums, and week 8 of this series covers them.

What is actually closed, and what is still open?

Closed: new ocean production for Black Friday, the early-scheduling fee discounts, and every October Prime event inbound route except optimised splits, which shuts on 16 September. Open: three decisions, each of which moves margin more than a late purchase order would.

1. Allocation. Where the units you already own go, by channel, and how many of them Amazon will physically let you send.
2. Air. Whether any units still at origin are worth flying, on landed cost rather than on fear.
3. Forecast honesty. Whether the demand number your plan rests on includes the demand you are about to pay to create.

Where should the inventory you already have go?

To the channel that earns the most on it, which is not automatically Amazon. In a constrained market, inventory is the one lever the retail side cannot take from you: Amazon's category team cannot sell what you do not ship. That makes allocation a margin decision before it is a logistics decision, and it is the decision the vendor manager across the table would rather you did not notice you hold.

The clearest case I have worked was a leading baby and toys vendor during the post-COVID supply crunch. In the annual negotiation, Amazon's category team asked for significant price decreases across the catalogue and higher co-op accrual rates at the same time, under a threat to de-list. Accepting would have pushed more than half the catalogue to negative margin. The vendor named its lowest-priced items as loss leaders it would rather de-list, and when Amazon threatened to stop purchasing, it stopped fulfilling Amazon purchase orders instead.

The crunch is what made that possible. Demand from a country stuck at home ran far ahead of manufacturing and shipping capacity, so for once vendors had more orders than units and could choose where the units went. Amazon was that vendor's lowest-margin channel, so every purchase order it accepted from Amazon earned less than the same units shipped to Walmart or Target. Moving Amazon down the allocation was mechanical: change the replenishment settings in Vendor Central so automated ordering stopped, confirm fewer units on the purchase orders that still arrived, and reject the rest outright. Amazon does not feel that immediately, because it tries to hold four to six weeks of cover on a product. So along with the notice that purchase orders would not be accepted, the vendor told the category team the week its out-of-stocks would start to show in sales. Having sat on that side of the table, I also knew which metrics beyond sales and in-stock rate would move, and that those metrics had leadership visibility. The hold ran more than ten weeks. Amazon's numbers moved about four weeks after shipments stopped, once the cover ran out, and from then on the category team was answering for a set of declining metrics its own leadership watched.

The mechanism is public: co-op and contra-COGS deductions were what the category team was pushing, and offering to de-list unprofitable selection turns the negotiation into a trade between Amazon's profitability and your growth. The relationship survived, as it usually does. Both sides want to sell the whole catalogue; the terms just have to work for both. There was strain with that category team, partly because it had said how things would go and they went otherwise, but Amazon rotates people fast enough that most of the faces in any negotiation have moved on within six to twelve months.

That is the 1P version. On the 3P side, Amazon caps what you can send before intent comes into it. Check the caps this week:

• Per-ASIN restock limits. Reported back since April 2025, without a Seller Central announcement. Operators commonly see a cap around 90 days of supply, but Amazon does not publish the rule, it varies by seller and category, and it can change without notice. Read your own restock limits page rather than planning to a number you read anywhere, including here.
• The cap is shared across sellers on the listing. Operators report it is calculated at ASIN level across every seller, not per account. On a multi-seller listing, a reseller's inbound eats your headroom.
• The warehouse route does not bypass the cap. Units in Amazon Warehousing and Distribution do not count toward FBA capacity, but operators report that moving them into FBA still sits under the ASIN restock limit, so the obvious workaround does not get you more room.
• Amazon does not publish the Inventory Performance Index threshold. Most 2026 guides work to 400, and one tracker reports 350 for the Q1 2026 reset. Below the line Amazon caps what you can send in, and it may limit inbound volume on its own demand forecast even when your score is healthy.

Vendors and sellers alike end up fighting Amazon for warehouse space at peak, and an override on a purchase order cap or an inventory cap is hard to get and sometimes impossible. The case that gets heard is made with data: recent changes in your inventory velocity, changes in ad spend, and the promotions you have booked. Bring those, not a forecast.

Vendors have one route to the customer that needs no override: Amazon Direct Fulfillment, the dropship route for 1P vendors in Vendor Central. Amazon still buys the units, but each order ships to the customer from your own warehouse, so the stock sits outside Amazon's inbound caps. It does not replace stocking Amazon. Amazon routes orders to inventory in its own network first, because that tighter delivery promise wins the buy box, unless your warehouse can meet delivery messaging equivalent to Amazon Retail's fast-track promise. Where it earns its place is peak. Direct Fulfillment inventory can carry a Black Friday deal, and it matters most when the deal sells through Amazon's own stock, because your warehouse keeps it selling. I led Direct Fulfillment for Amazon's furniture team when furniture was about a third of Amazon's North America Direct Fulfillment business, and my view is that every vendor should have it active.

If yours is not live, ask your vendor manager to switch it on this week. How long enablement takes depends on the team and the vendor manager, and there is no typical figure, which is why the request comes first. Amazon's Direct Fulfillment workflow guide says purchase orders can be confirmed in Vendor Central as well as through EDI, so a low-volume vendor has a manual route. For an automated connection, the nearest public benchmark is Comparatio's figure for Amazon EDI setup in general, not Direct Fulfillment specifically: 4 to 8 weeks, with 2 to 3 test cycles and 3 to 5 business days added by each round of corrections. My judgement, not a published deadline: if Black Friday opens around 20 November as it did last year, a vendor starting from zero should request Direct Fulfillment now and start the integration by the end of September to be live and stable before the deals run.

Whatever Amazon will not take is the allocation answer for your other channels. The sooner you know that number, the sooner the rest of the plan is real.

Is it worth flying?

Only if the landed cost of a flown unit still clears margin after every peak charge that lands on it, and for most catalogues that is true for a short list of SKUs, not the range. Work the arithmetic per SKU, in this order:

• Preparation. Amazon ended its US prep and labelling services for FBA and Amazon Warehousing and Distribution on 1 January 2026. A flown unit still has to be prepped and labelled before it inbounds, so this is a prep-capacity question before it is a freight booking.
• Inbound placement fees. These rose again on 15 January 2026, by roughly $0.05 per unit on standard-size minimal splits and roughly $0.27 per unit on large bulky items. For large bulky items on minimal splits, operators report about $2.16 to $2.67 per unit at 5 lb or less, rising to about $4.85 to $6.00 per unit at 42 to 50 lb. Those ranges come from operator reports, not Amazon's schedule, so confirm the live fee in your own account before you build it into a landed cost.
• The optimised-splits lever. Choosing Amazon-optimised shipment splits typically takes the placement fee to zero, and it buys the later cutoff: 28 October instead of 21 October for Black Friday. That week is the most useful thing in this section.
• The peak fulfilment surcharge, 15 October to 14 January, prices every unit that sells in the window. Amazon has published the period; check the current rate in your account.
• Aged inventory surcharges on what does not sell. Nothing at 180 days or less, then 30-day tiers from day 181: $0.50 per cubic foot, rising to $1.50 by 241 to 270 days, jumping to $5.45 at day 271 and $5.90 by 331 to 365 days, then $0.30 per unit or $6.90 per cubic foot (whichever is greater) at 12 to 15 months and $0.35 per unit or $7.90 per cubic foot beyond. Clothing, shoes, bags, jewellery and watches are exempt from the tiers before day 271. Over-sending in October is not a free hedge. It is a Q1 charge with a delay on it.
• The tail. A removal order back to you starts at about $0.84 per unit for a light standard-size item on published 2026 rate tables, rises with weight, and takes 7 to 30 days. Amazon Liquidations recovers a small fraction of list price and stops the storage clock at once. The September decision is whether to clear slow stock now or carry it into an aging charge.
• Amazon Warehousing and Distribution as the buffer. 2026 rates from 15 January: smart storage $0.51 per cubic foot per month in the West and $0.43 elsewhere, $1.40 per box in and out, and $1.40 per cubic foot to move into FBA on the base and smart storage tiers, or $1.26 on Amazon's managed tier, which also takes 20% off storage. Placement into FBA is included on that route. Amazon's program page states there is no seasonal surcharge, so the rates hold through Q4. Two cautions. Check-in into FBA has historically taken anywhere from days to weeks, so do not make this warehouse the only feed into peak. And the smart storage rate is reported to need at least 70% of your warehouse-to-FBA volume on auto-replenishment over the trailing 90 days plus 70 days of supply across both, recalculated monthly, so a September forecast miss can cost you the rate part-way through the quarter.

On the catalogues where I have made this call, it came down to ASIN economics and dimensions. Air freight is expensive, and worst on heavy, bulky items. It works for small, light units, where the freight spreads across more of them, at a selling price high enough for the margin to absorb it. A heavy, low-priced item fails both tests, and no amount of December demand changes that.

Is your peak forecast honest?

Probably not, if it starts from Amazon's number. Amazon's demand forecast reflects expected consumer demand only. It does not include the promotions, advertising or deal events Amazon does not know about yet, and you should not assume it counts the ones you have entered, so a plan built from it under-counts exactly the demand you intend to pay for. That is the most important caveat in Amazon forecasting, and it applies on both sides of the Vendor Central and Seller Central divide even though the tools differ.

On the 1P side, the Vendor Central demand forecast has been showing 13 weeks of order history and a 26 week forward view at mean, P70, P80 and P90, and it is not a purchase commitment. Amazon is reported to prefer that vendors plan to P90, which carries the highest overstock risk; that preference comes from an agency's guide, not a published Amazon requirement. Pick the probability level from the cost asymmetry between a stockout and an overstock for that SKU, not from a default.

Two habits from operators who plan this well: build production to your true lead time rather than to the date the purchase order lands, and move forecast review from weekly to daily through Q4. Their planning ranges, 30 to 60 days of supply in steady state and 60 to 90 in Q4 with 15 to 45 days of safety stock, are starting points to test against your own velocity, not a standard.

Week 2 of this series takes 1P forecasting and purchase-order discipline in full. The eCommerce strategy work we do with Amazon brands usually starts here, because a forecast that may leave out the demand you are creating misallocates every unit downstream of it.

What has to be decided this week?

The 16 September optimised-splits cutoff closes the October event on Wednesday. On 14 October the Black Friday warehouse route closes too. Between those dates, set the allocation by channel from what Amazon will actually accept, run the air arithmetic per SKU against the charges above, and add your own promotional demand to the forecast before anyone books freight against it. The ordering decision is behind you. These three are not.

If you want a second pair of eyes on your peak allocation, a 45-minute working session on your actual SKU list and channel margins is the fastest way to get one. Book it here.