Inventory Split: Amazon vs Wholesale Stock Allocation

The allocation question sounds like forecasting and is really about reversibility. Stock sent to an Amazon fulfilment centre is expensive and slow to get back. Stock held at your own warehouse can go to Amazon in a week. So the default should be to hold inventory in the reversible place and push it to the irreversible one only as fast as sales consume it.

That single principle solves most of the problem. The rest is working out how much cover each channel needs, and what to do when a purchase order arrives that your plan did not anticipate.

Why the two channels behave differently

Amazon demand is continuous and fairly predictable. Units leave at a rate you can see daily, and the rate responds to price, rank and advertising in ways you have learned to read. The risk is running out, because a stockout costs rank and the recovery takes weeks.

Wholesale demand is lumpy and discontinuous. Nothing happens for a month and then a distributor orders four hundred units with a ship window ten days out. The risk is not being able to fill it, because a supplier who misses a first order rarely gets a second.

Those two shapes need different buffers. A continuous channel needs cover measured in weeks of sales. A lumpy channel needs cover measured in the size of a plausible single order.

That distinction is the practical heart of the split, and it is why splitting by a fixed proportion of every run is the wrong instinct.

The cost of getting it wrong in each direction

Both errors cost money, but not symmetrically.

Too much in FBA and you pay monthly storage on units that are not selling, you risk long-term storage surcharges on the slow sizes and colours, and your cash is locked in a building you cannot ship pallets out of. You also end up filling wholesale orders by pulling stock back, which costs fees and time and returns units in mixed condition — the details are in FBA removal orders for wholesale.

Too little in FBA and you drop below the cover level where Amazon starts charging for low inventory, your delivery promise slips, and a stockout resets the momentum you paid advertising money to build.

Notice that the second error is recoverable with a fast inbound shipment if the stock exists somewhere domestic. The first error is recoverable only through removals. That asymmetry is the reason to hold stock at the 3PL and top FBA up frequently rather than shipping in large batches.

A rule that works

Here is a concrete allocation method that survives contact with reality. Use weeks of cover, not percentages of a run.

Start with your Amazon sales rate per week for the SKU. Decide the cover you want in the fulfilment centre — enough to survive the lead time from your warehouse to Amazon's receiving, plus a buffer for a receiving delay, plus enough to stay above the low-cover threshold that triggers extra fees. For most brands that lands somewhere between four and eight weeks of sales.

Then decide your wholesale reserve as one plausible large order plus one typical reorder. Not an average of past demand, because the average of a lumpy series is meaningless for capacity planning. If your biggest realistic single PO is three hundred units and a normal reorder is eighty, reserve three hundred and eighty.

Everything else sits at the warehouse as unallocated stock, available to either channel.

Review the numbers monthly and after any event that changes the sales rate — a deal, a price change, a new account. The rule is not the point; having one explicit number per channel that anyone on the team can check is the point.

What to do when a purchase order exceeds the reserve

It will happen, and it is a good problem. Work through the options in this order.

Check the unallocated stock first. This is why the reserve is not the whole wholesale allocation.

Ask the buyer about the ship window. Many windows have more room than the PDF suggests, and a partial shipment now with the balance in three weeks is usually acceptable if you propose it early and in writing. Buyers hate surprises far more than they hate constraints.

Look at expediting production. If your factory lead time plus transit fits inside a revised window, that is cheaper than a removal.

Only then consider pulling from FBA, and only if the units will not need rework. Units carrying Amazon-specific labels are not shelf-ready, and the packaging differences are set out in FBA prep vs wholesale packaging.

What you should not do is quietly cut the Amazon replenishment to cover the order without adjusting the sales expectation. A stockout that arrives three weeks later looks unrelated and is not.

Does wholesale demand deserve the stock at all

Worth asking honestly per SKU, because a unit can only be sold once.

Compare what the unit earns in each channel. On the marketplace it earns the retail price less the referral fee, the fulfilment fee, the storage it consumed and the advertising needed to sell it. Wholesale it earns the wholesale price less pick and pack and freight, with no advertising and no referral fee, but with terms attached and a longer cash cycle. The comparison for a typical product is worked through in ACoS vs wholesale margin.

For many brands the per-unit contribution is closer than expected, and the wholesale unit arrives without any ad spend. But cash timing differs sharply: Amazon pays quickly, a retailer pays net terms from receipt, and that gap is real working capital, as Amazon payout vs net 30 lays out.

The strategic answer usually overrides the per-unit one. A brand that exists only on one marketplace carries a concentration risk, and wholesale accounts reduce it. Allocating stock to a channel that is being built is an investment decision, not an optimisation.

Running one number instead of two

The operational failure that undoes a good allocation plan is two systems each believing they own all the stock.

Pick one place where available-to-promise lives, usually your 3PL's system or a layer above both. Every FBA replenishment is a deliberate transfer out of it. Every wholesale order is a deliberate commitment against it. Nobody quotes a delivery date from a Seller Central screen.

Flag the wholesale reserve so it is visible and not accidentally consumed by an Amazon restock suggestion. Reserve logic in most warehouse systems handles this; a spreadsheet column works at small scale as long as one person owns it.

And keep a single item master — variant, barcode, case quantity, dimensions, weight — so the same SKU means the same thing in both channels.

Planning the next production run

The split gets much easier at the purchase-order stage, because that is where quantity is actually decided.

Order for the Amazon sales rate over the next cycle, plus the wholesale reserve, plus whatever pipeline you have visibility of. Ask the factory to pack it retail-ready in cases, which keeps every unit eligible for either channel until the last possible moment and avoids committing stock to Amazon at the factory.

If you are trying to size that wholesale reserve for the first time and have no order history to use, the input you need is which kinds of accounts would buy. Paste your listing into WholesalePilot and the preview shows the retailers and distributors that stock comparable products — a distributor-heavy list points to larger, lumpier orders and a bigger reserve, while a list of independents points to small frequent ones and a much smaller one.