Amazon FBA or FBM - how to prepare your warehouse and WMS

In FBA (Fulfillment by Amazon) you ship stock in bulk to Amazon's fulfilment centres, and Amazon takes over storage, packing, shipping, customer service and returns. You are buying two things above all: the Prime badge, which can lift a listing's conversion rate by tens of percent, and elasticity - you do not hire seasonal packers for Q4, Amazon does.

You pay through three streams of fees: the referral fee on each sale (typically 8-15% depending on category), a per-unit fulfilment fee based on size tier and weight, and monthly storage fees per cubic metre - with rates that jump sharply in Q4 and separate, painful surcharges for inventory aged past 180-270 days. On top of that come preparation costs: every unit needs an FNSKU label, and inbound shipments must meet Amazon's carton and pallet requirements. Slow-moving or low-margin products can go underwater in FBA on storage fees alone - which is why the FBA-or-FBM decision is made per product, not per company.

In FBM (Fulfillment by Merchant) you fulfil Amazon orders from your own warehouse. The margin stays with you; so does control over stock and packing quality. The catch: Amazon measures your operational metrics automatically and enforces them without mercy:

• Late Shipment Rate below 4%. The share of orders confirmed as shipped after the promised ship date. Staying over the threshold for a couple of weeks is a real suspension risk.
• Valid Tracking Rate above 95%. Amazon verifies that the tracking number actually exists in the carrier's system and shows a first scan. Typing tracking numbers in manually "in the evening after dispatch" reliably ruins this metric.
• Cancellation Rate below 2.5%. Pre-shipment cancellations - almost always the result of selling stock that physically was not there, i.e. broken stock synchronisation.
• Order Defect Rate below 1%. Negative feedback, A-to-Z claims and chargebacks combined.

Those four numbers are, in practice, your WMS requirements list. Without a system that reserves stock in real time, watches carrier cut-off times and returns tracking automatically, FBM above a few dozen orders a day becomes a game of roulette.

The classic disaster scenario: the same stock sells in parallel on Amazon, eBay and your own shop. The last unit sells in two channels three minutes apart, because synchronisation runs every fifteen minutes. One of those orders you must cancel - and your Cancellation Rate just moved.

• One source of truth. Available-to-sell stock is managed by the WMS, not by a spreadsheet and not by any single channel. Channels receive available stock minus reservations, never the gross warehouse quantity.
• Reservation at order arrival. An order pulled from Amazon reserves its units immediately, before anyone picks anything. The double-selling window shrinks from minutes to seconds.
• A safety buffer on fast movers. For products selling quickly across channels, report stock to Amazon minus 1-2 units or a few percent. The cost: an occasional "out of stock" while goods are physically available. The gain: no cancellations.
• Update frequency. The realistic standard is every 5-15 minutes via API. Anything above an hour with fast rotation ends in overselling.

On the technical side Amazon exposes SP-API (Selling Partner API), the successor to the retired MWS. Orders come from the Orders API, stock and prices go through Listings and Feeds, shipment confirmations with tracking go back through the fulfilment endpoints. Add developer application registration, LWA tokens and request signing - the entry barrier is noticeably higher than with most European marketplace APIs.

In practice sellers choose one of two routes. First: a WMS integrated with Amazon directly over SP-API - the shortest data path and full control, provided the vendor actually maintains the integration (Amazon periodically forces API version migrations). Second: a middleware integrator that connects Amazon, eBay and your webshop on one side and exchanges orders and stock with the WMS on the other. The second route is faster to launch and convenient with many marketplaces at once, at the cost of an extra link in the chain that has its own delays and outages. Below a few hundred orders a month, middleware is usually enough; past that, the direct route starts paying for itself.

Even if you sell mostly through FBA, your warehouse does not disappear - it changes role. Instead of hundreds of small parcels, you ship a few large inbound deliveries a month to Amazon's fulfilment centres, and those have their own rituals:

• FNSKU labels on every unit. The FNSKU is Amazon's internal barcode tying a unit to your seller account. Labelling during inbound picking (scan the EAN, print the FNSKU, apply, verify with a scan) is a natural WMS task - done by hand on 500 units it takes a full day and produces the mislabels that Amazon punishes with rejected receipts or relabelling fees.
• The shipment plan. Amazon decides which fulfilment centre receives which part of your delivery - one batch is routinely split across two or three warehouses in different countries. Your WMS must be able to pick and pack the delivery according to that split.
• Carton and pallet requirements. Carton weight limits, shipment labels on cartons and pallets, delivery appointments. Small details - but a rejected inbound is a week of lost sales availability.

Mature accounts are rarely all-in on one model. The typical split: fast-rotating, small-sized bestsellers go to FBA for the Prime badge, while the long tail, oversized items and seasonal products stay in FBM so you are not paying Amazon rent on slow stock. Another common pattern is the fallback listing: an FBM offer on the same ASIN as insurance, keeping sales alive when FBA stock runs out before the next inbound lands.

A hybrid raises the bar for your system, though: the same SKU now lives in two stock pools - your warehouse and Amazon's - which must be tracked separately but planned jointly. A good WMS shows both pools side by side and suggests FBA replenishments based on sales velocity, before the listing loses Prime to a stockout.

Amazon runs the most customer-friendly returns policy on the market, and the operational cost of it lands on the seller. In FBA, Amazon accepts the returns, but periodically sends you consolidated shipments of stock flagged unsellable (removal orders) - and someone in your warehouse must inspect each unit and decide: back to sale, discounted through another channel, or disposal. In FBM the return comes straight back, often unannounced and far from factory condition. In both models the WMS should run a dedicated returns zone with mandatory quality grading before a unit re-enters sellable stock - otherwise a "new" product with a half-peeled FNSKU label ships to the next customer and comes back with negative feedback attached.

A simple exercise worth doing for 5-10 representative products instead of theorising about the whole catalogue. Take a product and its selling price, then price the FBA route: referral fee, fulfilment fee for its size tier, storage at your actual rotation speed (mind Q4), inbound freight and labelling per unit. Next to it the FBM route: your pick-and-pack cost (with an efficient WMS realistically EUR 0.70-1.50 per parcel), courier cost to the destination market, returns handling. Then one commercial correction: FBA conversion tends to be higher thanks to Prime, so you are comparing not just unit costs but expected volumes too.

The result is rarely uniform across a catalogue - which is exactly why hybrids are so common. Redo the numbers every six months; Amazon adjusts its fee schedules more often than carriers adjust theirs.

The FBA-versus-FBM decision is really a decision about where your margin and your risk live: in FBA you trade part of the margin for Prime and elasticity, in FBM you accept the SLA regime in exchange for control and lower unit cost. Most honest calculations end in a hybrid, decided product by product. The common denominator of both models is the same: without a warehouse system that reserves stock in real time, returns tracking automatically and handles FNSKU labelling cleanly, an Amazon account sooner or later runs into metric trouble - and from there the road to suspended listings is short.

Weaver WMS covers both scenarios: in FBM it guards cross-channel stock reservations, shipping cut-offs and automatic tracking upload; for FBA inbounds it runs shipment-plan-based picking with scan-verified FNSKU labelling. The integration works both directly and through middleware platforms - so you can start with an integrator and switch to a direct connection when volume justifies it, without changing anything in the warehouse process itself.