Inside an Amazon FBA Inspection Report: What the Sample Covers and What Counts as a Defect

"Pallet's here, cartons look sound. You want us to open any of them?" That is a receiving clerk with a forklift idling behind him. Answer no and the lot goes to a rack, then to Amazon, and the first person to open a box is a customer. Answer yes and three questions land on you at once: how many units, out of which cartons, and checked against whose list.

An inspection report is that phone call written down: who looked, how many units they opened, what they found, and what should happen to the lot. The usual habit is to search the file for the word pass and archive it. That was survivable while Amazon still bagged and stickered loose units at receiving and billed you for the trouble. Since January 2026 it does none of that, so a factory mistake either gets caught in the building that receives the pallet or it ships. The barcode half of the same change is what ended commingling for resellers.

A report is worth what you paid for it only if you can argue with it.

The header block, and the field most sellers skip

The top of the report identifies the lot. It looks like paperwork. It decides whether the rest of the document means anything.

• Lot identity. PO number, supplier, ASIN, SKU and FNSKU. A report that names a product but no PO cannot be tied to a payment.
• Declared against received. Cartons expected, cartons received, units on the packing list, units counted.
• Carton condition on arrival, photographed before anything is opened. This line separates transit damage from factory damage.
• Sample traceability. Which cartons were opened, and how they were picked.
• Checklist version and inspector. Same lot, same list, same table.

Traceability is the field sellers skip and suppliers count on. A report that gives a percentage inspected, without saying which cartons it came from, is not evidence. Units pulled from the two cartons nearest the dock door tell you about those two cartons alone. Units pulled across the pallet, top layer and bottom, tell you about the lot.

The sample line: how many units were really looked at

Sampling runs on AQL, the acceptance quality limit method inspection firms take from the ISO 2859 standard. The mechanics are simple once someone writes them down:

1. Lot size plus a chosen inspection level gives a code letter.
2. The code letter gives a sample size.
3. For each defect class, the table gives an accept number and a reject number.
4. Defects in the sample are counted by class and compared against those two numbers.

Inspection levels are the dial. Moving up a level buys a larger sample and a better chance of catching a defect that occurs at a low rate. Severity shifts with supplier history: a new factory or a lot that failed last time gets tightened inspection, and a run of clean lots earns a reduced plan until the first miss.

One sentence in the report gets misread more than any other. A pass means the defects found in the sample stayed at or below the accept number. The units nobody opened were never part of the question. AQL bounds risk, it does not remove it, and even a full unit-by-unit screen catches a share of problems rather than all of them, because attention degrades over a pallet.

Depth of check	What it answers	What stays unknown	Where it fits
Receiving check only	Did the cartons arrive intact	Everything inside a sealed carton	Repeat orders, supplier with history
Spot check	Are these the goods you ordered	Any defect rate below the obvious	Low value, low return rate
AQL sample by class	An accept or reject call with numbers behind it	Defects in the units nobody opened	New supplier, new product, first run
Unit by unit	Which specific units are bad	What sealed packaging hides, anything lab level	High price, fragile, regulated, or a mixed lot

You set that depth, and the warehouse quotes against it. Each level carries its own rate per unit, which turns the choice into arithmetic instead of taste: pick the depth per supplier and per production run, then record the setting somewhere you will see it when the next PO goes out. The depth that earned its keep on a first order is money burned on the fourth. Our product inspection service runs from a percentage sample through to unit by unit for that reason.

The defect list belongs to you

The defect table can only hold what the checklist told the inspector to look for. That checklist should exist before the container lands, and you should write it, with the warehouse editing rather than authoring. Useful lists carry a written definition of each defect, a measurable threshold where one class turns into the next, and reference photos so two inspectors classify the same scratch the same way.

• Every claim on your detail page that has to be true on the unit: color, size, model, variant, count per pack.
• Contents: accessories, manuals, inserts, cables, mounting hardware.
• The point where a cosmetic mark becomes a reason not to ship, split by surface. A mark on the back panel and the same mark on the front are not one defect.
• Barcodes: readable manufacturer barcode, correct FNSKU, label position, nothing printed across a seam.
• For dated goods, the remaining shelf life you accept on arrival.

Skip this and the warehouse uses a default checklist built for the average product. Nobody sells the average product.

The list should also get longer over time. Every return reason code that repeats and every question a customer asks twice belongs on it. A checklist written once at launch records the problems you imagined, not the ones your product has.

Reading the defect table by class

Inspection standards sort defects by severity: critical for a safety or regulatory failure, tolerated at zero; major for a unit that is not saleable or will come back; minor for a cosmetic deviation a customer will live with. Severity alone does not tell you what to do next, so a prep center report adds a second axis: findings are grouped by where the defect lives, and every group holds all three severities.

Cosmetic findings

Scuffs, scratches, print misalignment, color variance against your approved sample, abrasion from carton rub. These should appear with a photo and something in frame for scale.

Shipping them rarely costs a refund. It costs a three star review and a return filed as not as described, which drops the unit into unfulfillable stock and leaves a paragraph on your listing that outlives the lot.

The decision: accept, or sort. Sorting pays only when the defect is visible without unpacking. If someone has to open a sealed box to see the mark, sorting the lot costs more than the returns from the bad fraction. It is the one class where accept and move on is often right, and the report gives you a defect rate to decide with instead of a hunch.

Functional findings

The unit does not do what it is sold to do. A switch that will not hold, a zipper that jams, a pump that never primes, a battery contact that was never connected. Function checks are the slowest line in any inspection, so the report should state how many units were tested and against which test.

The cost is a refund plus the return trip, and in categories with high return rates Amazon adds a returns processing fee. The unit comes back unsellable, so you paid for the goods, the inbound, the fulfillment and the return with nothing left to resell.

Rework is the right answer when the failure sits outside the product: a missing hardware pack, an accessory in the wrong box, an insert that fell out during packing. When the failure is inside the product, hold the lot and open a claim instead. A functional defect that turns up more than once or twice in a sample is a process problem on the production line, and sorting at a warehouse does not repair a production line.

Packaging findings

This is where prep and inspection meet, and it is the class that changed most once Amazon stopped doing prep work at receiving. Most of the weight sits in a short list: retail packaging that will not survive a conveyor, a poly bag missing or wrong on an item that needs one, a label a scanner cannot read, and a multi pack with nothing on the outside saying the units are sold as a set.

Bag size, the seal, the suffocation warning and whether a code still scans through film come straight from the poly bagging requirements. Label findings work the same way, and it pays to know what a bad FNSKU label does downstream before you decide how hard to push. On the report itself you only need three things per finding: the class, the rate, and a photo clear enough to argue with.

Nearly everything in this class ends as rework, and it is the cheapest class to fix, because the fix happens in the building that already has the goods in hand. Bag it, warn it, wrap it, relabel it. Leave it and the bill arrives in another currency: a shipment turned away at the door, a delay before units go live, a charge nobody budgeted for.

Compliance findings

Wrong barcode, an FNSKU pointing at a different ASIN, missing country of origin, absent warnings, hazmat markings missing from a product that needs them, adult products without opaque packaging, dated goods with less remaining life than your rules allow.

This is the only class with no accept option. Compliance defects do not produce a bad review. They produce blocked ASINs, inventory held as unsuitable, document requests on a clock, and account health consequences that outlast the lot. Dated goods run stricter than most sellers expect, so build your remaining-life threshold from Amazon's own wording rather than from what a supplier promises.

The decision: hold, then rework or return. If the fix is a label, a warning or a bag, it becomes a rework line. If the fix is a certificate, a marking printed on the product itself, or a change to the formulation, it goes back to the supplier. Nothing in this class ships on probably fine.

The photo file

Photos are the part of the report with value outside your office. A defect described in a sentence is an opinion. A defect photographed against the carton it came out of is a claim.

• The pallet and cartons as delivered, before anything is opened.
• Each sampled carton with its number visible in frame.
• The defect itself, close, with a ruler or a coin for scale.
• The whole unit, so the defect can be located on it.
• The label, close enough to read every character.

Suppliers argue with defect rates. They argue much less with a carton number and a date stamp.

The recommendation line: five outcomes and what each costs

Outcome	When it is the right call	What it costs you	What it protects
Accept as is	Minor cosmetic findings, low rate, no compliance items	A share of returns and review risk	Your launch date and restock rhythm
Sort and ship the good units	Defect visible from outside, rate high enough to matter	Labor per unit, plus the time the lot sits	Most of the lot and the selling window
Rework	Packaging, labeling, outside-the-product failures	Labor and materials per unit	Inbound defect charges and refused shipments
Hold and claim	Functional or compliance failures pointing at the factory	Storage while the argument runs, plus a stockout	The purchase price, if your paperwork holds
Return or dispose	Unfixable, unsellable, or unsafe to list	The goods and the freight already spent	Account health and the listing

Sorting and rework are billed by unit or by hour, so both pay only when the defect rate sits in the middle. At a very low rate you accept. At a high rate you stop sorting and start negotiating, because paying labor to salvage a lot the supplier should be paying for is a transfer, not a saving. Inventory frozen during that argument still needs a shelf, which is what short term storage is for.

What the report cannot tell you

Prep centers leave this part off their sales pages. It decides how much risk you still carry after a clean report.

• Laboratory properties. Inspection is visual, dimensional and functional. Chemical composition, flammability, lead content, food contact safety and battery testing come from a laboratory, not a bench at a warehouse.
• Whether a certificate is real. A warehouse can confirm a document is present and matches the model on the box. It cannot confirm the test behind it happened.
• Whether the factory is stable. One passing lot is evidence about that lot. A factory running a steady defect rate still passes a sample on a good day. Only repeat inspections show the trend.
• Durability and shelf life. Nobody learns in an afternoon whether a hinge survives six months of use.
• Authenticity. A competent counterfeit passes a visual check. Sourcing controls handle that. Inspection does not.

Risk that survives a clean report is still risk. Inspection narrows it and documents it. Pricing your product as though it vanished is how sellers get caught out later. Keep a short list of what your checklist cannot reach, and decide once per product where each item gets handled instead: a lab report before the first PO, a supplier audit, a unit held back from each run so you have a reference on the shelf when a customer says the last one was different.

What a missed defect actually costs

The case for paying separately for inspection is arithmetic, not a speech about quality. A defect that ships carries a stack of costs, and most sellers count only the first:

1. The refund, which hands the sale price back to the customer.
2. The return trip and the returns processing charge, which Amazon applies in categories with high return rates and on every return in apparel and shoes.
3. The unit itself, which comes back unsellable and turns into a removal order or a disposal.
4. Storage along the way, because defective stock ages in the network like any other and picks up the aged inventory surcharge.
5. The replacement units and a second inbound shipment, placement charges included.
6. The review, which does not expire when the lot sells out.
7. The metrics: order defect rate, listing complaints, and in the compliance classes a blocked ASIN.

Amazon revises those charges more than once a year, so do the arithmetic on the numbers sitting in your own Seller Central fee reports this month rather than on a figure someone quoted last season. For a cheap item with a low return rate, light sampling wins on the math. For anything fragile, regulated, expensive, or new from a supplier, the deeper check pays for itself the first time it stops a bad lot at the dock.

Reading between the lines

Is inspection the same as the receiving check that comes with prep?

No. A receiving check confirms the cartons arrived, the count matches the packing list, and nothing is visibly wrecked. Inspection opens units, applies your defect list, tests function, and produces an accept or reject decision with photos behind it. Most centers include the first in intake and quote the second separately, which is why two quotes on one lot can look so different. It is also why sample size, inspection level and severity belong in the quote you approve, not in a conversation after the pallet lands.

Who pays when the inspection finds a bad lot?

You pay the warehouse for the inspection and any rework, then recover from the supplier if your paperwork supports it, which is what the header block is for. A claim backed by a PO number, carton numbers, dated photos and a defect list the supplier agreed to in advance tends to get settled. A claim backed by an email saying the goods were bad usually does not.

Can units already sealed in retail packaging be inspected?

Partly. Everything outside the seal can be checked: labels, barcodes, packaging condition, printed dates, contents where the box has a window. Anything inside means opening the box, after which the unit is either resealed to retail standard or no longer new. Decide before the lot lands how many sealed units you will sacrifice.

Does a clean inspection protect me from an inbound defect charge?

It removes the causes inside your control. Missing labels, missing bags, absent warnings, wrong barcodes and prep that never happened at the factory are exactly what a packaging and compliance check catches. Charges tied to routing, appointments and shipment accuracy sit with whoever books the freight, and those are worth reading up on separately before your next inbound plan goes out.