Managing Amazon Returns at Scale
At twenty orders a day, returns are an annoyance you absorb. At two hundred, they are a second business running inside your first one, and most sellers do not notice until the write-offs show up in a year-end P&L. This is the operational playbook. If you would rather hand it over, see our returns management service.
Returns stop being an exception and become a process somewhere around 50-100 orders a day. The sellers who handle it well do three things: they inspect and grade every return within 72 hours instead of letting a pile build, they resell the 60-70% that is genuinely resellable rather than defaulting to disposal, and they read return reasons as product feedback so the rate itself comes down. Doing none of this is a quiet five-figure leak.
What's in this guide
When returns become a process
There is a specific point at which returns change character. Below roughly fifty orders a day, a return is an event: you look at it, you make a judgement, you deal with it. Somewhere above that, returns arrive faster than anyone makes individual judgements about them, and in the absence of a process they go into a corner.
That corner is where the money goes. Not dramatically, which is why it survives for years: a few units a day that could have been resold, quietly ageing until nobody can remember what was wrong with them, at which point disposal is the only honest option. A seller doing two hundred orders a day with a 10% return rate is handling twenty returns daily, around six thousand a year. If half of those could have been resold and were not, on a $25 item, that is a mid five-figure number walking out of the business annually.
The fix is not clever. It is having a defined process that runs every day whether anyone feels like it or not.
The four buckets
Every returned unit belongs in one of four buckets, and the entire job is deciding which one, quickly and consistently.
Resellable as new. The unit is genuinely unused, in original packaging, with nothing missing. This is more common than sellers expect, because a large share of returns are change-of-mind, wrong-size or ordered-by-mistake rather than anything wrong with the product. These go straight back into sellable stock.
Resellable after rework. The product is fine but the packaging is opened, scuffed or missing. Repackaging, a new poly bag, a replacement insert or a fresh FNSKU label makes it sellable again, sometimes as new and sometimes as used in very good condition. This bucket is where most of the recoverable value sits, and it is the bucket that gets lost when nobody is doing the rework.
Sell as used or liquidate. The item works and is honestly describable but is not new. Used-condition listings, liquidation channels and clearance lots all recover something. Something is considerably better than nothing, and an honest used listing is far cheaper than a dishonest new one.
Dispose. Genuinely broken, unsafe, expired, or in a category where a returned unit cannot legally or ethically be resold. Some products belong here and pretending otherwise creates a worse problem than the write-off.
The discipline is that every unit gets allocated to a bucket within seventy-two hours. Not perfectly, consistently. A slightly conservative call made quickly beats a perfect call made in six weeks, because the unit is still worth something now.
Why 60-70% is resellable
Across client inventory we see sixty to seventy percent of returned units going back into sellable stock after inspection. Sellers consistently guess lower, usually because their mental model of a return is a broken item, when the modal return is actually an unopened box.
Amazon publishes return reasons, and the distribution tells the story. Change of mind, no longer needed, wrong size ordered, better price found, arrived too late: none of these involve anything being wrong with the unit. Even item-defective returns include a significant share of user error and setup confusion rather than genuine faults.
The rate varies by category in predictable ways. Apparel returns heavily but returns cleanly, because the usual reason is fit. Books return cleanly too. Electronics return less often but less cleanly, because a return is more likely to involve an actual fault and more likely to need functional testing before you can honestly resell it, which is why serial number capture matters in that category. Anything with a hygiene or safety dimension has a floor below which resale is not appropriate regardless of condition.
What all of this means practically: assuming returns are losses is an expensive habit. Inspecting them is cheap, and the majority turn out to be inventory.
FBA returns vs FBM returns
These are different problems and conflating them causes bad decisions.
FBA returns are handled by Amazon at the front end. The customer returns to a fulfillment centre, Amazon processes the refund, and then Amazon makes its own judgement about the unit's condition. Units judged sellable go back into your available inventory. Units judged unsellable sit as unfulfillable stock waiting for you to decide between a removal order and disposal.
The trap in FBA is that Amazon's condition judgement is made at speed by someone with no stake in your margin, and it goes both ways. Units get marked unfulfillable that are perfectly fine, and occasionally the reverse. Leaving unfulfillable stock sitting in a fulfillment centre costs storage and eventually gets disposed of automatically under whatever settings your account has. Pulling it out via a removal order to a 3PL that will actually inspect it is usually the better call at volume, because a human looking at the unit with your P&L in mind reaches different conclusions.
FBM returns are entirely yours, end to end. You set the return address, you receive the unit, you make every judgement. This is more work and more control, and it is where the margin advantage of merchant-fulfilled selling is most often thrown away. Sellers switch to FBM for the fee saving on oversize or slow-moving items, then absorb returns they never processed properly, and wonder why the saving did not show up. Our FBM fulfillment page covers the dispatch side of the same equation.
Getting reimbursed for what Amazon lost
Returns processing has a second recovery stream that is separate from reselling the units: claiming for what went wrong inside Amazon's own system.
The common cases are a customer refunded but the item never came back, a unit lost somewhere between the customer and the fulfillment centre, a unit damaged by Amazon rather than by the customer, and a return processed against the wrong order. Each of these is reimbursable in principle. In practice reimbursement is mostly claim-driven: you have to identify the discrepancy and raise it, within the time limits that apply to that claim type.
At low volume this is not worth structured effort. At several thousand returns a year it is a real line of recovery, and it is only possible if you are reconciling what was returned against what was refunded. That reconciliation is a by-product of having a proper returns process and is impossible without one. Our guide on FBA reimbursements goes through the claim types in detail, and there is a reimbursement estimator if you want a sense of the scale before committing time to it.
Return reasons are free product research
Everything above is about recovering value from returns that have already happened. The higher-leverage work is reducing the rate, and Amazon hands you the data to do it.
Return reason reports tell you why, per ASIN. The patterns are usually blunt once you look at them in aggregate rather than one complaint at a time:
Wrong size dominating is a listing problem, not a product problem. Your size chart is wrong, ambiguous, or missing the measurement buyers actually need. Fixing a size chart is free and permanently reduces returns on every future order.
Not as described means your images or copy are overselling. Sometimes it is a colour that photographs warmer than reality, sometimes it is a material description doing some work, sometimes it is a scale problem where buyers do not realise how small the item is. An honest listing converts slightly worse and returns substantially less, which is almost always the better trade.
Defective clustering on one SKU or one batch is a supplier conversation. If returns spike after a particular inbound shipment, that is a quality control signal with a lot number attached, and it is far cheaper to address at the factory than to keep absorbing.
Missing parts or accessories is usually a packing or kitting problem, which is fixable at the warehouse and invisible until someone correlates the reports.
Reviewing return reasons monthly, at the ASIN level, is one of the highest-return hours in an Amazon business. It is also one of the most consistently skipped, because the data is not in front of you unless you go and get it.
Building the process
A workable returns process has five parts and none of them are complicated.
One destination. Every return goes to one address where someone will actually open it. Returns scattered between a fulfillment centre, a home garage and a 3PL are returns nobody owns.
A daily inspection slot. Not when the pile gets big. Daily, so the pile never exists, and so the seventy-two hour window is met without anyone having to make an effort.
Written grading rules. What counts as resellable-as-new for your products, specifically. Written down, so two different people reach the same conclusion on the same unit. This is what makes the process survive staff turnover.
A rework station. Poly bags, replacement inserts, label stock, cleaning materials. The bucket-two units need somewhere to be fixed, and if that place does not exist they become bucket-four units by default.
Reconciliation. What came back, against what was refunded, against what went back into stock. This is what surfaces reimbursement claims and what tells you whether the process is working.
At scale you also want the return reason data flowing somewhere a human reads monthly, because otherwise the whole operation is recovery with no prevention.
When to outsource
Returns processing outsources well for a specific reason: it is repetitive judgement work against written rules, performed on physical units that are already arriving at a warehouse. If your inventory is in a 3PL and your returns are not, you have split one process across two locations for no benefit.
The honest threshold is somewhere around fifty to a hundred orders a day, or whenever you notice returns accumulating rather than being processed. Below that, a disciplined hour a day handles it. Above it, the daily inspection slot stops happening reliably and the pile returns.
We process returns at our own warehouse in Austin, Texas: inspected within seventy-two hours, graded against rules we agree with you, reworked where rework recovers the unit, and put back into your sellable stock, with the ones that cannot be saved reported rather than silently binned. See how returns management works, or get a quote with your actual return volume and we will tell you what it costs.