Amazon Returnless Refunds Reach Five EU Stores: Sellers Cap Items at 75 Euros

The Analysis

On Wednesday morning a seller I sit with in Hanover watched a return request come in at 09:14. At 09:16 it was closed. Refund issued, goods stay with the buyer, no label, no parcel, no scan, nothing. For eight years she has run her EU business on the assumption that a return takes nine to eighteen days: the buyer has the statutory fourteen days to get it in the post, then two to five days of transit, then one to three days at the fulfilment centre before anyone grades it and lets the money go. That corridor just became two minutes.

Amazon has switched on FBA Returnless Resolutions in Germany, France, Italy, Spain and the United Kingdom. Enrolment is voluntary and it is the seller who sets the ceiling: EUR 1.00 to EUR 75.00 per item, GBP 1 to 75 in the UK. You can cover every ASIN and carve out a CSV exclusion list, or you can keep it narrow and name the ASINs one by one. Every request is still assessed case by case.

Two lines in the small print decide whether this is a gift or a leak. Goods resolved this way do not qualify for inventory reimbursement, so the unit is gone and there is no claim to file. And the order still counts toward your return rate; the high-return-rate fee is waived, the number itself is not. I care about a third thing, which nobody has printed yet: those two minutes are where you lose the only window you ever had to do something about a return, and the bill for them arrives at a fulfilment centre five or six weeks later, in a week when you are already chasing a stockout.

Four numbers are worth pinning to the wall before anyone argues about it. Five marketplaces: Germany, France, Italy, Spain, the United Kingdom. One ceiling you type yourself: EUR 1.00 to EUR 75.00 per item, GBP 1 to 75 across the Channel, and at EUR/GBP 1.15, which is my assumption so check your own rate, the British ceiling is about EUR 86, meaning a single global setting in euros quietly gives the UK store a wider dial than the four continental ones. Two coverage modes: every ASIN with a CSV exclusion file, or a named-ASIN list, and nothing in between. And two consequences buried in the detail: no inventory reimbursement on those units, and the order still lands in your return rate with only the high-return-rate fee switched off. That is the instrument in full. It fits in four sentences, which is roughly why everyone is treating it as a small decision.

Why this week, and why these five stores? The reverse mile is the most expensive mile in e-commerce, and in Europe it is structurally worse than in the United States because the statutory withdrawal right hands the buyer fourteen days to change their mind about a product that was never faulty. Amazon has carried the cost of that corridor for years as the price of a frictionless customer experience, and it has now built a mechanism that hands both the decision and the bill back to the seller, as an opt-in. Voluntary is the detail to stare at. A mandatory programme makes Amazon the party that decided your forty-euro kettle was not worth the trip back. A voluntary one makes you that party. You type the number, you own the outcome, the customer gets a refund in two minutes, and Amazon's own return metrics look exactly as they did last quarter. Opt-in is not generosity. It is the place the liability gets parked.

Let me put the old corridor on a clock, in hours, because hours are the only unit that means anything on a return. H0 the buyer opens the request. Somewhere between H+24 and H+72 they find a box and get it to a drop point. Between H+96 and H+192 the parcel moves. Between H+216 and H+288 the fulfilment centre receives it, opens it, grades it and files it as sellable, unsellable or customer damaged. Between H+288 and H+432 the refund fires and the unit either goes back on the shelf or into disposal or liquidation. Call it nine to eighteen days, and inside that corridor sat four or five moments where a person could still change the outcome: a message at H+6 offering a replacement instead of a refund, a partial refund at H+24 that closed the case for eight euros instead of thirty, a grading decision at H+240 that recovered a unit the buyer had sworn was broken.

Returnless deletes the corridor. H0 the request. H0 plus roughly two minutes the decision, the refund, end of story. What is being sold as a customer-experience upgrade is, on the operations floor, the removal of a fourteen-day intervention window, and if your job is to protect the customer and still protect the margin, that window was the job. The hour that actually blows is not in the return flow at all, and this is where I stop agreeing with the coverage. The return flow gets shorter, quieter and cleaner, and for the first three weeks every dashboard you own will tell you it was a brilliant idea. The hour that blows is the restock hour, five to seven weeks out, when the units that used to come back into sellable stock do not arrive, and the gap shows up on a shelf in a building nobody was watching.

Who wins and who loses, in shape first. The win case is cheap, heavy and returned often: a fourteen-euro bathroom accessory at 1.8 kilograms. The reverse leg costs real money, the landed cost of the unit is EUR 3.70, and if it comes back there is a fair chance it is chipped, wet or missing a fitting and goes straight to disposal. Paying to fly that thing home is worse than letting it stay. The loss case is small and expensive: sixty-nine-euro earbuds in a small standard envelope. The reverse leg on that size band costs about what it costs on a nine-euro cable, so there is almost nothing to save, but the unit you hand over cost you eighteen euros.

Every keep-it you approve on that SKU is a donation of roughly eleven euros net. The trap is that the control Amazon gives you is a price ceiling, and price is the variable with the weakest correlation to the answer. Size band drives what the reverse leg costs. Landed cost drives what you are giving away. Price is what the box lets you type.

When each part of this lands. At T+0 the refund leaves your money and the unit leaves your sellable pool permanently, and your inventory reconciliation is now wrong by one with no inbound document to correct it. At T+1 day your return rate moves by precisely the amount it would have moved anyway; the fee relief shows up on the invoice, the metric does not move in your favour. Between T+20 and T+30 the conversion completes, because the requests filed in the first days after switch-on come out of orders shipped two to four weeks earlier, so the true run rate only appears once the order cohort has aged. Between T+28 and T+45 the abuse curve arrives. Buyers learn ceilings; they always do, and it always takes two to six weeks. Between T+35 and T+50 the shelf goes short. Around T+8 months the warranty tail shows up: a unit you refunded and never received comes back as a service claim with your serial number on it.

Now the thing I have not seen in any of the coverage, and it is not the reimbursement line everyone has already posted about. The setting is a single scalar: one maximum item price, per store, in one currency. The decision it stands in for is a two-dimensional grid, size band crossed with landed cost, where size band sets what the reverse leg costs and landed cost sets the value of the unit you are abandoning. No single number can express that grid. Worse, the range Amazon lets you type runs straight past the point where the economics turn, for exactly the SKUs the filter is built to catch. A price ceiling does not select for cheap to replace. It selects for expensive, and expensive things sitting in small size bands are the worst candidates this programme has.

Put it on paper with my assumptions and then swap in yours. Assume a steady state where your monthly volume is set by traffic and conversion rather than by how many units you hold, so a unit recovered in sellable condition simply replaces one you would otherwise have bought. Write R for the all-in cost of the reverse leg: return handling, the trip back into the FC, grading, and an average of refurb or disposal. Write s for the share of returns that come back sellable and C for the landed cost of a replacement unit.

Taking a return back costs you R and hands you back a unit worth s times C; letting it go costs you the whole of C. So returnless comes out cheaper whenever R is greater than s times C, and the break-even landed cost is R divided by s. My numbers, from my own floor: s at 70 per cent; R at EUR 6.50 for a small standard unit, EUR 11.50 for a large standard, EUR 19.00 for bulky or heavy, all-in. That puts break-even landed cost at EUR 9.29, EUR 16.43 and EUR 27.14.

At a landed cost ratio of 26.4 per cent, my assumption of EUR 7.40 of landed cost on a EUR 28 average item, those convert to selling prices of about EUR 35, EUR 62 and EUR 103. On a small standard SKU this programme goes bad at thirty-five euros and the setting runs to seventy-five. On a bulky SKU the entire legal range is profitable and seventy-five is timid. One number cannot serve both, and the dial is pointing at the wrong end of the catalogue.

Run it on a month I can defend. Twelve thousand units shipped into the five EU stores, average price EUR 28, landed cost 26.4 per cent or EUR 7.40, return rate 8 per cent, so 960 return requests. Split by size band: 55 per cent small standard, 528 requests; 33 per cent large standard, 317; 12 per cent bulky, 115. Average price inside each band is EUR 22, EUR 30 and EUR 48, so landed cost is EUR 5.81, EUR 7.92 and EUR 12.67. The saving per request is R minus s times C: EUR 2.43 on small, EUR 5.96 on large, EUR 10.13 on bulky. Over the month that is EUR 1,283 plus EUR 1,889 plus EUR 1,165, or EUR 4,337 saved, on the assumption every request clears.

Now the tail nobody prices. Inside the small standard band, 15 per cent of requests sit on items priced EUR 50 to EUR 75, average EUR 62, landed EUR 16.37, so s times C is EUR 11.46 against an R of EUR 6.50: you lose EUR 4.96 a request, on 79 requests, EUR 392 a month, and you would still have that money if those ASINs had never been on the list. Fixing the list is worth about EUR 4,700 a year on this catalogue, and on a catalogue skewed toward small and expensive it is the entire result.

Flip the mix and the sign flips with it. A catalogue that is mostly EUR 69 earbuds: 300 requests a month, landed EUR 18.22, s times C EUR 12.75, R EUR 6.50, so minus EUR 6.25 a request, EUR 1,875 a month, twenty-two and a half thousand a year, paid for the privilege of a setting that sounded modern when you turned it on.

The abuse curve points the same way, which is why this belongs on your desk this week rather than next quarter. Assume the return rate goes from 8.0 to 9.5 per cent within six weeks of switch-on, 180 extra requests a month on twelve thousand units. Assume the extra requests concentrate where retained value per unit is highest, 60 per cent of them in the EUR 50 to EUR 75 small-standard band. Each of those hands over EUR 16.37 of landed unit and saves EUR 6.50 of reverse cost, so EUR 9.87 runs against you; 108 of them is EUR 1,066 a month.

The other 72 land in the cheap heavy bands where returnless is genuinely cheaper, call it EUR 5.69 saved each or EUR 410, so the abuse term nets out around EUR 650 a month against you, and it does not appear in week two. It appears in week six. Judge the pilot on day ten and you will roll it out to the whole catalogue in time to learn the shape of it in December.

The inventory clock is the one that actually hurts, and it is arithmetic rather than opinion. Under the old rules 70 per cent of returned units came back sellable: 0.70 times 960 is 672 units a month of supply you did not have to buy. Switch the whole catalogue over and that supply stops, on a lag, because requests age out of orders shipped two to four weeks earlier. Q4 sell-through in our network is running at about 1.6 times baseline, my assumption, so call the shortfall 1,075 units a month, roughly 36 a day. My replenishment loop from a factory in Guangdong to a fulfilment centre in Germany is 38 days: 30 days at sea, 4 days of port and forwarder, 4 days to receive-ready at the FC.

The reorder point has to go up by 36 times 38, about 1,370 units, on the day you flip the switch and not on the day you notice the gap, because by then you are 38 days too late. Against a ten-day safety stock of, say, 6,400 units, a monthly hole of 1,075 empties the shelf somewhere in week six or seven, which on this calendar is the last week of November. There is no worse week of the year to be empty.

And then the part that makes me reach for the CSV file instead of the slider: there is no filter on return reason. Enrolment is by price and by ASIN. A request coded defective, a request coded arrived damaged and a request coded no longer needed all arrive at the same assessment, and the assessment is Amazon's, described in one line as case by case. So a unit that came back the first time because the housing cracked can be resolved returnless and stay in a kitchen you will never inspect. If there is a lithium cell in it, that cell is now in a drawer in a stranger's house with your brand on the outside and your name on the compliance file.

If it carries a serial number, it comes back to your service team eight months from now as a warranty claim, and because you never received it, nothing in your system says refunded, do not service. My triage rule is blunt and I do not negotiate it: anything with a cell, anything with a serial, anything under a warranty registration, anything you carry a take-back obligation on in Germany or France, comes off the list permanently, and comes off by CSV rather than by hoping the assessment catches it.

The reimbursement waiver deserves more than the one line it has been getting. In the normal flow, if a unit vanishes between the buyer and the FC, or comes home as a different product in the same box, you file an FBA inventory reimbursement claim and you get something back. Under returnless there is no event to claim against: nothing moved, nothing was lost, nothing to file. That sounds tidy and is not tidy at all, because what you have actually surrendered is the evidence trail. Your record of a sixty-two-euro blender leaving your business is one refund line and no goods movement.

If a chargeback or an A-to-z claim follows, you argue from a refund instead of a receipt. The credit-note and VAT-adjustment end of a refund with no goods behind it is Derek Xu's ground rather than mine, but the operational half is mine: split these refund lines out in your own reporting on day one, and stop your finance team filing reimbursement claims against them out of habit, because they will, and they will spend the time and then lose the argument.

Where I could be wrong, and you should test this before you borrow my numbers. Everything above rests on one assumption: recovered units substitute for purchases, because your volume is set by traffic and you are never inventory-constrained. If you are the opposite, permanently short, demand running ahead of supply, listings going dark for low inventory, then a returned sellable unit is not a replacement, it is an incremental sale, and its value is the full contribution, price minus fulfilment fee minus refurb, not the landed cost. Redo the small standard band that way: EUR 28 price, EUR 4.20 fulfilment fee, EUR 2.00 refurb, so EUR 21.80 of contribution, times the 70 per cent sellable rate, EUR 15.26, against an R of EUR 6.50.

Taking the unit back is now worth EUR 8.76 more than letting it go, and returnless is the wrong answer for that SKU even at twenty-two euros. Same product, same arithmetic, opposite decision, purely because of how tight your inventory is. Test that condition first, on one SKU, for two weeks. The second thing that flips it is the abuse gradient, and my 1.5 percentage points inside six weeks is an estimate from watching the same pattern run in the United States, not a published figure. If your category is one where nobody wants to keep the thing, consumables, anything with no second-hand value, anything bulky enough to be a nuisance, the abuse term may be near zero and the ceiling can run higher. If your category is small, desirable, easy to resell and priced near seventy euros, assume worse than mine and set the ceiling below forty.

So: which parcel do I rescue first, which customer do I steady first? The parcels are easy once the grid is on paper. Enrol the EUR 48 bulky SKU today and never enrol the EUR 69 earbuds, no matter what the ceiling permits. The customers are harder, because the tool cannot see them. Returnless is the fastest service recovery that exists, a case closed in two minutes instead of twelve days, and for a repeat buyer in your top five per cent by lifetime value that is worth more than the six euros of margin arithmetic says it costs.

You cannot target it by buyer, though, because the setting only knows ASINs, so you approximate: find the ASINs your best customers over-index on, let those run at a loss on purpose, and write that down in your own numbers so nobody files it as an accident. Then judge the whole thing on day 45 and not day 10, with three tables side by side, return rate, return reason mix and contribution per order. If return rate is up more than 1.5 points and contribution is down, pull the ceiling down ten euros and re-read in fourteen days.

If contribution is up and the reason mix has not moved, widen it. What you must not do is roll it out across the catalogue in week two because the first fortnight looked clean. The first fortnight always looks clean. The clock does not care about your excuses, and this one runs for six weeks.

↑ Back to the key points

— By Nadia Pryce