Can you actually return a parcel to a British shop from Germany? We tried 31 times ( Part 5)

31 orders returned through each shop's own customer flow. What it takes to get a label, and what happens when there isn't one.

This is the third of four articles from our British study: forty nine brands, one real self funded order each wherever the shop would ship to a German address, everything sent back through the standard customer return flow with no contact identifying us as researchers. Thirty one orders placed. The first article covered the shops that would not sell at all, the second the delivery promises they missed.

This one is about what happened when we tried to send the goods back. As before, we name categories, never shops.

There are five parcels we still have. We could not give any of them back. We paid for them, they arrived in good order, and then every route back to the shop that sold them turned out to be closed. Not late, not stuck in transit, not expensive to send back. Returnable: no.

Five out of 31 orders, which is 16%, and roughly 173 euro of goods that stayed with a customer who had asked to be rid of them. A sixth parcel, worth 38 euro, was sent back to an address that turned out to be wrong and has produced neither goods nor money since - on the most inclusive reading, 6 of 31, or 19%.

In our earlier German study, where the whole transaction stayed inside one country, the return label was a friction point. Only 4 boxes in 20 contained a printed one; the rest lived behind portals, accounts and QR codes, and when one QR flow broke it produced the longest case in that dataset. Cross border, the same mechanism stops producing friction. It starts producing failure.

Key findings at a glance

5 of 31 orders, or 16%, reached a dead end. Combined value: 173 euro of goods the customer could not send back through any path that existed. Counting a sixth case where the parcel was dispatched to an incorrect address and never accounted for, the figure is 6 of 31, or 19%.

Not one shop in the test offered a genuinely free, no questions return. In Germany free return shipping was the standard, with roughly 9 flows in 10 costing the customer nothing. Cross border, every documented return was either deducted from the refund or paid out of pocket. Thirteen shops deducted. Nine required the customer to fund the shipment themselves.

The most common failure was the absence of any returns portal at all, with support as the only route back - and support was frequently slower than the return window. In one case a portal did exist and simply had no function capable of generating a label for a German address.

Support's standard answer was to point at the portal. In several cases the portal was the thing that had just refused, or did not exist.

In several cases the return solution offered was a private address in the UK, with no label, no carrier and no instructions, leaving the customer to arrange and fund an Sinternational shipment themselves.

One order produced no confirmation email, therefore no order number and no access to the portal-making the case unresolvable by design. Three support tickets were never answered at all, across multiple follow-ups, and remain open as of writing.

And one shop refunded a parcel that was never sent back. The money arrived, the goods stayed with the customer, and the shop's own records now show a return that did not happen. The mirror image of every other failure in this article, and evidence that the same broken machinery loses money in both directions.

More than half of returns worked. 16 of 31 were logged by the tester as straightforward, against 12 logged as difficult or impossible. Same border, same carriers, same fourteen day windows.

Five dead ends, five different mechanisms

None of these is the same failure wearing a different hat. They are five separate ways a returns process can stop existing.

The order that left no trace. A menswear brand operating across separate UK and EU domains, in what our tester described as a partnership arrangement between two storefronts. The order went through, the payment cleared, the parcel arrived. The confirmation email never did. The returns portal asks for an order number and there is no order number. Support on both domains was contacted and neither produced one. The parcel is still sitting here.

This is the cheapest failure on the list to prevent and the most complete in its effect. A confirmation email is not a courtesy, it is the token that proves the customer exists inside your system. Without it, every downstream process that authenticates against an order number is closed permanently, and the customer has no route to reopen it.

The shop that simply said no. A cycling retailer. We ordered from one company's website and the parcel arrived from a different company in the same group, with a different name on the box. Because the two share a support operation, there was no route to a party that would take ownership of the transaction. When the return was raised, support did not permit it. No policy was cited that we could act on. The goods have still not been sent back and that was the end of the process.

The shop with no return infrastructure and no reachable human. A menswear label with no returns portal of any kind. The only path was support, and support did not resolve it - our tester's note records that contact was extremely difficult and that no help was ever provided towards actually returning the item. Nothing was refused here either. There was simply never a mechanism, and the one channel that could have created one did not answer.

The internal handover that dropped it. A footwear brand. Delivery was fine. The return request entered the shop's internal process and never came out the other side. Communication between the people handling it broke down, the return window closed while it sat there, and the item stayed with the customer. Nobody refused anything. The failure is that a return can quietly expire inside a company that is technically still working on it.

The clock that ran out during the wait. A streetwear brand. The return was initiated in good time. Then it took long enough, waiting on responses and waiting on instructions, that the return period lapsed before the process was ready. The parcel was left with the buyer for precisely that reason.

Slow support is not a service quality issue when the return window is fourteen days long. It is a mechanism for denying returns without ever declining one. If your median first response is four days and your window is fourteen, you have built a filter, whether or not anybody intended one.

The failure that runs the other way

One case in this test deserves separating out, because it is the only one where the shop lost and the customer did not.

A country and outdoor brand. The return was difficult: support did not respond for a long stretch, and our tester's note is explicit that the delay was the shop's fault rather than the courier's. Then the refund arrived - in full, to the customer's account, while the parcel was still sitting in Germany. Nothing had been sent back. Nothing had been scanned. The money simply moved.

From the outside this looks like generous customer service. From the inside it is the same defect as every other case in this article: a returns process where nobody is checking the state of the parcel against the state of the money. In four cases that gap cost the customer their refund. In this one it cost the shop both the goods and the payment.

If your reverse logistics cannot tell you whether an item is back in the building before it authorises a refund, you do not have a returns process that is too strict or too lenient. You have one that is not connected to anything.

What the support queue actually looked like

The dead ends are the visible failures. Underneath them sits the pattern that produced them, and it repeated across most of the cases where a return was difficult rather than impossible.

Support answered only after several chasers, or never. Multiple tickets opened during this test remain unanswered after repeated follow ups.

Support's standard answer was to use the returns portal. In several cases there was no portal to use, and in one case the portal existed but contained no function capable of generating a label for a German address. Either way the customer was being routed back to the thing that had already failed, which is exactly the structural trap identified in Part 2 of the German study: an exclusion or a fault discovered at the return stage escalates to email by definition, because the self service path is what just refused them.

Where a label could not be produced, some shops supplied a private address instead. No carrier, no prepaid label, no guidance. From Germany to a residential British address, arranged and funded by the customer. Nine of our 31 returns ended up funded this way. That is not a returns process. It is the absence of one, communicated politely.

On some sites, finding any route to a human was difficult or impossible. Not slow. Absent.

And procedures differed by shop and by carrier, so nothing learned in one case carried to the next. Every return was a fresh investigation. That cost is invisible to each individual shop and enormous in aggregate.

The German study measured this as an email count: four messages to close a clean case, eleven to close a broken one, in the same market with the same carrier and the same product type. Its conclusion was that support volume is not a support problem, it is the invoice for the other failures.

Cross border, that invoice carries a second line item. Some cases do not close at eleven emails. They do not close.

The returns page said fourteen days

Every shop in this test had a returns policy. Most of them read well. Several promised generous windows and free labels.

The German study's opening line applies here without a word changed: if you audited this market by reading policy pages, you would conclude that returns are a solved problem.

What a policy page does not tell you is whether a label can actually be generated for a German address. Whether the portal recognises a guest checkout. Whether support answers. Whether the confirmation email fires. Whether the return window keeps running while the customer waits for a reply. Whether anybody checks that the parcel came back before the money goes out.

All six of those are operational states rather than policy statements, and all six broke somewhere in this test.

A failed return does not appear as a loss on any dashboard you own. It appears as a completed sale. Revenue retained, no refund processed, no return shipping cost, no restocking, no write down. In five cases here the shop's accounts record a clean transaction, while the customer records a company that took their money and would not take the goods back.

You will never see that number directly. You will see the reorder rate eventually, and you will attribute it to something else. Probably to price.

What good looked like, in the same dataset

The counter pattern was not the exception here. It was the majority.

16 of 31 returns in this test were logged by the tester as straightforward. Portal worked or a human answered, label generated, parcel dropped, money arrived. Several of these were among the fastest refunds in the entire study, and one of them turned money around on the same day the parcel was scanned.

One brand had no returns portal at all and still landed in the simple column, because a human answered, supplied a return address and clear instructions, and the case closed. The customer paid for that shipment out of their own pocket. The tester's note on that one still reads that communication with support was excellent - which tells you what customers actually price. Not free. Reachable.

That is the full spread. Same border, same fourteen day windows, same couriers, same country. Sixteen shops made it work and five made it impossible. The difference between simple and impossible was never geography. It was whether anybody on the other side was reachable.

Check this against your own operation

• Try to generate a return label for a foreign delivery address in your own portal, right now, as a guest rather than as a logged in account. Does one exist?
• What happens to a customer whose order confirmation email never sent? Is there any path back into your system that is not the order number?
• Does your return window pause while a customer waits in your support queue, or does it keep running against them?
• If your support team's standard answer is to use the portal, has anyone confirmed the portal can actually serve an international address?
• Does anything in your process verify that the goods are physically back before a refund is authorised - and does anything verify the opposite, that a refund was actually issued once they arrived?
• How many return requests in the last quarter were opened and never closed? That number, not your refund rate, is your real failure count.

Then run the awkward version of the test on yourself: one real order, one real return, and one deliberately difficult case such as a guest checkout with no account, or a lost label. Our data shows the failures cluster exactly there.

"Before you launch in a new country, generate one return label to that country's address yourself. If you cannot, your customers cannot either."

Pawel Zakielarz, CEO, ShopReturns

What to take from this

Five parcels could not be returned at all, and not one of them was refused by a policy. They failed on a missing email, a missing portal, an unanswered ticket and a return window that kept running while the customer waited. Sixteen other shops, on the same border and with the same couriers, made the return simple.

The difference was never geography. It was whether anyone on the other side was reachable.

We will walk your return flow the way a customer does and show you the point where it breaks.

About the study: live test, May to August 2026. 49 British brands selling direct to consumers, approached as an ordinary shopper with a German delivery address. 31 real, self-funded orders, each returned through the shop's standard customer flow, with no contact identifying us as researchers. Results are reported by category, and no individual brand is named. Full methodology is in Part 4.