J Janson Wang · CEO & Founder, ASG Dropshipping · 18 min read
Quick Answer
Returns get recorded at the service desk. But many preventable, product-related returns become detectable much earlier — during production, at a pre-shipment check, or at the packing bench. The real question is not only how efficiently a return gets processed. It is where the problem could first have been intercepted.
One scope note first. This article is about preventable, product-related returns: physical damage, wrong variant, and product-versus-description mismatch. It is not about all returns. Fraud, buyer’s remorse and size preference live somewhere else entirely.
The National Retail Federation and Happy Returns put 2025 US retail returns at $849.9 billion, or 15.8% of annual sales, with an estimated 19.3% of online sales returned. Those figures come from a survey of 358 professionals at US merchants over $500 million in revenue, plus 2,006 consumers — not from independent stores. The number is not yours. The mechanism is.
Post Contents (12 sections):
- Quick Answer
- Returns Are Bigger Than the Service Desk — and Whose Numbers These Are
- The Location Model: Cause, Earliest Catch, and Actual Discovery
- Why Visibility Doesn’t Move the Number — and What a Late Catch Costs
- Per-Order vs Per-Batch: Why Pre-Shipment Inspection Advice Doesn’t Fit You
- From Return Reasons to Locations: Audit Your Last 50 Returns
- Steel-manning: The Case for Treating Returns as a Service Problem
- Frequently Asked Questions
- Final Thoughts
- About the Author
- External Sources
- ASG Data Note
Key Takeaways
- Returns get recorded at the service desk. Many preventable, product-related returns become detectable earlier — in production, at pre-shipment inspection, or at packing.
- A return cause has three coordinates: where it was caused, where it could first have been caught, and where it actually got found. Most arguments about returns confuse the first two.
- Not everything is catchable at a bench. Intermittent faults, durability and compatibility problems surface only in use, which is one reason the service desk stays necessary.
- Visibility tooling tells you a defect exists. It does not move the point where you could have caught it.
- Pre-shipment inspection advice is written for people approving a batch. You ship one order at a time.
| Half of the problem |
Where it happens |
Who usually writes about it |
What it cannot explain |
| Upstream |
Supplier line, pre-ship check, packing bench |
Inspection firms, sourcing agents |
Fraud, remorse, recovery experience |
| Downstream |
Carrier, doorstep, service desk, receiving dock |
3PLs, reverse-logistics software |
A defect that was already in the box |
2. Returns Are Bigger Than the Service Desk — and Whose Numbers These Are
You already know returns hurt. What you may not have seen is who the headline numbers actually describe.
Quick answer: US retail returns hit an estimated $849.9 billion in 2025 — 15.8% of annual sales — with online higher at 19.3%. The merchant-side sample behind those figures is 358 professionals at companies over $500 million in revenue. Read them as a picture of the category, not as your store’s benchmark.
| Figure |
2025 value |
Source and sample |
| Total US returns |
$849.9 billion (15.8% of sales) |
NRF / Happy Returns, 2025 Retail Returns Landscape; 358 merchant professionals at US merchants over $500M revenue |
| 2024 comparison |
$890 billion (16.9%) |
Same report |
| Online return rate |
19.3% of online sales |
Same report |
| Won’t shop again after a poor returns experience |
71% of consumers (up from 67% in 2024) |
Same report; consumer sample of 2,006 |
| Returns that are fraudulent |
9% |
Same report |
Read the source: Consumers Expected to Return Nearly $850 Billion in Merchandise in 2025 and the 2025 Retail Returns Landscape key findings.
That one report carries both halves of the argument. Fraud and recovery experience are downstream by definition — and every unit behind both numbers was made and shipped by somebody first.
A word on whose numbers these are. The merchant sample is half-billion-dollar companies. Your store is not that. I am not telling you these figures do not apply — I have no evidence for that, and neither does anyone quoting them at you. I am telling you what they were measured on, so you know what you are borrowing.
Where the fulfillment guides stop
You have read the returns-and-fulfillment guides. Here’s the thing about them — they are mostly right, and they mostly start in the same place.
In the eight returns-focused fulfillment guides we read in full, "fulfillment" almost always begins at the receiving dock. Flat Out Fulfillment puts real QC inside a building the goods already reached. fulfilmentcrowd calls returns management "a fulfillment issue, not a customer service task," and its most upstream lever is product detail and sizing tools. Descartes works the pile-up after goods come back. Pitney Bowes lists challenges living in labels, tracking, refund speed and fraud. EcomLogistics gets as far upstream as returns "processed back into available stock before they have been inspected."
One of the eight goes further. Shippo writes: "Inspect goods coming in from suppliers the same way you inspect customer returns." That is still your own receiving dock, after the goods were made, packed and shipped.
What I am not claiming. Not "nobody writes about upstream causes." That would be false, and I name counterexamples below. The honest form is narrower: in the eight guides we read in full, one raised inbound supplier quality at all, and it did so in a single paragraph. Eight is the denominator — a hand-read count of pages we opened, not a crawl of the category. Read it as an observation about our reading, not about the industry.
Where a competitor’s column is blank in this article, it means we did not see it in the pages we read in full. It does not mean they lack the capability.
3. The Location Model: Cause, Earliest Catch, and Actual Discovery
This is the spine of the article, so here it is plainly.
The six-point location model used throughout this article.
Quick answer: A return cause has a location. Six points sit between a supplier’s line and a refund decision. Each return needs three of them named: where it was caused, where it could first have been caught, and where it was actually found. Most return conversations only cover the last one.
These are not six topics. They are six positions on one axis.
| # |
Point |
What can be caught here |
What it costs if it passes |
| 1 |
Supplier production line |
Build and material faults |
Whole batch carries the fault forward |
| 2 |
Pre-shipment check |
Build, variant and spec |
Fault leaves the country |
| 3 |
Packing and labelling |
Variant and packaging |
Fault leaves with a correct-looking label |
| 4 |
Line haul and carrier |
Transit damage only |
Freight spent on a unit that comes back |
| 5 |
Customer unboxing |
Everything, by the person least able to fix it |
Refund, return freight, lost repurchase |
| 6 |
Returns adjudication |
Everything, after the fact |
Handling, restocking, possibly no resale |
What ships before anyone looks
Three failure types get decided before a package moves. Physical damage: transit damage is a carrier conversation, but damage that was already there when the carton got sealed is not — and the return ticket looks identical either way. Wrong variant: same product, wrong colour, size or plug type. This is the quiet one, because your system says the right SKU shipped. Spec mismatch: the listing says 500ml, the unit is 450ml. Nothing broke. The product simply is not the one the page sold.
A seller on Shopify Community in 2021 asked the version of this that matters: "how does returns actually work? Do they send the defective product to me or to the producer? I’m a very long way from America, postage is expensive." Read what that person is asking. Not what do I say to the customer. They are asking where the physical object goes — because a return is a movement problem, not a script problem.
I do not have a source that splits returns across those three causes, and I am not going to invent one. What I can say is where each becomes visible.
Three columns, not one
The first version of this model collapsed two questions into one word. Where a problem happens and where it can first be seen are not the same place. Treating them as one is how blame lands on the wrong desk. So every return gets three coordinates: cause origin, earliest detection point, and actual discovery point.
| Return |
Cause origin |
Earliest detection point |
Actual discovery point |
| Wrong colour shipped |
Picking and packing (Point 3) |
Point 3 — variant check against the order |
Point 5 — customer unboxing |
| Product page states the wrong capacity |
Content and merchandising — off the physical chain |
Pre-publish listing review, before the SKU goes live |
Point 5 — customer measures it |
| Intermittent internal fault |
Supplier line (Point 1) |
Point 1 — functional test at the line, if one exists |
Point 5, often days after delivery |
| Crushed in transit |
Carrier handling (Point 4) |
Point 4 — handover condition check, nothing earlier |
Point 5 — customer unboxing |
Read the second row again. The product page has no station on the physical chain at all. Nobody at Point 1, 2 or 3 looks at your listing. The copy was written by someone who never touches the box. Its cause origin is content and merchandising; its earliest detection point is a listing review before the SKU goes live. That is the position it holds here — outside the six points, which is exactly why it goes unowned and keeps producing "wrong item" tickets no inspection could have stopped.
Not everything is visible at a bench
The middle column has a trap in it. What can be caught where depends on the kind of defect, and inspection is not one skill.
| Detectability tier |
What it takes to see |
Typical failure |
| 1 — Visible on sight |
Appearance check |
Dent, scuff, wrong colour |
| 2 — Checkable against a spec |
Measure, weigh, compare to the listing |
450ml sold as 500ml |
| 3 — Testable in seconds |
Power on, run one function |
Dead unit, wrong plug type |
| 4 — Needs sampling or lab work |
Batch sampling, materials or durability testing |
Stitching fails after a month |
| 5 — Only appears in use |
Nothing at the bench finds it |
Intermittent faults, device compatibility, environment-specific problems |
Points 1 through 3 handle tiers 1 to 3 reliably. Tier 4 needs sampling, a different budget. Tier 5 is not catchable upstream at all, and pretending otherwise is how sellers pay for inspection that was never going to touch their failure mode. If your returns cluster in tier 5, the answer is design, supplier selection or listing accuracy — not more eyes at the bench.
Why this is a location problem
Here’s where I land after eight years of this. Per ASG operational records, the pattern reads like this: if your supplier can ship, but tracking updates, QC, packaging and after-sales issues keep increasing, the problem may not be one shipping line. It may be your fulfillment system. The sharper version, per ASG operational records again: your supplier problem may actually be a fulfillment system problem.
Those four symptoms are one chain with no interception point, expressing itself four ways. Split them across four owners and you get four people who each solved their piece, and a return rate that barely moved.
What I am not saying: that support work is wasted. Support is where a bad experience converts back into a second order, and 71% of consumers say a poor returns experience makes them less likely to come back. That money is won at the desk. The claim is narrower than it sounds — the volume arriving at the desk gets set elsewhere.
These are company-observed patterns from our own fulfillment work, not an industry law. Your mix of SKUs, suppliers and markets will move the picture.
Related reading: Common Dropshipping Order Fulfillment Issues and Solutions covers the day-to-day version of this seam problem.
4. Why Visibility Doesn’t Move the Number — and What a Late Catch Costs
Real talk — tracking dashboards are the most satisfying purchase in this category, and the most commonly mistaken for a fix.
Quick answer: Visibility tools tell you a defect exists and where the package is. They do not change the point at which it could have been intercepted. Knowing sooner is worth money. Knowing sooner is not the same as catching earlier.
At Point 4, tracking tells you a package is late — irrelevant to whether the unit inside is correct. At Point 5, photo evidence tells you what the customer saw; by this stage the outbound cost has already been incurred, and a refund may already be unavoidable. At Point 6, reason codes tell you what got claimed, and those codes are shakier than they look. None of that moves the interception point.
A Shopify Community poster in 2026 — promoting his own return-fraud app, RefundSentry, in that same thread, so read him as a vendor rather than a neutral observer — put the ceiling in nine words: "By the time a return is requested, you’ve already shipped." One vendor’s line is an illustration, not proof. The argument stands on the model above.
The same defect also costs more the further it travels. At the supplier line it is rework or reject: one unit, no freight. At pre-ship or packing it is a hold or a swap. At the customer it is outbound freight, return freight, a refund, and a unit whose resale status is unknown. At adjudication it is all of that plus handling and warehouse space. A Descartes account manager described a client at that end state: hundreds of received returns sitting in piles, unidentified, with refunds already issued against them.
Then there is the leg nobody puts on the invoice. 71% of consumers say they are less likely to shop with a retailer again after a poor returns experience — up from 67% in 2024, per the NRF/Happy Returns consumer survey of 2,006 shoppers. Four out of five said they would tell friends and family.
One caveat about that 71%. It measures the experience of returning, not the cause of the return. A well-handled return of a defective unit may not trigger it at all. Which is why the service desk is not optional, and why it cannot be the only place you work.
5. Per-Order vs Per-Batch: Why Pre-Shipment Inspection Advice Doesn’t Fit You
This is where I have to be most careful, because the obvious conclusion here is one somebody else already published.
Quick answer: Advice to inspect before shipment is well established and easy to find. What is hard to find is that advice written for someone shipping single orders continuously, rather than approving one container at a time.
So let me be honest with you about the prior art. "Inspect at the source, not just downstream" is not my line. Silq, an inspection provider, publishes it near-verbatim: "Prevent returns by inspecting at the source, not just managing them downstream." BuckyDrop has a whole quality-control guide built around risk-tiered inspection. The Inspection Company walks the pre-shipment process end to end. The upstream half of this argument has been written, and written well.
What is different for you is one thing, and it is structural.
|
Batch importer |
Per-order seller |
| Unit of inspection |
A shipment or lot |
A single order, or a SKU across many orders |
| Decision available |
Approve or reject the shipment |
There is no "shipment" to approve |
| Timing |
Once, before a container moves |
Continuously, as orders arrive |
| Who the advice addresses |
Importers — Silq’s own page calls itself a "free guide for importers" |
Not addressed |
| Failure mode |
Bad lot ships |
Bad SKU or bad supplier keeps trickling out, one order at a time |
The right-hand column describes the gap as we read it in the pages we opened in full. Blank space would mean we did not see it, not that those providers lack the capability. Verify against the specific page, plan and product you are buying.
Pre-shipment inspection assumes a decision point: a lot exists, someone samples it, someone approves or rejects. Your orders do not have that shape. There is no container to hold. There is a stream.
And the naive fix does not scale. Photographing and filming every order sounds rigorous and collapses immediately — you pay for evidence nobody will look at, and slow dispatch to produce it.
Here is what works in a stream, and it is three layers instead of one:
- By SKU risk. Fragile, multi-variant, size-critical and functional items get checked deeper than a flat, single-variant item. Depth follows likelihood and cost of failure, not fairness.
- By supplier state. A supplier with a clean recent record gets sampled. One who just changed material, tooling or subcontractor goes back to full check until they have earned sampling again.
- By change event. Any change — new batch, new packaging, new variant, line moved — resets that SKU to a higher check level for a defined window, then relaxes.
That puts a check at Points 1 through 3 without pretending every order deserves the same scrutiny. It is the per-order translation of a per-batch idea, and the translating is the work.
Related reading: How to Dropshipping Fulfillment and Drop Shipping Fulfillment Services cover how these checks sit inside a per-order flow.
6. From Return Reasons to Locations: Audit Your Last 50 Returns
A return reason that stops inside your helpdesk is a defect that ships again next week.
Quick answer: Return reasons arrive in the customer’s language and live in the service team’s system. Neither form helps the person choosing suppliers. Reclassify by location — cause origin, earliest detection point, actual discovery point — and the data starts pointing at something you can change.
Reason fields are descriptions, not diagnoses. "Didn’t fit" could be a sizing chart failure (content), a spec drift at the factory (Point 1), or a customer who ordered the wrong size (Point 5, not fixable upstream). Same code, three different owners.
They are also not always sincere. The RefundSentry developer posting on Shopify Community in 2026 — again a vendor describing his own product’s problem space, not an independent finding — described the pattern as "Reason switching — The same customer uses ‘defective’ one time, ‘wrong item’ another, ‘didn’t fit’ another. Individually none are red flags. Combined, it’s a pattern."
He is talking about fraud. The same point cuts the other way for anyone reading reason data at face value: the field records what somebody chose to type, not what happened to the object.
So do the translation yourself. Take fifty recent returns and ask four questions of each.
| Ask this |
If yes |
Look here first |
| Could someone at the packing bench have seen this defect? |
The failure is upstream |
Points 1–3: supplier line, pre-ship check, packing |
| Is the same SKU or the same supplier repeating? |
The failure is systemic, not incidental |
Points 1–2: supplier state, change events |
| Is the same customer cycling through different reason codes? |
The reason field is unreliable |
Point 6: adjudication and fraud screening |
| Product was correct and the customer changed their mind? |
The failure is genuinely downstream |
Point 5: listing accuracy, sizing, recovery experience |
You are not looking for a percentage. You are looking for which column fills up. If column one fills, the helpdesk is not where that number moves. If column four fills, more supplier inspection is not where it moves either.
Then send that classification — not the raw reason text — to whoever talks to your suppliers. A supplier can act on "Point 2, variant mismatch, three occurrences this month, SKU-441." "Customer said it was wrong" rarely gives a supplier enough information to take corrective action.
Do it on paper: the 50-Return Root Cause Audit Sheet. One row per return, ten columns: Order ID, SKU, customer reason, evidence, cause origin, earliest detectable point, actual discovery point, supplier, corrective action, follow-up result. The three bold columns turn a complaint log into a map of where to stand. Fill in fifty rows, then count the columns. [download link pending]
This works on Shopify, WooCommerce, or a spreadsheet. The classification is the work; the platform is not.
7. Steel-manning: The Case for Treating Returns as a Service Problem
I have spent this whole article arguing for upstream. Now let me make the other side’s case as well as I can, because it is stronger than my framing makes it look.
Quick answer: Three arguments support treating returns primarily as a service problem, and all three are correct. None of them require the upstream half to be wrong — which is precisely why the original claim says not only.
One: the returns experience is itself a differentiator. Stacia Americas frames returns exactly that way, as a competitive differentiator rather than a cost centre. The data supports it: 71% of consumers say a poor returns experience makes them less likely to come back. That is won at the desk, not at the packing bench.
Two: the service side is controllable and fast. fulfilmentcrowd’s ten-step returns process is something you can implement in a quarter with people you already employ. Supplier-side checks take longer, cost more, and need leverage over parties you do not manage. With one quarter and one team, the service side is the rational first move. I would make the same call.
Three: some returns have no upstream at all. Per the NRF/Happy Returns report, 9% of returns are fraudulent. Buyer’s remorse and "changed my mind" have no factory-side signature. No check at Points 1 through 3 touches them, and neither does any check for tier 5 failures that only appear in use. For those slices, the service desk is the whole toolkit.
Where I still land. The original sentence is not only a customer service problem, and that phrase is load-bearing. Every argument above is about returns you cannot prevent. Mine is about the ones you can. You cannot tell which is which until you have asked the location question — so run both, and ask location first.
8. Frequently Asked Questions
QUICK ANSWERS ABOUT ECOMMERCE RETURN CAUSES
Does a return automatically mean a product defect?
No. Fraud accounts for 9% of returns per NRF’s 2025 report, and preference-based returns have no defect at all. Ask where the defect first became visible; if the answer is "nowhere," it is a downstream return.
Can pre-shipment inspection be applied to per-order dropshipping?
Not in its standard form. Pre-shipment inspection assumes a lot to approve or reject. Per-order sellers need a layered version keyed to SKU risk, supplier state and change events.
Is "inspect at the source" an original idea?
No. Inspection providers publish that exact framing. What is less commonly written is how it translates to continuous per-order shipping rather than batch importing.
Do tracking and returns-visibility tools reduce return rates?
They reduce reaction time, not return rate. Visibility changes when you learn about a defect, not where it could have been caught.
Should return reasons be classified by customer wording?
No. Classify by location — cause origin, earliest detection point, actual discovery point. Customer wording records a choice; location records a mechanism.
How do I verify a fulfillment partner inspects upstream, instead of taking their word for it?
Ask what each check leaves behind. ASG’s documented inbound flow separates arrival check, appearance check, functional test, photo record, packaging check and a final outbound review. The useful question about any of them is not whether it happens, but what you can retrieve afterwards. A control you cannot audit is a claim, not a control. Apply that to us as readily as to anyone else.
What should a return root-cause record actually contain?
Four things, and none of them is the customer’s wording. Where the defect first became visible; what evidence exists from that point; who was in a position to see it; and what changed as a result. The third field does the most work, because naming the station rather than the person converts a complaint into a location you can put a check at.
How should I respond to a customer who wants to return a product?
Approve quickly, tell them where the item goes and who pays for it, and confirm refund timing before they have to ask. This half lives entirely at the service desk — 71% of consumers in NRF’s 2025 consumer survey said a poor returns experience makes them less likely to shop with that retailer again. Separately from the reply, log where that defect could first have been seen. The conversation and the classification are two different jobs done from the same ticket.
9. Final Thoughts
Every return you get already happened somewhere. The ticket tells you when you found out. It does not tell you when you could have.
I have watched sellers rebuild the helpdesk, buy the tracking suite, rewrite the policy page — and get a return rate that moved by nothing, because all of those changes live at Points 4 through 6 and the defect entered at Point 2. That is not a failure of effort. It is a failure of location.
Which is the whole argument in one line. For the preventable, product-related share of your returns, this is not a helpdesk problem, it’s a detection problem — and detection has an address.
You do not need a factory, and you do not need to inspect every order. You need to know which point your returns trace back to, whether anyone could have seen them there, and then put one check at that point instead of six checks after it.
Start with fifty returns and three columns. The answer will be lopsided, and the lopsided side is where your money goes next.
Ready to look upstream? If your returns keep tracing back to Points 1 through 3 and you cannot stand at those points yourself, that is the specific thing a China-side fulfillment layer exists to do — put someone at the packing bench before the box seals. Start with our dropshipping service overview or dropship agent service, or contact us and bring your last fifty return tickets. See also extra services.
What we don’t do: we are not the cheapest option, we do not promise zero defects, and we do not take on your product-compliance, importer or platform-account responsibilities. Those stay yours.
10. About the Author
I’m Janson, CEO and founder of ASG Dropshipping. We run a China-side fulfillment operation out of Shenzhen and Dongguan, working with a network of verified factories to handle sourcing, quality control, packaging and per-order fulfillment for ecommerce sellers. Most of what is in this article comes from watching the same defect show up in three different tickets and tracing it back to one packing bench.
11. External Sources
12. ASG Data Note
The observations attributed to ASG in this article — the pattern of tracking, quality, packaging and after-sales issues rising together, and the framing of a supplier problem as a fulfillment-system problem — come from our internal fulfillment records and are company-observed, not industry findings. All return-rate, return-value, repurchase and fraud figures in this article come from the National Retail Federation and Happy Returns 2025 Retail Returns Landscape report, published 15 October 2025, based on a survey of 2,006 US consumers and 358 ecommerce professionals at US merchants with over $500 million in annual revenue. No ASG performance figures are cited anywhere in this article.
