Return prevention starts with evidence from the actual fulfillment handoff. By Janson Wang, Founder of ASG Dropshipping · September 28, 2026 · 30 min read
Three refunds land on the same Tuesday. One buyer says the stitching came apart. One sends a photo of a crushed box. One says the parcel never showed up.
You press the same button three times.
The button is the same. The three problems are not. Each one was decided somewhere else, months earlier, by a different call you already made.
A product quality control inspection moves one of three fulfillment-related return causes directly. The other two are each moved by a different document.
This article covers three groups — quality, transit damage, and late or non-arrival — not a complete taxonomy of all return reasons; fit, preference, and cases that stay unresolved also exist and sit outside these three.
Within the fulfillment-related group, quality is decided on the factory floor. Transit damage is decided in the packaging spec and in the carrier’s own rules.
Late is decided in your shipping promise, where the FTC attaches a refund duty.
Classify the failure before choosing the control. You already sort your returns in your head. Broken.
Crushed. Never arrived.
Almost nobody sorts their spending the same way.
Here’s why that gap matters. Each of those three words points at a different month, a different vendor, and a different document.
Sorting them by cause is the whole job. Everything below is what to do once you have.
The split is not something I invented. The payment layer already made it, in public.
Stripe’s dispute categories page maps its own buckets onto card network reason codes. Product not received carries Visa 13.1 Merchandise/Services Not Received. Product unacceptable carries Visa 13.3 Not as Described or Defective Merchandise/Services.
The same page lists the equivalents at Mastercard, PayPal and Klarna. A buyer who says “it never came” and a buyer who says “it came wrong” reach the bank through two different doors.
Now the part that is missing, and the reason this article exists.
Card networks do not separate a product that left the factory broken from a product that got broken on the way. Both land in Product unacceptable.
I searched that same page for damaged in transit, transit damage, shipping damage and damaged during. Zero hits on all four.
The control search for 13.3 Not as Described or Defective did hit. So the absence is real on that page, not a failed search.
That is a statement about card networks, not about the world. Plenty of people separate the two.
Your warehouse does. Your factory does.
The reason code does not.
So the payment layer gives you two doors for your three causes, and glues two of them into one. You have to do that separation yourself, before the money moves.
Here is a diagnostic clue, not a proof test — run it on your own return log as a starting point for investigation, not a verdict. Ask where the fault was sitting.
Was it already in the box before it was sealed? Somewhere between the seal and the doorstep?
Or nowhere near the unit at all, because what is really in question is a date?
A quality fault typically keeps showing up on the same SKU regardless of which carrier or which country it ships through, with the box arriving intact and the paperwork matching.
Transit damage typically clusters by lane, carrier, or packaging spec rather than by SKU — though a fragile SKU, or packaging that is specifically wrong for that SKU, can also rack up damage across several lanes at once, which is exactly why this is a clue to investigate rather than a rule that assigns the cause on its own.
Late does not touch the unit at all. The complaint, or the chargeback, lands before the delivery date does, sometimes before the parcel has even left the warehouse.
Confirm any of these with an actual inspection or arrival-condition record before you write the cause down — the pattern tells you where to look, not what you will find.
Where a return becomes visible, and where it began, are two different questions. We took the first one apart in Where returns actually start .
This article takes the second one: which link in the chain you reach for, cause by cause.
Why Reducing Returns Is Three Problems, Not One
Return cause
How you tell
Where the lever sits
Nearest payment-layer category
Sources read for this row
Quality: wrong, faulty, not as described
Same SKU, same description, repeats across carriers and countries; box and paperwork check out
Factory floor, and how deep you inspect before shipment
Product unacceptable, Visa 13.3CPSC 16 CFR 1107.21 and Stripe
Transit damage: it arrived crushed
Often clusters by lane, carrier, or packaging spec — a clue to confirm with an inspection or arrival-condition record, not a rule; a fragile SKU can still show up across several lanes
Packaging specification, plus the carrier’s own packaging rules
Product unacceptable, Visa 13.3 — the same bucketISTA and UPS
Late, or never arrived
The unit is not the question; the complaint or chargeback lands before the delivery date
Your shipping promise, and the refund duty sitting behind it
Product not received, Visa 13.1FTC and Stripe
These three rows are fulfillment-related causes, not the full universe of why a buyer returns something.
Fit, preference, and cases you genuinely cannot resolve from the evidence in front of you all exist — put those in an “other / unresolved” bucket rather than forcing them into one of the three.
A “wrong item” complaint in particular can just as easily be a warehouse picking error or a listing mismatch as a factory defect; check the order paperwork before you assume it belongs on the quality row.
Within that fulfillment-related group, most returns still sort cleanly into one row.
When a shipment fails on more than one front — a unit that was already defective and then got crushed in transit — name one primary cause, not three.
The test is which fault made it non-conforming first: a defect present before the box was sealed outranks damage that happened after.
A second, contributing cause can ride along in your notes. It should not turn one return into three rows on a spreadsheet.
Order, packing, and handling records help test what a return metric actually represents. Before you spend a dollar lowering the number, find out what the number counts.
“How do I lower my return rate?” is the right question. It is also being asked one level too high.
That rate is three different numbers stacked into one, and on some stores it is not even counting returns.
A merchant posting on Shopify Community in June 2026 went and checked:
Shopify’s “Returns” metric counts every refund event — including pre-shipment order edits for size adjustments — as a return. This is wildly inaccurate.
My Shopify-reported return rate is ~25%. My actual return rate is ~3%.
— YC3, Shopify Community, 16 June 2026
He went further in the same thread, down to one week:
I had zero actual physical returns in the last 7 days. Shopify reported -$572.
Every single line item driving that number was a pre-shipment size swap where items were “Removed” from an order before fulfillment — not a return.
— YC3, same thread, post 3
Read the thread rather than the headline. Another poster says the behaviour was fixed in March 2026.
The original poster goes back line by line and says it was not. I am not flattening that disagreement.
It is the useful part. Two people opened the same dashboard and came away with opposite readings.
A second merchant, in a different thread, is asking for the same separation from the other end:
Currently, Shopify mixes: RTO (delivery failed / courier return) with Customer Returns (buyer intentionally returned product) These are fundamentally different fulfillment events. …
RTOs are operational/logistics failures, not customer satisfaction issues.
When Shopify combines both into “returns,” it creates misleading return-rate data and makes performance analysis inaccurate.
— nardevfashion, Shopify Community, 18 May 2026
RTO is his word, and it comes out of a specific market.
He is describing cash-on-delivery direct-to-consumer selling in India, where parcels come back undelivered at a rate that gets its own acronym.
Do not import the acronym into a card-paid US or EU store without checking whether the event even exists there.
What travels is what happened next. A second person in that same thread, not the author:
We’ve seen staff accidentally treat high RTO products as “bad products” when the actual issue was courier coverage or fake COD orders.
Then inventory/planning teams start reacting to the wrong problem.
— order_ops_guy, same thread, post 3
A mislabelled number does not sit still. It gets acted on.
Somebody kills a product that was fine.
The same ambiguity exists one layer down, at your processor. Stripe reports two different dispute figures : dispute activity, counted by dispute date, and dispute rate, counted by charge date.
Same account, same month, two answers. “What is my rate?” needs a second sentence naming the clock.
Patterns are clues; direct records support the cause. Deciding to inspect harder is easy. Deciding which SKUs deserve it is the part that keeps people up.
A regulator has published a list of factors for exactly that decision. Two entries on it land squarely on the reader of this article.
The US Consumer Product Safety Commission sets how often children’s products must be re-tested. It lists ten things to weigh.
Two of them, word for word:
(iv) Consumer complaints or warranty claims;
(x) Inability to determine the children’s product’s noncompliance easily through means such as visual inspection.
— 16 CFR 1107.21
Put those two together and you get a usable filter. Escalate where complaints have already appeared, and where the fault cannot be caught by looking.
That is a regulator’s reasoning, published and citable, not a rule of thumb somebody made up.
The same body of rules ties sampling depth to how steady the process is:
If the manufacturing process … consistently creates finished products that are uniform in composition and quality, a manufacturer may submit fewer samples …
If the manufacturing process … results in variability in the composition or quality …, a manufacturer may need to submit more samples
— 16 CFR 1107.20
Now the boundary, and it matters who is reading this.
That rule governs manufacturers and importers of children’s products in the United States — and if that is you, this is not a borrowed analogy, it is a compliance obligation you may already be subject to.
For every other product category, it is not a standard you are held to, and it is not an industry norm.
I am borrowing a regulator’s published way of reasoning, not handing you a compliance obligation. Read it as a worked example of how a serious body decides where to look harder.
The same regulation adds two things. Periodic testing must be done by a third-party conformity assessment body.
And the baseline interval — testing at least once a year — may not exceed one year under paragraph (b).
A manufacturer running a documented production testing plan can move to once every two years under paragraph (c); one using an ISO/IEC 17025-accredited lab can move to once every three years under paragraph (d).
Even the longest of those qualifying intervals carries a lesson outside its scope: whoever wrote the rule did not think a testing interval, at any tier, could be set once and forgotten.
Two things follow for a quality control inspection on your own line.
First, a quality control inspection and a laboratory test are not interchangeable. One looks at the units in front of it.
The other establishes whether the design and materials comply at all. Passing the first says nothing about the second.
Second, escalation should be aimed, not blanket. The two factors above give you the aim: complaints already in hand, plus a defect that hides from the eye.
How many units to pull, and how to grade what you find, is a separate craft with its own arithmetic. That one is written up in our guide to AQL sampling for China dropshipping .
I am deliberately not repeating it here.
The warehouse handoff is one place to preserve inspection and packing evidence. Look — a supplier telling you the carton “passed drop testing” sounds like the transit-damage problem is handled. Sometimes it is.
Often the test they ran was never built to predict damage.
The International Safe Transit Association says so on its own overview page:
Non-Simulation Tests 1A, 1B, 1C, 1D, 1E, 1G, 1H Challenge the integrity of the product and package combination.
Useful as screening tests but not designed to simulate environmental occurrences.
General Simulation Tests 3A, 3B, 3E, 3F, 3H, 3K, 3L, 3N, 3M, 3P Provide a simulation of the damage-producing motions, forces, conditions, and sequences of transport environments.
Useful as a predictive tool to understand risk of damage.
— ISTA test procedures overview
Screening versus predictive. That is ISTA’s own wording, not mine.
So the first question to a supplier is not “did it pass” but “which series”.
The one that matches a direct-to-consumer parcel has a name and a weight range:
Test Procedure 3A is a general simulation test for individual packaged-products shipped through a parcel delivery system.
It covers packaged products of 70 kg (150 lb) or Less. Source: the ISTA 3A overview .
The same document draws a line most sellers never hear about:
ISTA 3 Series tests are advanced tests and are designed to: Challenge the capability of the package and product to withstand transport hazards, but Utilize general simulation of actual transport hazards, and Do not necessarily comply with carrier packaging regulations.
That last clause is doing a lot of work. Passing ISTA is not the same as satisfying the company that will actually carry the box.
Which brings in the carrier. UPS states its position on packaging on every page of that eight-page document, twice on the first:
Please note: UPS does not make any warranties, express or implied, regarding this information.
It is the responsibility of the Shipper to ensure that proper packaging is used and that contents are adequately and securely packed in a container, wrapped, and cushioned for transportation.
The same document gives a concrete list:
Use a rigid box with flaps attached. Avoid re-using boxes.
Use the Box Strength Guidelines chart to determine the proper strength of your shipping container. Use adequate cushioning material.
Use strong tape designed for shipping. Never use string or paper over-wrap.
Wrap all items separately.
Here is the bridge, written out rather than implied. The packaging decision and the damage claim are one decision, taken months apart.
You choose a carton in March. In August a buyer photographs a crushed corner.
Whoever assesses that claim opens a document written by the carrier, not by your supplier. Nothing in the March conversation mentioned that document.
Three limits on everything in this section, stated plainly.
The actual ISTA test parameters — drop heights, vibration profiles, pass criteria — sit behind a paid standard. I did not buy it, so this article contains none of those numbers.
Third-party laboratory blogs publish figures that look like them. None were used here.
UPS is describing its own allocation of responsibility. It is not a neutral referee on who broke the box.
And this is UPS. The equivalent FedEx page returned a 404 on the day of research, so there is no second carrier to cross-check against.
Treat the passage above as one carrier’s stated approach, not as how carriers behave in general.
Route and timing evidence sit downstream from the product inspection decision. Of the three causes, late is the one that came back with a statutory refund duty attached in the sources I read. Not a policy.
A duty.
The FTC’s business guide to the Mail, Internet, or Telephone Order Merchandise Rule sets it out:
If you make no shipment statement, you must have a reasonable basis for believing that you can ship within 30 days.
That is why direct marketers sometimes call this the “30-day Rule.” If, after taking the customer’s order, you learn that you cannot ship within the time you stated or within 30 days, you must seek the customer’s consent to the delayed shipment.
If you cannot obtain the customer’s consent to the delay … you must, without being asked, promptly refund all the money the customer paid you for the unshipped merchandise.
— FTC business guidance
Read the verbs. The rule is about shipping, not about arriving.
A parcel can leave a Shenzhen warehouse two days after the order. It can reach the buyer weeks later.
That has not tripped this clock. Conflating dispatch with delivery is the common misreading here, and being pedantic pays, because the two words carry different obligations.
The refund side has its own clock. The same guide sets seven working days after the order is cancelled.
Where you extend the credit yourself, it is one billing cycle.
The second chain is faster than the legal one, and it runs through your processor rather than a regulator. A merchant described it in 2021:
I have recently gotten a chargeback on my site with the reason being cited that the product was not received.
The thing is, this request for a chargeback was asked just a couple days after the person bought the product, meaning of course they wouldn’t have received it yet.
— DND, Shopify Community, 1 September 2021
That post is five years old and describes a pre-order situation. Treat it as a mechanism, not a typical case.
The mechanism is what matters. An unmet delivery expectation can reach the payment layer before the delivery date arrives.
Once there, it lands in Visa 13.1, Merchandise/Services Not Received .
Stripe’s own guidance for that category accepts, as valid evidence, that the agreed delivery date simply hasn’t arrived yet, together with your customer-communication records — you are not limited to proof that delivery already happened, which you may not have on a shipment still in transit.
I have no data connecting days in transit to a rise in returns. So this article states none.
The shape is all I can state. The late chain has a legal floor and a payment-layer trigger.
The two run on different clocks.
The cheapest remedy depends on the full cost stack. A refund is not one number. It is a stack of lines, and each line is decided by a different party.
Start with the one that is settled before you touch it. Shopify’s documentation, as of September 2026, on duties collected at checkout:
If you refund an order that has already been fulfilled, then you can refund the duties and import taxes to the customer.
However, any funds that you have used to purchase DDP labels aren’t returned to you by the carrier.
— Shopify Help Center
The label money is gone. The tax money is a separate fight, and it has its own section below.
One more line is easy to miss here: the original payment processing fee. Shopify’s own refunding guide states that on a transaction processed through Shopify Payments, you don’t get the credit card fee refunded — full refund, partial refund, fulfilled or not.
If you use a different payment provider, the treatment is whatever that provider’s own terms say; Shopify’s page only speaks for Shopify Payments.
Inside the refund screen itself, the lines come apart. Shopify’s refunding guide , as of September 2026, puts it this way:
Optional: If the order includes duties and import taxes, then select whether you want to refund Duties and import taxes on refunded items and Additional fees: Customs clearance.
When refunding an order, you can manually edit the Refund amount to deduct a restocking fee. You can also refund any shipping fees.
One line in that same guide is a trap worth memorising: you can't create a return after a refund has been issued.
That closes off Shopify’s own return-record workflow — the structured path where you send return shipping instructions, the system tracks the shipment, and you inspect the item after it arrives before issuing the refund .
It does not mean the unit is barred from physically coming back, or that you cannot inspect it if it does; refunding first just means you lose that built-in return-then-inspect-then-restock sequence.
If you want it, create the return before you refund.
Return shipping is not a fixed cost of doing business either. It is a setting. Shopify’s return rules page says so plainly:
You can choose how you handle return shipping costs by selecting one of the following options: Offer free return shipping.
Add a flat rate return shipping fee that’s charged one time per return. Have your customers buy their own return shipping label.
The same page allows different rules per market. A choice that looks obvious in one country can be set differently in another.
Then the parcel lands somewhere, and the last line appears. ShipBob’s developer documentation lists what can be done with a returned unit: Default, Restock, Quarantine, Dispose.
Four options, and someone has to choose one. Quarantine is not free.
Dispose is not free either.
I have no source for what share of returned units make it back to sellable. So I am not giving you one.
The choice itself is a real step in the workflow. A person makes it.
It costs something.
What a Refund Costs You, Line by Line
Refund line
Who decides it
Can you recover it
Where it is set
DDP label cost
Carrier
No — aren't returned to you by the carrier
Shopify duties tasks page
Original payment processing fee
Your payment provider
No, on Shopify Payments — the credit card fee is not refunded; other providers set their own terms
Shopify refunding orders
Duties and import taxes
You, per refund
Refundable to the buyer at your option; recovery from customs is a separate process
Shopify refunding orders
Additional fees: customs clearance
You, per refund
Optional toggle at refund time
Shopify refunding orders
Return shipping
You, as a standing rule, per market
Three options, one of which is charging the buyer
Shopify return rules
Restocking fee
You, per refund
Deducted manually from the refund amount
Shopify refunding orders
Disposition of the returned unit
Your warehouse
Restock, Quarantine or Dispose — a decision, not a default
ShipBob returns guide
Which building the parcel goes back to, and who opens it, changes several of those lines at once. We wrote that up separately in how to choose a fulfillment center for returns .
Physical disposition and inventory recovery need their own warehouse decision. You paid duty at the border so your buyer would not have to. Then the parcel came back.
Is that money recoverable?
The statute says yes.
Whether you actually collect depends on following a specific procedure on a specific clock — and on checking, before you assume a past claim is dead, whether a waiver already covers you.
Take the United States first. It has the most complete published chain.
The statute gives you standing:
… (A) upon which the duties have been paid, … (C) which is- … (ii) ultimately sold at retail by the importer, … and for any reason returned to and accepted by the importer, … and (D) which, within 5 years after the date of importation or withdrawal, as applicable, has been exported or destroyed under the supervision of U.S.
Customs and Border Protection, an amount … shall be refunded as drawback.
— 19 U.S.C. 1313(c)
That last clause is where most of this quietly dies. Not exported.
Not destroyed. Exported or destroyed under supervision.
For a returned retail unit specifically, §190.45 adds a substitution route with its own ceiling and matching test — this is one specific path, not a universal checklist for every US duty-recovery claim:
Designating an entry of merchandise that was imported within 1 year before the date of exportation or destruction …
Certifying that the same 8-digit HTSUS subheading number and specific product identifier (such as part number, SKU, or product code) apply to both the merchandise designated for drawback (in the import documentation) and the returned merchandise. …
The total amount of drawback allowable will not exceed 99 percent of the amount of duties paid
— 19 CFR 190.45
Then the piece of paperwork almost nobody hears about until it is too late:
The claimant, or the exporter …, must file at the port of intended redelivery to CBP custody a Notice of Intent to Export, Destroy, or Return Merchandise for Purposes of Drawback on CBP Form 7553 at least 5 working days prior to the date of intended return to CBP custody
Drawback will be denied on merchandise that is exported or destroyed after the statutory 5-year time period.
— 19 CFR 190.42
That five-working-day figure is specific to redelivering the goods to CBP custody.
Destroying the goods instead runs on a different section and a different clock: 19 CFR 190.71 generally requires the same Form 7553 notice at least 7 working days before the intended destruction date, filed with the CBP port where the destruction will actually happen — not the 5-day figure that governs redelivery.
Two waivers sit inside that same rulebook, and they change the ending.
Under §190.42(c), the notice requirement does not apply if CBP has already granted you a prior-notice waiver under §190.91, or if you qualify under the one-time waiver procedure at §190.36 — a single-use application that, once approved, covers your future exports or destructions without a fresh 7553 each time.
So here is the line to take away from this whole article.
Destroy or re-export a returned parcel without following the applicable notice procedure, and you put the duty claim at serious risk — but check for a waiver before you write it off as gone. The notice comes first because its whole point is to let customs look at the goods.
If no waiver applies to you, that notice is what stands between “recoverable” and “forfeited.”
The United Kingdom runs the same principle on a shorter clock. There is one trap in the eligibility list. HMRC’s guidance , last updated 2 April 2026, reads:
You can claim on goods which: do not meet the conditions of their contract; are defective; are damaged before customs clearance
The time limit for submitting a claim for repayment of import duty and VAT is: 3 years for overpayments; 1 year for rejected imports; 90 days for withdrawal of an import declaration
Read that third condition twice. Damage after clearance is not on the list.
And damage after clearance is most of what a direct-to-consumer brand ever sees.
For the European Union I only reached the Commission’s explanation of the Union Customs Code. The article text itself did not load.
What the Commission’s page says:
Import or export duty can be repaid on any of the following grounds (Art 116(1) UCC): (a) overcharged amounts (Article 117 UCC); (b) defective goods or goods not complying with the terms of the contract (Article 118 UCC)
For repayments/remissions on the basis of Articles 117 and 118, there is no threshold foreseen.
That is the Commission describing the Code. It is not the Code.
I did not read the article text, so do not quote this back to a broker as the article itself.
And the platform view, which sounds contradictory until you set it beside the law:
You might want to withhold import tax refunds if the funds that you’ve remitted aren’t recoverable from the customs agency and you don’t want to assume that cost on your customer’s behalf.
— Shopify Help Center , as of September 2026
Both statements hold at once. The platform describes what usually happens.
The regulations describe what is available if you file on time, under supervision, with matching codes.
Recovering Import Duty on a Returned DDP Parcel
Jurisdiction
What I actually read
The clock
What takes it off the table
United States
19 U.S.C. 1313(c), 19 CFR 190.42, 19 CFR 190.45, 19 CFR 190.71
Export or destroy within 5 years of import; file Form 7553 at least 5 working days ahead of redelivery to CBP custody, or at least 7 working days ahead of destruction (§190.71); on the §190.45 substitution route, designate an entry imported within 1 year before that
Acting without CBP supervision; letting the 5-year period run out; on the §190.45 route, a mismatch between the 8-digit subheading and product identifier on your designated import entry and on the returned merchandise — unless a §190.42(c) or §190.36 waiver already covers you
United Kingdom
HMRC guidance page, last updated 2 April 2026
1 year for rejected imports, against 3 years for overpayments
Damage that happened after customs clearance; the published list says damaged before customs clearance
European Union
European Commission explanation of Article 116(1); article text not retrieved
not covered by the sources we read
not covered by the sources we read
Two boundaries on this section. It covers the destination country only.
What China’s customs authority does on the way back was not researched this round. Nothing here covers the full loop.
Also, if you use the §190.45 returned-retail substitution route, it is the 8-digit subheading and product identifier on your designated US import entry that has to match the returned merchandise — not an outbound export declaration.
That part is written up in our guide to HS codes and product descriptions for DDP shipping from China .
If your DDP volume is meaningful, spend an hour on Form 7553 this quarter. Spend it before the next batch of returns gets consolidated and scrapped.
Then ask your customs broker two questions. On the last returned shipment, was a notice of intent filed on time — and if not, does a prior-notice or one-time waiver already cover it?
If the answer to both is no, the claim is likely gone, and nobody told you.
Put prevention spend at the handoff where the evidence points. Separate the causes and the budgeting question becomes answerable. Which of these deserves money before the fact?
Which only ever costs money after?
There is a mature framework for exactly that, and returns are named inside it:
Cost of quality (COQ) is defined as a methodology that allows an organization to determine the extent to which its resources are used for activities that prevent poor quality, that appraise the quality of the organization’s products or services, and that result from internal and external failures.
— ASQ on cost of quality , read August 2026
Four categories: prevention, appraisal, internal failure, external failure.
The line that matters for this article sits under external failure, word for word:
Returns: Handling and investigation of rejected or recalled products, including transport costs
Return freight is not a miscellaneous logistics line. In this framework it is a failure cost, and it sits opposite prevention and appraisal by design.
Where Prevention Spend Belongs in the Cost Framework
Cost category
What ASQ names in it
Where our three causes land
Prevention
Product or service requirements, quality planning, quality assurance, training
Writing the packaging specification and the inspection criteria before production starts
Appraisal
Measuring and monitoring activities related to quality
The quality control inspection itself, and any package testing you commission
Internal failure
Waste, scrap, rework or rectification, failure analysis
Units caught and fixed before the parcel leaves — cheapest place to find a defect
External failure
Repairs and servicing, warranty claims, complaints, and returns including transport costs
All three causes land here — including late or non-arrival, where the buyer never has the parcel at all
Now the part this section has to say about itself.
That framework gives you accounts, not numbers. It tells you which bucket a cost belongs in.
It does not tell you what a dollar of prevention returns. I searched the ASQ page for a ratio, a rule of thumb, a percentage of revenue, and a multiplier.
None of them are there. I also did not find a public, measured comparison of prevention spend against return cost for parcel direct-to-consumer selling.
So I am not going to hand you a multiplier. Anyone who does, in this specific area, is quoting something they cannot show you.
The framework is good for something else. It ends the argument where return freight counts as logistics and inspection counts as overhead.
In this accounting they are two sides of one account.
Need to see where your return loss starts?
Share one SKU, one shipping lane, and a recent return sample. ASG can help you map the evidence before you change inspection, packaging, or shipping.
Review Your Return Evidence
Packing evidence can separate a preparation issue from a later transit event. Let me put the opposing view at its strongest. In this channel it is not a fringe position.
On some product lines it is the right call.
The argument runs like this. The unit is coming back to a warehouse thousands of miles from the factory that made it.
Return freight can exceed the landed cost. In this channel, refunding without asking for the goods back gets discussed as a normal option rather than a last resort.
One Shopify Community thread from May 2022 has a poster suggesting exactly that: refund, then ask the buyer to donate or recycle the item.
She is answering in a helper role, not describing her own store, so read it as evidence about the conversation rather than about practice.
The platform’s own documentation strengthens the case:
Remember that refunding the order in full is the only way to prevent a customer from disputing an order with their payment provider.
— Shopify Help Center , as of September 2026
Return shipping is a setting you control, as shown earlier — but it only decides who pays for the label, not whether the unit comes back.
A returnless refund is a separate decision layered on top of it: you can offer free return shipping and still require the item back, or skip requiring it back at all regardless of who pays for the label.
Put it together: refund fast, keep the customer, skip the freight, avoid the dispute. That is a coherent policy.
I have seen it work on low-value, high-variance lines.
Now the boundaries, which are just as concrete.
A chargeback is not a refund with extra steps. Shopify says the bank takes the disputed amount from you right away and also takes a chargeback fee. Stripe describes the same sequence independently: the dispute immediately reverses the payment and pulls the amount plus network dispute fees.
Two separate companies, same mechanic. Money leaves first, argument happens later.
Above that sits a ceiling that does not care how generous your refund policy is. Stripe’s page on dispute monitoring programs is worth reading in its own words:
For the purposes of monitoring programs, a dispute or chargeback occurs when funds move out of an account due to a disputed payment, regardless of the reason.
Monitoring programs don’t consider refunds when identifying disputes.
Similarly, monitoring programs don’t consider dispute outcomes. … They’re also more interested in how successfully you prevent disputes than in whether you win them.
Failure to comply with the requirements of a program within the specified time period can result in the network refusing to process further payments to you.
Three fences around those thresholds, and they matter more than the numbers themselves.
The published figures on that page are dispute-and-fraud thresholds, not return-rate thresholds, and they are joint conditions, not stand-alone tripwires.
Visa’s VAMP program folds two different data sources into one count: disputes reported through the network, plus fraud reports from card issuers (early fraud warnings, sourced from TC40 reporting).
A twenty percent return rate feeds neither number.
Mastercard’s Excessive Chargeback Program works on a joint test too — an account is only placed in it when the monthly dispute count (100-299 for the first tier) and the chargeback rate (1.5-2.99%) are both met in the same month, not by crossing either line alone.
Anyone who quotes one of these numbers in isolation as “the trigger” is oversimplifying a combined condition.
The thresholds are regional. Stripe lists 2.2% for CEMEA against 1.5% elsewhere, and a count of 150 in CEMEA against 1,500 elsewhere.
And Stripe caveats its own page:
This page is a general guide for Stripe users, not a comprehensive reference for card network monitoring programs
If you quote the numbers, quote that sentence too.
So the steel-manned position survives, with a limit attached. Refunding freely is a defensible cost decision.
It is not a defence against the monitoring programs, because those count the dispute and ignore your refund.
What to submit when you do fight a dispute is a different article, and this one stops here.
A product inspection answers a different question from a carrier scan or delivery promise. What is the difference between a return and a chargeback?
A return is a workflow you run: the buyer requests, you approve, goods come back, you inspect, you refund.
A chargeback is the buyer going to their bank instead. Shopify and Stripe both describe the money being pulled from you immediately, plus a fee, before anyone reviews the merits.
Is package testing the same thing as product inspection?
No, and they answer different questions. A quality control inspection asks whether the units match specification. Package testing asks whether the packed combination survives handling.
ISTA separates its own procedures into screening tests and predictive simulation tests, and neither one certifies that the product itself conforms.
When does refunding an order also refund the duty?
Only if you choose it. Shopify’s refund screen , as of September 2026, exposes duties and import taxes and customs clearance fees as separate optional lines.
Refunding your buyer’s duty is one decision. Recovering that duty from the customs authority is an entirely separate process with its own deadlines.
What disqualifies a duty claim on a returned parcel?
In the United States: acting without CBP supervision, or letting the five-year period expire.
On the §190.45 returned-retail substitution route specifically, a mismatch between the 8-digit HTSUS subheading and product identifier on your designated import entry and the returned merchandise also disqualifies the claim.
Missing the Form 7553 notice — at least 5 working days ahead of redelivery to CBP custody, or 7 working days ahead of destruction under §190.71 — puts a claim at serious risk, but check first whether a prior-notice waiver under §190.42(c) or the one-time waiver under §190.36 already covers you.
In the United Kingdom, damage after customs clearance is not on the eligibility list .
Do card networks watch your return rate?
No.
They watch disputes and fraud reports. Stripe’s summary of the monitoring programs says a dispute counts regardless of reason, and Visa’s VAMP program combines dispute counts with fraud reports into one threshold.
Refunds are not considered. Outcomes are not considered either.
A high return rate is your margin problem. A high dispute-and-fraud ratio is your ability to accept cards at all.
A disposition decision still needs warehouse evidence after the customer-facing refund. Does the payment layer separate transit damage from a factory defect?
No. Stripe maps both to Product unacceptable, Visa reason code 13.3.
The split you make internally has no counterpart there.
What notice must reach customs before returned goods are destroyed?
CBP Form 7553, filed at least seven working days before the intended destruction date, with the CBP port where the destruction will occur (19 CFR 190.71 ).
That is different from the five-working-day notice for redelivering goods to CBP custody — destruction and redelivery are two separate procedures with two separate clocks.
Skip the applicable notice with no waiver on file, and the claim is at serious risk.
Is the 30-day rule about shipping or about delivery?
Shipping. The FTC rule sets a deadline for dispatch, not arrival.
A parcel that ships in two days can still travel for weeks.
Does a returned unit go straight back into sellable stock?
Not automatically. ShipBob lists Default, Restock, Quarantine and Dispose.
Somebody has to choose, and that choice carries a cost.
When is more inspection supported by a published set of factors?
When complaints or warranty claims already exist, and when the fault resists visual inspection. Both are CPSC factors.
They bind manufacturers and importers of children’s products directly; for every other product category, they are a published example of how a regulator reasons about escalation, not a requirement.
Key Takeaways
Your three return causes reach the bank through two doors, not three. Card networks separate Product not received from Product unacceptable, and put factory defects and transit damage in the same bucket (Stripe ).
Check what your return metric counts before you spend anything lowering it. One merchant found a reported rate of ~25% against an actual ~3% on his own store (Shopify Community ).
Escalate a quality control inspection where complaints already exist and the fault is hard to see. Those are two of the ten factors a US regulator published (16 CFR 1107.21 ).
Passing a package test is not the same as satisfying a carrier. ISTA writes that its 3 Series tests Do not necessarily comply with carrier packaging regulations (ISTA 3A overview ).
Duty on a returned DDP parcel is recoverable under the statute, but the claim is at serious risk if the applicable notice isn’t filed on time and no waiver applies — check with your broker before assuming a claim is dead (19 CFR 190.42 ).
Platform and regulator wording here was reopened and reconfirmed word for word in September 2026, and pages change.
The packaging-standard and cost-of-quality sources — ISTA, UPS, ASQ — could not be reopened this round after repeated timeouts.
Those passages carry an August 2026 date and are marked as unverified this round rather than reconfirmed. The paid ISTA test parameters were not purchased, so no test figures appear.
The European Union passage cites the Commission’s explanation, not the article text, which did not load.
China-side customs treatment was not researched; only the destination country is covered.
19 U.S.C. 1313 — Drawback and refunds , Office of the Law Revision Counsel19 CFR 190.42 — Procedures and supporting documentation , eCFR19 CFR 190.45 — Returned retail merchandise , eCFR19 CFR 190.71 — Destruction under Customs supervision , eCFR19 CFR 190.36 — One-time waiver of prior notice for future claims , eCFRClaim a repayment or remission of charges on rejected imports , HM Revenue and CustomsCustoms debt: remission and repayment , European CommissionAdditional tasks for collecting duties and import taxes at checkout , Shopify Help CenterSetting up return and cancellation rules , Shopify Help CenterRefunding orders , Shopify Help CenterReturns , Shopify Help CenterChargebacks and inquiries , Shopify Help CenterDisputes , StripeDispute categories , StripeDispute monitoring programs , StripeMeasuring disputes , StripeA Business Guide to the FTC’s Mail, Internet, or Telephone Order Merchandise Rule , Federal Trade CommissionTest Procedures overview , International Safe Transit AssociationISTA 3A overview , International Safe Transit AssociationUPS Packaging Guidelines , United Parcel Service16 CFR 1107.21 — Periodic testing , eCFR16 CFR 1107.20 — Certification testing , eCFRReturns , ShipBob developer documentationWhat is Cost of Quality? , American Society for QualityReturns metric in Analytics is misleading , Shopify CommunityShopify should add separate RTO status , Shopify CommunityChargeback for product not received before I have had reasonable time to deliver , Shopify CommunityHow to handle returns in dropshipping with AliExpress , Shopify Community
This article contains no ASG figures. Every number, quotation and rule above comes from the public sources listed here.
Each one can be reopened at the link given.
One boundary worth stating plainly, since this is the work we do. We do not give legal or customs advice.
We do not decide how a chargeback, a rejected-import claim or a drawback claim comes out. That call does not sit with a fulfillment provider at all.
Most of this article has been about where it does sit.
What sits with us is narrower: making the records exist before the parcel leaves China. The criterion the goods were inspected against.
The packaging specification the carton was built to. The date it actually shipped.
Each one tied to its order and its SKU. A bank asks for that pairing when it wants proof of delivery.
A broker needs it when a claim has to name an entry.
Turn the diagnosis into a fulfillment control plan
Bring the product, packing, carrier, and customer records together before you add another blanket inspection or refund rule.
Build the Right Control Plan
A return rate is an average of three different failures. Averages hide levers.
Split the causes and each one points at a document you can open. An inspection criterion.
A packaging specification. A shipping promise.
I did not learn that from a framework. I watched a client argue for a month about a quality control inspection.
Meanwhile the real money was leaving through crushed cartons and forfeited duty. The inspection was not wrong.
It was aimed at one part of the problem, not the whole of it.
You do not need a new system to start. Take your last fifty returns.
Sort each one with the same question from earlier: was the fault already in the box before it was sealed, did it happen between the seal and the doorstep, or is the unit not really the issue and the date is?
Some will not fit any of the three — mark those “other / unresolved” rather than forcing a fit.
For the ones that do fit, put each in three columns: cause type, where the lever sits, and what evidence you already have for it.
That third column is a pre-shipment inspection record, a packaging spec and carrier file, or a shipping promise and a tracking timestamp, depending on the row.
It is the column that surprises people. If you cannot fill it in for a given row, the exercise has not given you an answer yet.
It has given you the question worth going and getting evidence for.
