> ## Content Index
> Fetch the complete content index at: https://conversion.club/llms.txt
> Use this file to discover other available public pages before exploring further.

# Fashion E-commerce Return Rate: How to Cut It Without Killing Your Sales
- URL: https://conversion.club/fashion-e-commerce-return-rate-how-to-cut-it-without-killing-your-sales/
- Published: 2026-07-21T06:00:24.000Z
- Updated: 2026-10-19T20:26:17.000Z
- Description: In fashion, 20-30% of orders come back. Learn where your return rate comes from, how to isolate it by size and how to manage it without hurting sales.
- Author: Jeremie Werner
- Tags: CRO, Post-Purchase, Fashion, Guides

## Return rate: the enemy eating your margins

When a product comes back, you've paid for it twice: once to sell it, once to get it back. It's the quietest expense in online fashion, and often the heaviest.

On the dashboard, gross revenue climbs and everyone celebrates. Then the gap with net revenue widens week after week, and the cash you were counting on melts away along the way.

We tend to look for the leak in acquisition, when it's almost always somewhere else: in the return rate, the thing nobody watches that slowly but surely destroys your profitability.

One brand the authors work with learned this the hard way. After switching carriers for its international orders, its return rate went from 25% to 45% in two months, without a single product or product page being touched.

The whole difference came down to the delivery experience. Returns aren't a fate that falls from the sky: they almost always respond to a decision made elsewhere, whose effect nobody saw coming.

## What is a good return rate in fashion?

In fashion, online returns hover around 20-30%, four to five times the rate in stores. One number deserves your attention, because it tells you where to act: 70% of fashion returns come down to a problem with size or fit. The rest follows from it.

Rather than comparing yourself to distant averages, find your own position. The Shopify baseline is around 18-19%, and across the fashion stores the authors see, the observed median sits between 8% and 15% product return rate, calculated on units.

One brand the authors work with holds at 5-6%, another is above 20%: same market, up to a fourfold gap, depending on what each one does or lets slip.

Either you know your own number SKU by SKU, or you aren't really managing it. And as long as it stays fuzzy, you think you're running a store when you're mostly running an expensive rental service.

![Two men talking next to a cart of documents, from the film Boomerang (1992) with Eddie Murphy](https://storage.ghost.io/c/b5/a7/b5a79c8d-ed77-4459-9729-b694bd04720c/content/images/2026/10/Two-men-conversing-while-standing-by-a-document-cart.png)

Still from the film Boomerang (1992) with Eddie Murphy

## Where is this bleeding coming from, exactly?

Before fighting it, remember that it isn't natural. Free returns aren't a law of physics. They're a dated business decision that someone made for you a long time ago.

In 1999, Zappos sells shoes online, the most absurd idea of the day, and defuses the risk with a stroke of genius: free returns both ways, for 365 days. The trade-off is clear: it costs more, but it keeps the customer longer.

Over twenty years, ASOS, Zalando and Amazon turn this differentiator into a mere ticket to entry. "Try before you buy" becomes the norm, Covid blows volumes up, and the whole sector ends up absorbing an externality it created itself.

Since 2023, the pendulum has swung back. In the UK, Zara [deducts £1.95 per return](https://www.which.co.uk/news/article/how-to-avoid-paying-zaras-new-1.95-returns-charge-aW4qW6I5oBPc?ref=conversion.club), H&M and Boohoo [follow suit](https://www.retailgazette.co.uk/blog/2023/09/hm-charge-online-returns/?ref=conversion.club), then ASOS [starts charging](https://www.retailgazette.co.uk/blog/2024/09/asos-returns-fee/?ref=conversion.club) its biggest "returners." After a quarter century of free returns treated as a given, the given is shrinking. The lesson of this shift applies to you: returns are a cost the market chose to bear, then to bill back, which means you keep control over where you decide to act.

## How much does an e-commerce return really cost?

When a customer sends a product back, the loss never stops at the refunded amount. To see all of it, you have to follow the whole cycle, down to the part of the iceberg below the surface.

First comes the logistics cost, the most visible one: return shipping, often free, then the time to receive, inspect and restock. Count on €8 to €18 of pure handling per returned item in apparel. A brand that processes 40,000 returns a year already loses €320,000 to €720,000 there, before discounting anything.

Then comes depreciation. A returned item rarely keeps its full value: a crumpled package, a sign of wear, and it's headed for the sale rack or secondhand (or the dumpster, but I don't even want to talk about that).

The service cost shows up on no report: answering, processing and refunding all take support time away from tasks that would actually drive sales. And the most invisible of all is the opportunity cost, because during its round trip the product isn't on the shelf for the next customer, the one who might have kept it.

Hence the trap the bill hides: a 25% return rate doesn't cost you 25% of your margin. It costs far more. The sale is refunded, but reverse logistics, handling and depreciation stay on your books. On a single unit, these costs can eat up to two-thirds of the contribution margin. A return isn't just a canceled sale: it's a canceled sale plus an invoice.

## The only metric that really judges

**Ask your team: does your conversion rate exclude returns? Nine times out of ten, the answer is no.**

We celebrate the order the day it comes in and forget to subtract it three weeks later, when it comes back. As a result, the displayed conversion rate is inflated and the revenue it suggests is partly fictional.

At one brand the authors work with, the e-commerce lead rebuilt her return rate by hand every month, separating the refunded share, the exchanged share and the share turned into store credit, for lack of a tool that could do it automatically.

It's laborious, but at least she was facing reality. Most brands never look at it.

The deciding metric is neither gross revenue nor order volume: it's contribution margin per visitor, after returns are taken into account. And from that comes a rule we apply to every test:

> A test that lifts conversion **and** returns is a false positive. Always read revenue per session and return rate together, never one without the other.

Almost nobody does this, because it's counterintuitive. A product page that "wins" on add-to-cart while sending back five more points of merchandise hasn't won anything: it has dug a loss that the dashboard dresses up as a success, because nobody looked at the other half of the equation.

![The IOAN Undyed store by Industry of All Nations](https://storage.ghost.io/c/b5/a7/b5a79c8d-ed77-4459-9729-b694bd04720c/content/images/2026/10/Clothes.webp)

The IOAN Undyed store, owned by Industry of All Nations

## Act at the source, not at the finish line

Most brands treat returns as fate. They optimize logistics, renegotiate shipping, and never touch the root. Yet the right question isn't "*how do I handle returns more cheaply*" but "*why are my customers sending things back*."

**In the vast majority of cases, the answer comes down to one thing: the gap between what they expected and what they received.**

Your whole job is to close that gap upstream. But be careful, because two very different paths lower the same number.

The first path tackles the cause: honest product pages, a reliable size guide, products shown on real bodies, everything that helps shoppers buy the right thing the first time.

The other prefers to discourage the return itself, with surprise fees, a maze of a form or a blocked account, until the customer gives up a right that is hers. Here, doing right over pleasing means going after the cause without ever punishing the symptom.

![Knix product page with many reassurance elements about size and returns](https://storage.ghost.io/c/b5/a7/b5a79c8d-ed77-4459-9729-b694bd04720c/content/images/2026/10/knix-return.jpg)

Knix highlights many reassurance elements to back up the shopper's choice (and ultimately lower the return rate)

### Product pages that tell the whole truth

Forget the solo packshot on a white background. What you need to show is the garment worn on several body types, how it drapes in motion, the fabric up close. In a recent workshop with a premium fashion brand, the diagnosis from customers was unambiguous. The head of the brand summed it up in one sentence: "What they're missing is knowing how it looks on someone 5'3", how it looks on a size 42, and we've completely underinvested there."

---

> *"What they're missing is knowing how it looks on someone 5'3", how it looks on a size 42, and we've completely underinvested there."*

---

Unable to picture themselves in it, some shoppers buy anyway, somewhat blind, then return the item because it doesn't fall like it does on the single, slender model on the page.

This isn't just a hunch. On another brand, a mobile A/B test replaced the description accordion with always-visible reassurance icons, and the variant won 23% more add-to-carts among new visitors.

But before rolling it out, the CRO team put five customers through user testing, and a barrier the numbers hadn't caught showed up: they couldn't picture themselves in the product, for lack of photos of it being worn.

The test was good, and yet it was the conversation with real people that opened the next project. Underinvesting in these visuals therefore costs twice: in lost carts at purchase, then in returns after delivery.

Showing the garment on varied bodies is first a question of accuracy (not to mention respect for the individual). It's also, and without the two contradicting each other, a margin driver.

💡

****The test I'd run first —** A short video of the product being worn, in motion, on the SKUs that come back most. Where a still image leaves doubt about the fit, movement answers it.

---

### The size puzzle

As we said, it's the leading cause of returns in fashion. And it's precisely this number-one problem that the industry handles worst: Baymard Institute found that [83% of apparel sites don't provide enough sizing information](https://baymard.com/blog/apparel-size-information?ref=conversion.club). Most still bury it in an accordion or a PDF at the bottom of the page, when it should sit right next to the size selector, at the exact moment the decision is made.

An interactive guide that recommends a size based on measurements or on a known size from another brand beats a static chart.

And sometimes it takes almost nothing: a note like "*the model is 5'9" and wears a M*" gives an immediate reference point.

On the brand mentioned above, the information already existed, but buried in an accordion nobody opened. The first decision wasn't even to create any, just to take it out of hiding.

**Copy-and-paste microcopy, seen in the field:**

| **Under the selector**          | *"The model is 5'9" and wears a M."*                                                                             |
| ------------------------------- | ---------------------------------------------------------------------------------------------------------------- |
| **On an item that runs large:** | *"This style runs large, go down one size."* A precise instruction, never the vague *"if you're between sizes…"* |
| **At add to cart**              | *"You chose a 38\. Confirm?"*                                                                                    |

💡

****The hypothesis to test** — Show the size the model is wearing directly under the selector. If returns for sizing errors drop, it means the shopper finally had a reliable point of comparison before buying, not after.

## Join the club

The e-commerce brand community.

Subscribe 

Email sent! Check your inbox to complete your signup. 

Clear, concrete, actionable content.   
Unsubscribe in one click.

### The return that has nothing to do with fit

Here's the signal most brands miss. At the same brand, customer service was receiving returns tagged "wrong size" that had nothing to do with fit.

The founder put it bluntly: *"We get quite a few 'it wasn't my size,' but not because the product runs too small or too big. It's just: I'm a 44 and I clicked a 36."* The garment was perfect; the selector, for its part, hadn't made anyone confirm.

The case isn't isolated. At a shoe brand, size 35.5 came preselected by default, so every 35.5 order had to be confirmed one by one to check that the customer really wanted a 35.5, and not the size she'd forgotten to change. One poorly designed variant, and your support team becomes a call center.

So part of your "size" return rate isn't a pattern-making problem but an interface problem. Good news: this one can be fixed without touching the product. Add a visual size confirmation step before add to cart, and you'll see returns for selection errors recede, because shoppers validate their choice instead of enduring it.

![Silhouette of a rider on the dunes of the Mojave Desert, photo by Terry Eiler, 1971](https://storage.ghost.io/c/b5/a7/b5a79c8d-ed77-4459-9729-b694bd04720c/content/images/2026/10/Silhouette-of-a-Horse-Rider-on-Sand-Dunes.png)

Photo by Terry Eiler, Mojave Desert, 1971

### The return you'll never make disappear: bracketing

Some of your returns are decided at the moment of purchase. Shoppers order two or three sizes of the same item, knowing full well they'll send most of them back; [Narvar](https://corp.narvar.com/resources/2022-returns-policy-benchmark?ref=conversion.club) estimates that 63% of shoppers practice this "bracketing."

For a wedding or an event, she takes five dresses and keeps one. It's neither a mistake nor fraud, just rational behavior that twenty years of free returns have ended up establishing.

The trap is that it looks like a disappointment return when it's the exact opposite. Quite often, the sale only happened because the shopper could hedge her size bet. It's easy to spot, in orders containing two size variants of the same product, but eliminating it would lower the return rate and revenue in the same move.

So the real question isn't how to crush it, but whether you consciously make it easier or discourage it. Some brands embrace it and let shoppers add two sizes from the same page, knowing exactly what they're doing. In any case, the best antidote remains the size guide upstream, which reduces the need to bet without killing the sale.

### Your returns are your most honest feedback

Every return tells you something about the product or its promise, provided you listen. In the return form, swap the generic dropdown for precise questions: did the color differ from the screen, did the fabric seem thinner than expected, was the fit tight at the hips?

Each reason also needs an owner, otherwise you pile up complaints nobody takes responsibility for. A reason code is only worth something if it points to a team and a driver.

| Structured reason         | What it says                                   | Who acts                     |
| ------------------------- | ---------------------------------------------- | ---------------------------- |
| "Doesn't fit as expected" | Wrong pattern or size guide                    | Product / Product page       |
| "Not as shown in photos"  | Misleading visuals or description              | Product page / Art direction |
| "Wrong item received"     | Fulfillment error                              | Logistics                    |
| "Size error at selection" | Selector that doesn't prompt confirmation      | Tech / UX                    |
| "Changed my mind"         | Poorly qualified traffic, unclear expectations | Acquisition                  |
| "Disappointing quality"   | Sourcing, material, finishing                  | Purchasing / Production      |

Once the reasons are centralized, read them SKU by SKU. At that same brand, a manager spotted a skirt whose returns were concentrated in three sizes, while the other two caused no problem: the garment simply ran large in those sizes.

The answer wasn't a vague *"if you're between sizes"*, but a clear instruction on the page, *"go down one size"*, while the pattern went back to the supplier for the next production run.

A generic message assumes the shopper is hesitating and simply walks her through it; a precise message shows her where the problem is, and makes returns drop for good.

![Skater, photograph by Sébastien J. Zanella](https://storage.ghost.io/c/b5/a7/b5a79c8d-ed77-4459-9729-b694bd04720c/content/images/2026/10/Skater.png)

Photograph by Sébastien J. Zanella

## Returns as a loyalty driver

Even with the best prevention in the world, you'll always have returns. Still, how you handle them can just as easily damage the relationship as strengthen it.

Start by simplifying.

A self-service portal where shoppers print their own label in a few clicks is no longer an advantage, it's a standard, and its absence is a reason to abandon: 76% of shoppers factor in free returns when deciding where to buy.

Then think about offering something other than a plain refund, which remains a dead loss.

Track your three outcomes separately, too: refund, exchange and store credit, because they don't weigh the same on cash flow. A bonus credit or a direct exchange for another size, shown before the "refund" option, preserves cash while keeping the shopper with the brand.

Then communicate at every step: "request received," "package in transit," "credit issued." Three lines in an inbox, right when the shopper wonders whether she was wrong to trust you, and that's where transparency works for you.

💡

****The hypothesis worth testing —** Move the exchange and the bonus credit ahead of the refund in customer support. If the share of returns converted into credit or exchange rises, it means the default option was enough to steer the decision without ever forcing it.

## Measuring and managing your return rate

This is where most articles let you down, with a more or less useless "*track your return rate*." The overall return rate is a vanity metric that hides more than it shows, even more than you imagine: a perfectly stable rate can easily mask a SKU that's deteriorating, simply because your catalog mix has shifted. The average holds steady while the ground gives way under one SKU.

Before measuring, make what you measure reliable. At a brand we audited, GA4 showed a checkout rate of 130%, which is to say unusable data.

A wrong dashboard does more damage than no dashboard, because it gives you confidence without the truth that should come with it.

Once the data is clean, three units matter. First, return rate by SKU, then, within a SKU, by size, exactly as the skirt showed. Second, the reason, structured and aggregated, which turns an isolated complaint into a product signal. Finally, net revenue after returns, the only number that really tells you what you take in.

## Join the club

The e-commerce brand community.

Subscribe 

Email sent! Check your inbox to complete your signup. 

Clear, concrete, actionable content.   
Unsubscribe in one click.

### The trap almost everyone ignores: latency

A return doesn't land on the day of the sale. It arrives two, three, sometimes six weeks later. Launch a product video on a Monday, read your return rate on Friday, and you'll mix orders still inside their return window with others already closed out: the number will mean nothing.

The right unit of analysis is the cohort. You isolate the orders placed during the test, let them run to the end of their window, at D+30 or D+45 depending on your policy, and then compare variant against control on that closed scope.

The framework to lock in before launching, to copy and fill in without deviating:

- **Primary metric:** contribution margin per visitor, net of returns.
- **Scope:** orders placed between \_\_\_ and \_\_\_.
- **Observation window:** D+30 (or D+45 depending on your return policy).
- **Minimum detectable effect:** \_\_\_% (below that, the test isn't worth running).
- **Stopping rule:** read the result when the last order in the cohort has finished its window, never before, even if the curve is tempting.

### The four-level diagnosis

To avoid confusing a symptom with its cause, here is a four-level thinking protocol you can adopt as is, each level calling for a different mindset. It works as well for conversion as for returns.

#### Analysis, the analyst's mindset

Have I properly separated the symptom from the cause? **"Wrong size"* is a symptom whose cause can be a faulty pattern, a hidden guide or a selector that doesn't prompt confirmation: one symptom, three causes, three teams.

#### Hypothesis, the experimenter's mindset

Is my hypothesis falsifiable, and can it be phrased as "if X, then Y"? Did I set my metric, duration and threshold before launching, and above all consider the opposite hypothesis, to know what a failure would teach me?

#### Hypothesis, the experimenter's mindset

Is my hypothesis falsifiable, and can it be phrased as "if X, then Y"? Did I set my metric, duration and threshold before launching, and above all consider the opposite hypothesis, to know what a failure would teach me?

#### Review, the reviewer's mindset

Have I confronted the result with real people? A winning test isn't an end in itself, it's more the door to the next barrier, the one the data alone doesn't reveal.

Each level demands its own mindset, and the rigor of the diagnosis determines the quality of the test; a good test poorly conceived is still a bad test.

The protocol is no theoretical luxury: by structuring our work this way for more than a year, we ended up gaining 4.4 points of success rate year over year.

### The tester's humility: even the obvious gets it wrong

The best proof that this framework is useful is the failures it brings to light.

At one brand, a hypothesis as consensual as *"free shipping from €300"* lifted add-to-cart by 2.6%, while lower in the funnel, conversion collapsed by 24% and average order value by 36%.

Without rigorous tracking, we'd have kept the +2.6% and trumpeted a win; with the framework, we saw the loss. CRO works by compounding, each real gain building on the last and even benefiting paid, on the sole condition of never mistaking an apparent gain for a real one.

### Instrumentation

No need for a €40,000 tool: everything can be pulled straight from Shopify with ShopifyQL. Here are four queries to copy and paste, adapting them to the column names in your schema if needed.

#### The first points out the culprits: return rate by SKU

```ShopifyQL
-- 1. Taux de retour par référence
FROM sales
SHOW
    sum(quantity_returned) AS units_returned,
    sum(quantity_sold) AS units_sold,
    (sum(quantity_returned) * 100.0 / sum(quantity_sold)) AS product_return_rate
BY product_title
SINCE -12m
ORDER BY product_return_rate DESC
```

![ShopifyQL query: return rate by SKU](https://storage.ghost.io/c/b5/a7/b5a79c8d-ed77-4459-9729-b694bd04720c/content/images/2026/10/return-sku-1.jpg)

#### The second drills down to size, the level of detail everything else depends on

It's what brings out the skirt that runs large in three sizes and stays healthy in the other two.

```ShopifyQL
-- 2. Taux de retour par référence ET par taille
FROM sales
SHOW
    sum(quantity_returned) AS units_returned,
    (sum(quantity_returned) * 100.0 / sum(quantity_sold)) AS product_return_rate
BY product_title, variant_title
SINCE -12m
ORDER BY product_return_rate DESC
```

![ShopifyQL query: return rate by size](https://storage.ghost.io/c/b5/a7/b5a79c8d-ed77-4459-9729-b694bd04720c/content/images/2026/10/return-size.jpg)

#### The third tells you what you actually take in: net revenue after returns, with its retention rate

```ShopifyQL
-- 3. Revenu net après retours
FROM sales
SHOW
    sum(gross_sales) AS revenu_brut,
    sum(net_sales) AS revenu_net,
    (sum(net_sales) * 100.0 / sum(gross_sales)) AS taux_de_conservation
BY product_title
SINCE -12m
```

![ShopifyQL query: returned amount and cash flow impact](https://storage.ghost.io/c/b5/a7/b5a79c8d-ed77-4459-9729-b694bd04720c/content/images/2026/10/return-cash.jpg)

#### The fourth gives the trend, month after month

ideal for spotting a drop after a change of carrier or supplier.

```ShopifyQL
-- 4. Taux de retour mois par mois
FROM sales
SHOW (sum(quantity_returned) * 100.0 / sum(quantity_sold)) AS product_return_rate
BY month
SINCE -12m
```

![ShopifyQL query: return rate by month](https://storage.ghost.io/c/b5/a7/b5a79c8d-ed77-4459-9729-b694bd04720c/content/images/2026/10/return-month-1.jpg)

The day these four tables run every week, returns stop being a fate and become a list of culprits.

### The weekly review of at-risk product pages

The query gives you the culprits, but not their motives, and no number alone will tell you why a page disappoints. Quantitative data points to the page, qualitative data explains the friction. Hence this ritual: 45 minutes a week, with a creative, a CRO and a product representative in the same room.

#### Selection

Five product pages, taken from the SKUs whose conversion falls below their category average or whose return rate is rising. The two lists often overlap.

#### Observation

Six session recordings per page, three on desktop, three on mobile, watched with no sound and no commentary during playback, because you observe before you interpret.

#### Cross-checking

We compare what we saw with the hard data: GA4, return rate by size, structured reasons. Hesitation on the selector makes total sense once you know that size accounts for most returns.

#### Prioritization

Two or three dominant signals per page, one test per signal, in this order of impact: architecture, then copy, then visuals, then microcopy. Move a piece of information before rewriting a word.

Without this framework, a session review turns into free-form commentary and everyone reads the same footage their own way. With it, creative, CRO and tech finally speak the same language, and a real test comes out of the meeting rather than a list of "ideas to explore."

And always with the same rule as a safeguard: the test you pick must prove it wins on conversion without pushing returns up.

### Where to start on Monday morning

If you could keep only one playbook, it would be this one.

1. **Export 90 days of returns** into a spreadsheet, or run query #1 above.
2. **Set up a taxonomy of 5-6 actionable reasons** (the table above) and assign an owner to each.
3. **Sort by SKU, then by size,** and flag any SKU above 15-20%.
4. **Read the dominant reason** for each culprit. Rule of thumb: 80% "too small" betrays a faulty size guide, 60% "not as shown in photos" means visuals to redo.
5. **Fix the page of the worst culprit** (precise size message, worn photo, model reference), one variable at a time.
6. **Measure by cohort, not by calendar,** then document the result so you don't rerun the same test next year.

## Not all returns are equal

Before going to war, a nuance every seasoned operator knows: there isn't one type of return but several, and confusing them means shooting at the wrong target.

First, the disappointment return: a wrong size, a betrayed color or quality below the promise. A dead loss paired with an error signal, it's the only one you really need to crush.

Next, the confidence return, when the shopper dared to order precisely because she knew she was free to send it back: on a premium basket, that assurance often makes the difference between purchase and abandonment, and part of your returns is just the price of revenue you wouldn't have had otherwise.

Finally, the strategy return, the bracketing we met above, planned and rational, a cousin of the previous one, to be handled rather than eliminated.

One last trap, in mirror image: a very low return rate isn't necessarily good news:

- either your product wins people over and perceived value matches received value, which is the ideal;
- or your return policy is so locked down that people give up on returning, and then you're not keeping happy customers, you're piling up disappointed ones who won't come back.

Zero returns can hide a leak more expensive than the returns themselves.

That's why "reducing returns" is a bad goal. At some stores, the customers who return the most are also the ones who spend the most and stay the longest.

The goal was never to crush volume, but to eliminate disappointment without sacrificing the trust that makes people buy. Once you hold that course, the rest falls into place on its own: you reduce disappointment upstream, protect the relationship on arrival, and measure SKU by SKU to turn both into decisions. Gross flatters; only net judges.

## FAQ

#### What is a good return rate in fashion?

There's no universal threshold. Fashion carries structurally high rates, around 20-30% online, while other categories stay much lower.

The Shopify baseline is closer to 18-19%, and across the stores we follow, the median sits between 8% and 15% product return rate.

The right benchmark isn't the market average but your own history, tracked by SKU and over time, keeping in mind that an unusually low rate can signal a locked-down policy rather than a well-loved product.

#### How do you reduce your return rate without hurting the customer experience?

By acting upstream: precise product pages, visuals in motion and on several body types, model size under the selector, reliable size guides, and systematic analysis of reasons to fix faulty pages and products alike.

Also think about returns that aren't fit problems but simple selection errors, which a size confirmation often solves.

The goal isn't to hold back returns, it's to prevent the buying mistake that causes them, never adding friction to discourage a legitimate right.

#### How do you know whether a CRO test really reduced your returns?

By reading both curves together, revenue per session and return rate: a test that wins on one while degrading the other is a false positive.

And by thinking in cohorts, because of latency: isolate the orders from the test period and wait for the end of their return window, at D+30 or D+45, before concluding.

Reading the return rate a few days after a change measures nothing but noise

#### Should you charge for returns to protect your margin?

It's a trade-off, not a no-brainer. Paid returns reduce volume but can discourage the initial purchase, especially on a high basket, while 76% of shoppers factor in free returns when choosing where to buy.

It's the path Zara, H&M and ASOS have taken since 2023, effective on volume but risky for acquisition. Before touching your policy, it's more profitable to tackle the causes of returns and steer unavoidable returns toward exchange or store credit.

####