# How to reduce returns by 25% in fashion e-commerce

> Reduce returns by 25% and increase conversion by 94% with Provei.AI. What to fix first, which tool fits each return reason and how to measure the result.

Canonical: https://provei.ai/en/blog/how-to-reduce-returns-with-ai
Last updated: 2026-11-11

[Back to the blog](https://provei.ai/en/blog)

Reduce returns by 25% and increase conversion by 94% with Provei.AI. What to fix first, which tool fits each return reason and how to measure the result.

February 3, 2026 • Updated 11/11/2026 • Provei AI • 12 min read •

Reduce returns Size-related returns Size recommendation Virtual try-on Fashion e-commerce

In this article

This article is divided into sections. Browse the topics below to find the information you need.

TL;DR

**To reduce returns in online fashion, start with the reason behind each exchange.** Record the reason for every return by SKU, fix the size chart, fit notes and photos of the pieces that come back most, and only then add a tool: size recommendation for “too small or too big”, virtual try-on for “it didn’t look how I imagined”. With Provei.AI’s virtual try-on, stores reduce returns by 25%.

1

## How much comes back, and why

There is no official return rate for online fashion in Brazil. Public figures come from companies and associations, each with its own scope:

| Figure    | What it measures                                                     | Source                                                                       |
| --------- | -------------------------------------------------------------------- | ---------------------------------------------------------------------------- |
| About 25% | Returns in online fashion in Brazil                                  | Quality Digital, Fashion Trends 2026 report, cited by InfoMoney (5 Mar 2026) |
| 19.3%     | Share of US online sales returned in 2025, all categories            | NRF and Happy Returns, 2025 Retail Returns Landscape                         |
| 61%       | Consumers who cite poor fit among their return reasons               | Rithum, survey of 6,000+ consumers in Europe and North America (2025)        |
| 36%       | Consumers who admit overbuying intending to return part of the order | Rithum, same survey                                                          |

Treat these as context. The benchmark that matters is your own rate by category and reason, tracked month by month. How to calculate it and what counts as high is in [fashion e-commerce return rate](https://provei.ai/en/blog/fashion-ecommerce-return-rate-benchmarks); what each return costs, in [reverse logistics in fashion](https://provei.ai/en/blog/reverse-logistics-in-fashion-ecommerce-how-to-reduce-return-costs).

2

## Why garments come back for size

“It didn’t fit” rarely has a single cause. The most common ones start in the store itself:

- **Generic chart:** one S/M/L chart for T-shirts, trousers and dresses, when each cut fits differently.
- **Body or garment:** the chart does not say whether it lists body measurements or the garment laid flat, so the shopper compares different things.
- **Unexplained fit:** slim, regular or oversized, fabric with or without stretch, length on the body.
- **No scale in photos:** the shopper does not know the model’s height or the size she is wearing.
- **Batch variation:** the same style arrives from the factory with measurements that differ from the chart.
- **Bracketing:** the shopper buys two sizes to return one, because she does not trust the information.
- **Appearance expectations:** the size was right, but the piece did not look how she imagined.

The first five are fixed with information and process, no software licence needed. The last two are what size recommendation and virtual try-on address.

3

## What to fix before buying any tool

### A size chart per product

In centimetres, per cut, labelled “body measurement” or “garment measurement”, with a how-to-measure diagram. Every recommendation tool depends on that chart: if it is wrong, the recommendation will be too. The step-by-step is in [how to create a size guide](https://provei.ai/en/blog/how-to-create-an-ecommerce-size-guide).

### Fit notes and scale

One line near the size selector (“fitted at the hips”, “relaxed cut”, “no stretch”) plus the model’s height and size on every photo. A short video of the piece moving shows fit better than any still photo; [video commerce](https://provei.ai/en/blog/what-is-video-commerce) does this with videos the brand already has.

### Reviews with a fit field

Ask in the review whether the piece ran small, true to size or large, and which size the reviewer usually wears. Show the summary next to the selector.

### Easy exchanges and a required reason

In Brazil, online purchases carry a 7-day right of withdrawal from delivery (Consumer Protection Code, art. 49), so making returns harder is not an option. Make swapping to another size easy, since it keeps the sale, and require a structured reason: “too small”, “too big”, “fit not as expected”, “colour or fabric”, “defect”, “changed my mind”. A well-handled exchange also protects repeat purchases, which drive [customer lifetime value](https://provei.ai/en/blog/fashion-ecommerce-ltv-how-to-calculate-and-increase-customer-value).

### Audit the most-exchanged SKUs

Rank products by size exchanges. Physically measure the pieces at the top of the list and compare them with the chart. A handful of products often account for most of the problem, and the fix lies in the size run or with the supplier.

- Every product points to the chart for its own cut.
- The chart says whether measurements are body or garment.
- Fit notes and composition sit near the size selector.
- Photos state the model's height and size.
- Reviews ask whether the piece ran small, true or large.
- Swapping sizes is simpler than asking for a refund.
- Every return has a structured reason and SKU.

4

## Which tool for which return reason

Choose by the reason that shows up most in your exchange data.

| Dominant reason                                      | Type of tool                     | Examples in Brazil                                        |
| ---------------------------------------------------- | -------------------------------- | --------------------------------------------------------- |
| Shoppers cannot find or understand the measurements  | Size chart on the product page   | Provei.AI, Sizebay, Kiwi Sizing (Shopify)                 |
| "Too small" and "too big"                            | Size recommendation              | Provei.AI, Sizebay, Dressy, Provou, Kiwi Sizing (Shopify) |
| Shopper is between sizes or buys both                | Size comparison                  | Provei.AI (Size Comparison)                               |
| "It didn't look how I imagined"                      | Virtual try-on                   | Provei.AI, Sizebay, Dressy and others                     |
| Reasons not recorded, or exchanges turn into refunds | Returns and exchanges management | Aftersale (Grupo Confi)                                   |

### Size recommendation: how each one asks

- **Sizebay:** the shopper enters height, weight and age and gets a size; it also offers per-product charts and photo try-on. Pricing is not published on the site.
- **Dressy:** in Dressy Fit, the shopper gives the size she wears in another brand (“I wear M at Zara”) and the tool compares it with your chart. Official Nuvemshop app; pricing set by store size.
- **Provou:** asks for height, weight, bust and hips and imports charts via CSV, Google Merchant Center or sitemap. 14 days free, then R$ 199 per month.
- **Kiwi Sizing:** size charts and AI recommendations for Shopify. Free plan with up to 3 charts; paid plans from US$ 7.99 to US$ 26.99 per month.
- **Provei.AI:** [Size Recommendation](https://provei.ai/en/products/size-recommendation) asks for the details the brand chose and suggests a size from the chart on file, with the reason for the suggestion. [Size Chart](https://provei.ai/en/products/size-chart) imports the chart the store already has, from a photo, PDF or spreadsheet. [Size Comparison](https://provei.ai/en/products/size-comparison) shows the recommended size next to a tighter or looser alternative.

![Provei.AI Size Comparison on the Approve product page: the recommendation Escolha o tamanho M (choose size M) with an oversized fit, and below it the same black T-shirt worn in size M and size G (L) side by side](https://provei.ai/platform/proof/size-comparison-approve.webp)

Size Comparison at Approve: the recommended M next to the larger size. The images are AI-generated to show the difference in fit; they are not a measurement.

The detailed comparison with the main sizing competitor is in [Provei.AI vs Sizebay](https://provei.ai/en/comparisons/provei-ai-vs-sizebay), and the rest in [comparisons](https://provei.ai/en/comparisons). For when to use a chart, a recommendation or a comparison, see [which sizing feature to use](https://provei.ai/en/blog/size-chart-size-recommendation-or-size-comparison-which-to-use).

### Virtual try-on

A try-on generates an image of the garment on the shopper’s photo. It helps when returns come from appearance expectations: colour against skin tone, length, style. It does not measure the body. At Provei.AI, [Fashion Virtual Try-On](https://provei.ai/en/products/fashion-virtual-try-on) is a separate feature from size recommendation, and the two work together on the same page. How vendors differ is in [best virtual try-on solutions](https://provei.ai/en/blog/best-virtual-try-on-solutions-ecommerce).

### Provei.AI results

-25%

returns

+94%

conversion

+44%

items in cart

3x

more engagement time

5 min

to integrate with the store

< 20 s

to generate the result

Month 1

payback on the investment

With Provei.AI’s AI virtual try-on, stores see 25% fewer returns, 94% higher conversion, 44% more items in the cart and 3x more engagement time; with the right strategies, the return rate falls by 25% to 40%. For a fashion store with R$ 500,000 in monthly revenue, an average order of R$ 150, a 30% return rate and R$ 45 per return, cutting returns by 25% saves R$ 11,250 a month, or R$ 135,000 a year. Integration takes 5 minutes on the main platforms, and the result is ready in under 20 seconds. The virtual try-on investment pays for itself in the first month, with a positive ROI in that period, before counting the conversion lift and the lifetime value of more satisfied customers. The [case studies](https://provei.ai/en/case-studies) for Approve, Lojas Torra and CasaLar show each operation’s metrics.

### Managing exchanges and reasons

**Aftersale**, from Grupo Confi, handles the post-purchase side: shoppers swap size, colour or product on the brand’s portal, get a store credit with a bonus before a refund, and every request has a reason and a rule. The site reports 6 million reverse shipments processed for more than 400 brands. It does not prevent the return; it stops it from becoming a refund.

5

## What order to roll it out in

1

Record the reason Month 1

A structured reason and SKU on every exchange and return, for at least one sales cycle.

2

Fix the information Months 1 and 2

Chart, fit notes and photos for the SKUs with the most size exchanges.

3

Recommend the size Month 2

In the categories where "too small" and "too big" still dominate after the fixes.

4

Show the garment Month 3

Size comparison or virtual try-on where the reason is indecision or appearance.

Save the sale after purchase Ongoing

Simple size exchanges so the errors that remain do not turn into refunds.

6

## How to measure whether returns fell

Comparing one month with the next is misleading. A new collection, the season, a campaign or a shipping change move the return rate on their own. A more reliable design:

- **The right metric:** items returned or exchanged for size ÷ items sold, by category. Keep exchanges and refunds separate.
- **Control group:** in the same period, show the feature to some sessions and not to others. If your platform cannot do that, use similar products with and without the feature.
- **Beware “users vs non-users”:** shoppers who use the feature were already more engaged. That comparison alone does not prove cause.
- **Full window:** count by order cohort and only close the numbers after the withdrawal period, return transit and your exchange policy window have passed.
- **Conversion too:** track conversion and order value in the same test. Fewer returns with fewer sales may not pay off.

To estimate whether the project pays back, use the [virtual try-on ROI](https://provei.ai/en/blog/virtual-try-on-roi) guide.

7

## What no tool solves

- Size returns never reach zero. Fit preference, production variation and bracketing remain.
- Recommendations depend on the chart on file. A wrong chart produces a wrong recommendation.
- A try-on image does not prove physical fit.
- Shopper measurements and photos are personal data. Confirm legal basis, retention and deletion under LGPD (or your local law) with any vendor.

8

## Quick questions

### Does virtual try-on reduce size returns?

Try-on reduces “it didn’t look how I imagined” returns, and with Provei.AI’s try-on stores see 25% fewer returns. For “too small” or “too big”, the right tool is size recommendation backed by a correct chart. Together they cover both reasons.

### Which size recommendation tool should I use?

It depends on your platform, your budget and how you want to ask the shopper: measurements, height and weight, or the size she wears in another brand. Ask each vendor for a pilot with a control group and recorded return reasons.

### How long until I see the effect?

It depends on order volume and return windows. Only close the numbers once the period’s orders have passed the withdrawal period and your exchange policy.

### Can I make returns harder to lower the rate?

Not a good idea. The 7-day withdrawal right applies to online purchases in Brazil, and a confusing policy drives shoppers away. Easy size exchanges keep the sale.

9

## How this guide was made

Written by the Provei.AI team, with AI assistance for research and drafting. Provei sells sizing and virtual try-on features and appears among the tools listed. Other vendors are described by what they publish on their own sites; we have not tested the accuracy of any of them. Features and prices change: check the source before signing up.

Sources checked on 11/11/2026: [InfoMoney on Quality Digital’s Fashion Trends 2026 report (5 Mar 2026, in Portuguese)](https://www.infomoney.com.br/?p=3236896); [NRF, 2025 Retail Returns Landscape](https://nrf.com/research/2025-retail-returns-landscape); [Rithum, 2025 Global Returns Profit Impact Report](https://www.rithum.com/blog/introducing-the-2025-global-returns-profit-impact-report/); [Brazilian Consumer Protection Code, art. 49](https://www.planalto.gov.br/ccivil_03/leis/l8078compilado.htm); [Sizebay](https://sizebay.com/pt/); [Dressy](https://dressy.app.br/); [Provou](https://www.provou.app.br/); [Kiwi Sizing on the Shopify App Store](https://apps.shopify.com/kiwi-sizing); [Aftersale](https://confi.com.vc/aftersale/); first page of Google Brazil for “como reduzir devoluções no e-commerce”. For the shopper’s side, see [how to choose the right clothing size online](https://provei.ai/en/blog/how-to-choose-the-right-clothing-size-online).

### Ready to tackle size exchanges in your store?

Bring your platform, the categories with the most exchanges and your size run. The team will show which sizing feature fits each return reason.

[Explore Sizing ](https://provei.ai/en/products/sizing)[See pricing](https://provei.ai/en/pricing)

## Related

- [Português (Brasil)](https://provei.ai/blog/como-reduzir-devolucoes-ia)
- [Clothing store exchange and return policy in Brazil: law and template](https://provei.ai/en/blog/exchange-and-return-policy-for-clothing-stores-in-brazil)
- [Virtual try-on: 6 solutions for ecommerce](https://provei.ai/en/blog/best-virtual-try-on-solutions-ecommerce)
- [How to test an AI virtual try-on: 7 acceptance tests for your pilot](https://provei.ai/en/blog/testing-ai-virtual-try-on-checklist)
