E-commerce
June 26, 2026
In fashion e-commerce, the wrong size is not a minor irritant. It is a cost: return, exchange, tied-up inventory, refund, disappointed customer, and an extra support ticket.
An AI chatbot size guide does not replace trying clothes on. It transforms an often-ignored chart into a useful exchange: cut, measurements, material, fit preference, customer reviews, and exchange policy.
This article #5 maintains the requested vertical angle: size guides + chatbot + reduction of wrong-size returns, without writing a general article about returns once again.
Summary
Why does sizing have such a big impact on returns?
The customer does not touch the material, does not see the fit on their body, and sometimes compares multiple brands whose sizes do not mean the same thing. They guess, order, and then return.
The NRF / Happy Returns 2025 report estimates that 19.3% of online sales will be returned. In apparel, the pressure is higher, particularly due to fit and bracketing (NRF Retail Returns Landscape).
Angle #5
Unlike general articles on returns, this one targets the pre-purchase moment: helping the customer choose the right size before the package is shipped.
What does a classic size guide often miss?
Context: an S/M/L size chart does not tell you if the fit is slim, straight, or oversized
Method: many customers do not know how to measure their chest, waist, or hips
Product: stretch jeans and rigid jeans are not chosen the same way
Moment: if the guide is hidden in a tab, doubt returns once they reach the cart
A good chatbot does not replace the guide. It makes it readable, engaging, and adapted to the product page currently being viewed.
What data should be connected before recommending a size?
The recommendation must never come from a generic template. It must be based on your data.
Size guide: measurements per size, units, tolerances
Fit note: runs small, normal, large, slim or loose fit
Product: material, stretch, length, size worn by the model
Fit review: customer feedback on running small, large, or true to size
Stock: do not recommend an unavailable size
On Shopify, metafields are useful for structuring this data: fit_note, model_size, fabric_stretch, size_guide_url (Shopify Help, metafields).
Practical format
For each pilot SKU, add at least: fit, stretch level, size worn by the model, advice if between two sizes, guide link, known exception. Without these fields, the bot will mostly just rephrase.
How should the chatbot ask its questions?
A sizing assistant should not ask questions like an administrative form. It must be fast, because the customer is already on a product page.
What size do you most often wear in this type of product?
Do you prefer a fitted or a more comfortable fit?
Do you have your main measurements?
Are you hesitating between two specific sizes?
If it can make a recommendation with just two answers, it stops there. The longer the dialogue, the higher the risk of abandonment.
How do you frame a recommendation without overpromising?
Good phrasing helps the user choose, but leaves some room. It explains the reasoning instead of declaring an absolute truth.
Example
"Based on your measurements and the slim fit of this model, M seems the most consistent. If you prefer a looser fit or if you are between two sizes, L will be more comfortable. Returns are still possible within 30 days."
To avoid
"Take M, it's 100% sure." No serious chatbot should promise a perfect fit.
If the customer provides contradictory information or if the product is sensitive, the best advice is sometimes to hand over to a human with the context already collected.
Where to place the size guide on Shopify?
Product sheet: clear button: "Help choosing size" near the selector
Collection page: help filter by fit, use, morphology
Cart: offer a verification if two sizes of the same product are added
Post-purchase: explain size exchange before refund
The best time remains the PDP, right before the variant choice. That's where the doubt is strongest and correction costs the least.
Useful triggers
Trigger help if the customer opens the guide twice, changes sizes multiple times, stays on the selector for a long time, or adds two sizes to the cart. Avoid aggressive pop-ups as soon as they land on the page.
How do you reduce returns without frustrating the customer?
Bracketing is rational on the client side: they order two sizes to reduce their risk. On the brand side, it increases logistics costs, returns, and blocked stock.
Concrete Action
When the cart contains two sizes of the same product, the bot can suggest: "Are you hesitating between M and L? I can check the fit in 20 seconds to avoid an unnecessary return."
Happy Returns highlights that bracketing is growing among young buyers, especially Gen Z. The goal is not to punish this behavior, but to restore confidence before purchase (Happy Returns / NRF).
What are the rules by fashion category?
Denim: ask for waist and hips, specify stretch and rise
Shoes: clarify EU/UK/US conversion, width, socks, usage
Outerwear: take layering with a sweater or jacket into account
Lingerie: factual, respectful response, escalate if in doubt
Unisex: explain the sizing chart used instead of assuming
A single script is not enough for all fashion. Each category must have its own fit rules.
Example responses
Denim: "These jeans have little stretch. If you are between sizes, choose the larger one to avoid a waist that is too tight at the hips."
Outerwear: "If you often wear a thick sweater underneath, size up. The fit is straight, not oversized."
Shoes: "This model runs true to size in length, but is rather narrow. If you have wide feet, choose a half size up if available."
What signals show that the size guide is incorrect?
Returns: strong wrong size pattern on certain SKUs
Support: repeated questions on the same model
Reviews: mentions of "runs small" or "odd fit"
Cart: frequent multi-size orders
Conversion: abandonment at the moment of choosing a variant
Zalando points out that size and fit vary greatly depending on brand, cut, and category, and uses data, AI, and customer feedback to improve recommendations (Zalando Corporate).
Which KPIs should you track to avoid kidding yourself?
Wrong-size return rate: share of returns related to size
Bracketing: orders with multiple sizes of the same SKU
Assisted conversion: purchase after sizing interaction
Exchange vs refund: is the size corrected or lost?
Sizing tickets: volume before/after on pilot products
Measure on 10 to 20 pilot SKUs, not on the entire catalog. Compare assisted and non-assisted products for 60 days.
Frequent reading error
If sizing conversations increase, it is not necessarily bad. It may mean that the bot is finally catching a doubt that already existed. Look especially at whether wrong-size returns and bracketing are decreasing.
How does Qstomy help fashion brands?
Qstomy can answer sizing questions, explain the fit, compare two sizes, remind of the exchange policy, and hand over to a human if doubt remains high.
DTC Scenario
Women's ready-to-wear brand: wrong-size returns are concentrated on 12 best-sellers. Qstomy reads guides by collection, fit notes, filtered reviews, and stock. The "Size Help" widget triggers on the PDP and multi-size cart.
The expected gain is not just fewer returns. It is also fewer sizing tickets, more exchanges instead of refunds, and product pages enriched from actual questions. See AI sales assistant, AI customer support and request a demo.
Which scripts can we use starting tomorrow?
Usual size script
"If you usually wear M and like a close fit, stick with M. If you prefer more room, especially on this fitted model, go with L."
Hesitation between two sizes script
"Between S and M, I advise M if you are between two measurements or if you don't like tight fits. S will be more fitted."
Lack of data script
"I prefer not to guide you randomly: a reliable measurement is missing for this model. I can transfer you to the team or show you the exchange policy."
Pilot playbook
Choose 10 SKUs with high wrong-size returns
Add fit notes and model size in metafields
Write three standard responses per category
Activate the widget on PDP only
Review 30 sizing conversations per week
Correct guide, product page, or AI response every Monday
QA testing before going live
Ask the bot 20 questions: usual size, between two sizes, loose fit, guide unavailable, out of stock, exchange, lingerie, child, unisex, stretch product. Every response must be clear, sourced, and cautious.
Sources and useful linking
Shopify: Product page best practices
Shopify: AI Personal Shopper
Qstomy: Pre-purchase product questions
Qstomy: Analyze return reasons
Enzo
June 26, 2026
