Size recommendation: how to choose a solution for a fashion brand
Size and fit are the leading cause of online apparel returns: 53% of the apparel brands and retailers surveyed by Coresight Research name them as the top reason. That is why size recommendation has become a category of its own in fashion e-commerce.
Every solution in this category answers the same question — "which size should I order?" — but not in the same way, and above all not by asking the same thing of the shopper. What a tool requires from the visitor is what decides its real usage rate, and therefore its effect on your returns.
This page sets out the three approaches that exist, what each one assumes, and the criteria to check before choosing. We publish OptiDress, one of the solutions in this category; the criteria below apply whichever one you pick.
The three approaches
1. Recommendation based on purchase history
The principle: cross what the shopper has bought and returned in the past, at this brand and at others, to infer their size.
This is the sector's historical approach, carried by the most established American and European players — True Fit, Fit Analytics. Their strength is the volume of purchase data accumulated over years, which cannot be reproduced from a standing start.
What it assumes: that the visitor is recognised, and therefore has a usable history. For a first-time visitor, the data does not exist. The entry price of these solutions also places them beyond the reach of most independent brands.
2. Recommendation based on a photo
The principle: the shopper submits a photo of themselves, and the solution derives measurements from it, or renders the garment onto it.
What it assumes: that the shopper agrees to photograph themselves mid-purchase. Every visitor who declines is a visitor the tool never serves — and no vendor in this category publishes that share, ourselves included.
It also raises a personal-data question: a photo of an identifiable person is personal data; it becomes biometric data when it is processed by technical means that allow the person to be identified. Check the retention period and the processing location before you commit.
3. Recommendation based on garment measurements
The principle: cross the characteristics the visitor states — height, weight, sex, body shape, measurements — with the garment's real measurements, size by size, supplied by the brand.
What it assumes: that the brand has its product tech packs. This is the main constraint of the approach, and it is organisational rather than technical. The measurements almost always exist, but they are held by the studio or the product development team, not by the e-commerce team.
In exchange, it asks the visitor for no photo, no account and no history. It works from the first visit.
Comparison table
| Purchase history | Photo | Garment measurements | |
|---|---|---|---|
| What the shopper provides | Nothing, but must be recognised | A photo of themselves | Height, weight, body shape |
| Works on a first purchase | No | Yes | Yes |
| Sensitive personal data | Transaction history | Image of the person | Stated characteristics |
| What the brand must provide | Order and return feeds | Product imagery | Measurements by size |
| Shows the garment | No | Depends on the solution | Depends on the solution |
Five criteria to check before choosing
1. What the tool asks of the shopper. This is the criterion that sets the usage rate, and the usage rate sets everything else. A tool nobody opens produces no measurable effect, however good its algorithm.
2. Where the garment data comes from. A recommendation computed from a generic "usual size" cannot tell a fitted cut from an oversized one in the same catalogue. Ask whether the solution uses the real measurements of each reference, and how many measurements per size.
3. What happens between two sizes. This is the most frequent and the most expensive case. A useful recommendation states not only the size, but how the garment will sit: where it will be close to the body, where it will fall loose.
4. Who enters the measurements, and how. If the solution demands a clean format, the integration fails as surely as manual entry does. Ask whether it accepts your tech packs as they actually exist — spreadsheet, PDF, mess included.
5. Whether visualisation is built in or separate. Most vendors treat try-on and size recommendation as two distinct products, sometimes from two suppliers. One journey, one integration and one visitor profile avoid a double friction.
What about size charts?
A size chart is not a recommendation: it is a table the customer has to read, interpret and compare against their own measurements, which they rarely know. It is the most widely installed category on Shopify, and the least engaging for the visitor.
It remains useful as a complement. It does not answer the question being asked in front of the size selector.
Where OptiDress sits
OptiSize is the native size recommendation inside OptiDress. It belongs to the third approach: it crosses the characteristics the visitor states — no photo, no account, no history — with four measurements per size supplied by the brand for each reference, and states the size to order along with a description of how it will sit.
The same profile drives the virtual try-on on the visitor's 3D avatar: one integration, one journey, and the avatar kept for thirty days on the store.
The whole thing installs self-serve on Shopify and by assisted setup on other platforms, on a subscription sized to try-on volume with a decreasing overage rate, and no commitment.
Frequently asked questions.
What is the difference between a size chart and a size recommendation?
A size chart displays a table of measurements the customer has to interpret themselves. A size recommendation states a specific size for this customer and this product, computed from their characteristics and the garment's measurements.
Do you have to send a photo to get a size recommendation?
It depends on the solution. Photo-based approaches require one; approaches based on garment measurements and stated characteristics do not.
Does a size recommendation work on a first visit?
Solutions based on purchase history need a recognised visitor. Solutions based on stated characteristics work from the first session.
What product data do you need to provide?
The real measurements by size, taken from your tech packs. The question to ask internally is who holds them: it is rarely the e-commerce team.
Does size recommendation reduce returns?
It addresses one cause of returns — size and fit, named as the top reason for online returns by 53% of the brands surveyed by Coresight Research. How large the effect is depends on the share of visitors who actually use the tool. No vendor in this category, ourselves included, publishes a figure you can verify on your own store.
See also: Size recommendation, explained · Virtual try-on without a photo · Virtual try-on apps for Shopify · OptiDress vs Veesual · OptiDress vs Genlook · OptiDress vs Kleep · Reducing returns: which lever for which cause
Compiled in October 2026 from publicly available information (vendor websites and official communications). True Fit and Fit Analytics are trademarks of their respective owners. This information may change; tell us about any inaccuracy at contact@optidress.fr.