AI FRAME RESEARCH · MERCHANT CASE STUDY

21,095 size recommendations in nine weeks.

What happened when an international fashion retailer made personalised sizing and Virtual Try-On part of the shopping experience — and what the shopper activity reveals.

FASHION & APPAREL · SIZE RECOMMENDATIONS · VIRTUAL TRY-ON · SEPTEMBER–NOVEMBER 2026
PERSONALISED SIZE RECOMMENDATIONS SHOWN 21,095 Across 4,929 shoppers
4,929 Shoppers who received a size recommendation
37 Products with size recommendations
WATCH THE CASE STUDY

The whole story in one video.

Prefer to watch? Here is what this fashion retailer learned from 21,095 size recommendations, in a few minutes.

The video loads from YouTube when you press play. Watch on YouTube

THE OPPORTUNITY

Make choosing a size — and seeing the product on yourself — part of the shopping experience.

Choosing a clothing size online often means interpreting measurements, comparing tables and deciding which option feels right. And even after choosing a size, shoppers may still wonder how a garment will actually look on them.

This international fashion retailer introduced AI Frame's Size Recommendations and Virtual Try-On to give shoppers two different kinds of guidance directly while browsing.

Across the store, thousands of shoppers used personalised sizing. On the dress store, shoppers also actively requested Virtual Try-On images and waited for the results.

IMPLEMENTATION
Same day

The Size Recommender was installed and the first live recommendation was available on 4 September 2026.

01 · SIZE RECOMMENDATIONS · COMMERCIAL SIGNAL
OBSERVED ORDER RATE

Shoppers who used size recommendations were 21× more likely to order.

Shoppers who received a size recommendation had a substantially higher observed order rate than product-page visitors without a recorded recommendation.

Shoppers who used a size recommendation 15.6%
Placed an order
Product-page visitors without a recommendation 0.7%
Placed an order
21× Higher reported order rate among shoppers who used the recommendation feature.
Important context: This is an observed association, not proof that the recommendation caused the purchase. Shoppers who chose to use the feature may already have had stronger purchase intent. The displayed percentages are rounded; confirm the underlying unrounded rates before publishing the 21× comparison.
02 · SIZE RECOMMENDATIONS · VISITOR CONTRIBUTION

2.4% of visitors accounted for 37% of orders.

A relatively small group of visitors used the size recommendation feature, but shoppers in that group placed a disproportionately large share of the store's orders.

Share of visitors and orders

4,929Of 207,104 visitors
744Of 2,020 orders

A small group, with a meaningful presence in the order data.

Only around 2.4% of the recorded visitor population received a size recommendation. Those shoppers were associated with 744 of the store's 2,020 orders.

This highlights why feature adoption alone can hide the commercial relevance of a more engaged group of shoppers.

The opportunity is to make personalised sizing easier to discover and use across more relevant product pages.

These are descriptive comparisons. They do not establish incremental orders or causation.

03 · SIZE RECOMMENDATIONS · REPEAT USAGE

One in three shoppers used their size on two or more products.

Once shoppers had a Fit Profile, they could use it across multiple products without entering their details again.

1,618

Shoppers received size recommendations on two or more products — approximately one in three of the 4,929 shoppers.

32.8%
Of shoppers who received a recommendation

Enter once. Use across products.

A traditional size chart presents sizing information for an individual product. Shoppers have to interpret that information again as they move through the store.

With an AI Frame Fit Profile, shoppers can reuse their saved details when exploring other supported products.

The data shows that this was not just a one-product interaction for a meaningful group of shoppers.

01
Enter detailsCreate or use a Fit Profile.
→
02
View a sizeGet a recommendation for the first product.
→
03
Explore moreReuse the profile on other products.
04 · SIZE RECOMMENDATIONS · LESS FRICTION

10,150 size answers were worked out from height and weight alone.

Personalised sizing does not always require shoppers to measure and manually enter multiple body measurements.

AI Frame can use height and weight to create an estimated Fit Profile, which can then be used with the merchant's product size chart.

In this merchant's data, 93% of saved Fit Profiles used estimated measurements rather than manually entered body measurements.

⌁
Manual measurements Shoppers measure and enter body measurements themselves.
↗
Estimated profile Shoppers provide height and weight to create a profile.
93%

Of saved Fit Profiles used estimated measurements.

Estimated profiles93%
10,150
Size answers from height and weight alone

The 93% figure relates to saved Fit Profiles. It is not a percentage of all visitors or all size answers.

05 · SIZE RECOMMENDATIONS · RECOMMENDATION ALIGNMENT

Eight in ten buyers purchased the recommended size.

The size recommendation is most useful when it helps shoppers identify a size that they ultimately choose.

In this merchant's analysis, eight in ten buyers purchased the size that AI Frame had recommended.

This provides an indication of how closely the recommendations aligned with the sizes shoppers selected when placing an order.

RECOMMENDED SIZE PURCHASED
8 in 10

Eight of every ten buyers in the analysed group purchased the recommended size.

The exact buyer count and denominator should accompany this statistic in the final published methodology.

06 · VIRTUAL TRY-ON · REAL SHOPPER ACTIVITY
DRESS RETAILER · THREE-WEEK OBSERVATION

Shoppers waited 14 seconds to see the dress on themselves.

14 sec
Median time to receive a generated Virtual Try-On image.
184 Shoppers used Virtual Try-On
194 Successful renders — each one requested by a shopper
A FEATURE SHOPPERS CHOSE TO USE

They didn't just open the feature. They waited for the result.

In the three weeks following 26 October, 184 shoppers used Virtual Try-On on this dress retailer's store.

Across the period, 194 renders successfully generated the images shoppers had requested. The median wait was 14 seconds.

This is a simple but important signal: shoppers were willing to pause their browsing and wait to see how a dress could look on themselves.

CONSISTENT DAILY USE
19 / 20

Days with successful renders

Successful renders were recorded on 19 of the 20 observed days. Usage was still continuing on 14 November 2026.
11 NOVEMBER 2026
22 renders
Successful Virtual Try-On images generated on that day.
PRODUCT-LEVEL ENGAGEMENT

Three dresses accounted for most of the activity.

250 / 284

Of the 284 Virtual Try-On attempts, 250 were made across just three dresses.

Three most-used dresses88% of attempts

Shoppers concentrated their try-ons on a small group of dresses they were particularly interested in seeing on themselves.

13 Shoppers used the same uploaded photo to try on a second dress. Some shoppers continued exploring different products using the same personal photo.

When Virtual Try-On is available on the product page, shoppers can stop, upload their photo and wait for an answer to a simple question: "Will it suit me?"

Virtual Try-On activity is reported for one dress retailer from 26 October to 14 November 2026. There were 184 participating shoppers and 194 successful renders. The separate attempt analysis recorded 284 tries, of which 250 related to three dresses. The median generation time was 14 seconds. These figures describe usage and engagement, not sales, conversion or revenue impact.

THE BIGGER PICTURE

Help shoppers answer two important questions before they buy.

Which size should I choose? And how could this look on me?

In this merchant's store, thousands of shoppers received personalised size recommendations, reused their Fit Profiles across products and actively requested Virtual Try-On images. The two experiences address different parts of the same shopping decision.

For fashion merchants, the opportunity is to make product pages more informative and personal — helping shoppers explore sizing and visualise products in the way that suits them.

Explore AI Frame for Shopify →
About this case study and methodology

This case study presents aggregated AI Frame Size Recommender and merchant store data. The supplied analysis reports 21,095 size recommendations across 4,929 shoppers and 37 products over nine weeks.

The reported order-rate comparison is 15.6% among shoppers who used a size recommendation and 0.7% among product-page visitors without a recorded recommendation. The reported 21× comparison is observational, not a controlled experiment. Shoppers who chose to use the feature may have had different purchase intent from those who did not.

The visitor/order share comparison uses 4,929 of 207,104 visitors and 744 of 2,020 orders. These figures describe observed activity and do not establish incremental orders or causation.

The repeat-use figure represents 1,618 shoppers who received a size recommendation on two or more products. The 10,150 height-and-weight answers and 93% estimated Fit Profile figure describe different measures and should not be treated as interchangeable.

The eight-in-ten buyer result is presented as supplied. Its precise denominator and buyer count should be confirmed in the underlying analysis.

Virtual Try-On activity is reported separately for one dress retailer from 26 October to 14 November 2026. The analysis includes 184 shoppers, 194 successful renders, a 14-second median wait, and a separate 284-attempt analysis. Of those attempts, 250 related to three dresses. Thirteen shoppers reused the same photo on a second dress. These figures describe feature usage, not sales or conversion.

Merchant identity has been withheld. Figures are based on aggregated merchant data. Results describe this merchant's experience and should not be interpreted as a guaranteed outcome for other stores.