The Myth of AI Sizing: Why Product Data Beats Big Data in Apparel Ecommerce
Understand how AI sizing works and why garment data and human body measurements remain the most effective way to work. Most AI sizing solutions rely on purchase history or cross-brand comparisons to estimate sizing; but is this the best approach with limited sales volumes?
The Assumption Behind “AI Sizing”
If you have already looked into AI sizing, you have probably heard of big data. The concept is intuitive: if an algorithm analyses enough purchases and returns, it should eventually understand how a brand fits.
Indeed, most sizing tools follow one of two paths. Some analyze purchase and return behavior, looking at what customers buy, keep, or send back. Others rely on brand comparison, asking shoppers which brands and size they usually wear and cross reference this data with shoppers with similar inputs.
These approaches can work fine for very large retailers with thousands and thousands of weekly transactions. Most Shopify brands, however, operate at a different scale. Collections change often, product runs are smaller, and individual SKUs rarely accumulate enough data for these models to stabilize.
For brands releasing new collections every season, the model rarely gathers enough data before the cycle begins again. This is what is known in the AI world as the Cold Start problem!
When a new product launches, the recommendation algorithm has very little to learn from. To produce a recommendation, it often falls back on rough comparisons with similar products—or even sometimes products from different categories, and needs months to learn from behavioral or returns history, making it innaccurate in the meantime.
A Different approach: The Digital Centimeter
When Measmerize first started building sizing technology, this limitation became obvious very quickly.
Many of the brands we worked with— even large luxury retailers —simply did not generate the massive datasets algorithms depend on. Waiting for thousands of orders per SKU before recommendations became reliable was not a realistic option.
We took a different approach.
Rather than relying primarily on historical transactions, the system begins with something more stable: the physical relationship between the garment and the human body.
Internally, we refer to this principle as the Digital Centimeter approach.
At its core, the logic is fairly simple. Fit can be understood by looking at two key elements:
- the anthropometric size charts and garment measurements that define the product
- the shopper’s body dimensions, collected through a short questionnaire that estimates exact body dimensions.
Once these two datasets are established, the system compares the centimeters (or inches) of the garment to those of the shopper. By factoring in the items' material (elasticity) and intended fit—whether a garment is designed to be fitted, relaxed, loose, or oversized—the engine can generate a reliable size and fit recommendation from day one.
Sales and return data still play a role, but mostly as a feedback and fine-tuning mechanism. Over time, this behavioral data helps refine the model and highlight consumer preferences that may not be visible in measurements alone.
This approach makes a difference for merchants that release new SKUs and collections every season. The recommendation algorithm has a clear understanding of the product itself, and behavioral signals deduced from sales and returns history simply help fine-tune the recommendations rather than teaching the algorithm from scratch.
The Two Levers of Day-One Accuracy
In order to turn our Digital centimeter approach into to an accurate recommendation system for Shopify retailers, two elements are critical: the size chart and calibration.
Both elements connect the brand’s product knowledge with our algorithm’s decision logic.
1. Size chart: the starting point
When it comes to sizing, every creative director or pattern making team needs a starting point, and this starting point is generally a body (anthropometric) or product dimension size chart. This chart might just look like a table with numbers in it; but in reality, it says a lot about the brand and their targeted customers: who the clothes are designed for and how the garments are meant to fit.
That being said, size charts alone rarely gives the full picture. As important as who the chart is designed for, is how those measurements compare with the actual body dimensions of the population. A chart may be internally consistent and yet still sit far from real customers proportions.
This is where Measmerize adds value. When a size chart is ingested into our system, our algorithm analyses it and compares its structure with real body measurement of the population. The goal isn’t just to read the chart—it’s to understand the fit logic behind it.
Our proprietary analysis focuses on identifying potential inconsistencies between your data, consumer’s bodies, and extensive industry benchmarks.
Taken together, these signals help our Size Hub interpret the brand’s sizing philosophy in a much more practical way. Instead of simply displaying measurements, the system starts to understand how the garments are likely to behave on real bodies—allowing recommendations to be accurate from the very first day.
2. Calibration: Adding Human Expertise
Garment measurements tell an important part of the story, but anyone retailing fashion apparel and footwear knows that fit involves nuance. Designers understand which jacket was meant to feel oversized. Merchandisers know when trousers tend to run narrow compared to the rest of the collection. Those insights are hard to capture in raw measurements alone.
For our enterprise clients, Measmerize teams validate this by conducting mystery shopping—visiting boutiques and physically trying on garments to observe how pieces behave on real bodies.
Shopify merchants can apply a similar principle digitally.
Whilst adding their size chart in the Measmerize solution, the merchant actually tests and adjusts the recommendations to ensure that the team’s knowledge is also considered by our algorithm. Verifying and calibrating the size advisor, offers a peace of mind for the brand knowing and results in more qualitative recommendations for the shoppers!
Bringing this calibration step during the setup ensures the recommendation engine reflects the brand’s vision from day one, rather than learning slowly through purchase and returns data. If you want to learn more about this feature, you may want to read: How to: Calibrating Recommendations Before You Launch
Where to start for accurate size recommendations
The starting point for accurate size recommendations usually isn’t data volume. Many brands already hold the key pieces of information needed to guide shoppers properly:
- Brand’s size charts
- Product data and garment measurements
- And the team’s knowledge of how each collection is meant to fit.
When these pieces are structured properly, they provide a strong basis for reliable size guidance.
For Shopify merchants using our solution, putting this into practice is relatively simple. Measmerize integrates directly into Shopify themes, allowing brands to add tools like AI Size Advisors, Smart Size Charts, and measurement guides directly on the product page.
The size chart remains the reference point. Measmerize is just turning your static table into actual guidance that helps shoppers choose their size with confidence.