SmartSize doesn't ask your customers to grab a measuring tape. Instead, it predicts their chest, waist, and hip measurements from a few simple questions. Here's how it works.
What the Quiz Asks
The quiz collects a small set of easy-to-answer inputs:
Height
Weight
Age
Body shape questions — simple visual selectors where customers pick the silhouette that best matches their body (e.g., chest shape, belly shape, seat shape for men; body shape and bust circumference for women)
These questions take most customers under 30 seconds to complete.
How the Prediction Works
SmartSize uses separate machine learning models for men and women, because male and female bodies distribute weight differently.
For women
The quiz collects height, weight, age, bust circumference (self-reported or band size), body shape, and belly shape. SmartSize then:
Uses the self-reported bust circumference directly as the chest measurement
Predicts waist and hip circumference using a trained ML model
For men
The quiz collects height, weight, age, chest shape, belly shape, and seat shape. SmartSize then:
Predicts chest, waist, and hip circumference — all three from trained ML models
In both cases, the models were trained on large anthropometric datasets and produce measurements in centimeters.
Three Ways Measurements Can Be Obtained
Not every customer goes through the full quiz. SmartSize supports three paths:
Path | When it happens | How it works |
Quiz prediction | First-time visitor answers the quiz | ML models predict chest, waist, and hip from quiz answers |
Direct input | Customer knows their measurements | They type exact values (e.g., "my chest is 92 cm") — no prediction needed |
Returning visitor | Customer has used SmartSize before | Measurements are remembered for 90 days — no quiz needed on repeat visits |
What Happens After Prediction
Once SmartSize has the body measurements (predicted, entered, or recalled), it compares them against the merchant's size chart to find the best-fitting size. This matching process is covered in the How SmartSize Recommends a Size article.
Why Absolute Precision Is Not Needed
You might wonder: can a short quiz really predict someone's waist to the centimeter? The honest answer is no — and that's perfectly fine. Here's why.
Size charts work in ranges, not exact numbers. A Medium might fit waists from 73 to 78 cm. The matching engine doesn't need to know that a customer's waist is exactly 75.2 cm — it just needs to know the waist is somewhere in that range.
The matching engine is designed for this. SmartSize doesn't do a simple "is this number inside the range?" check. It uses a scoring system that evaluates how well each measurement fits across all sizes, weighs measurements by priority, and accounts for garment fit style. A prediction that's off by 2–3 cm will almost always land on the same recommended size as the exact measurement would.
Think of it like a shop assistant. When a shop assistant looks at a customer and says "you look like a Medium," they're not measuring anything — they're making an educated guess based on overall build. SmartSize does the same thing, but with more data points and mathematical rigor. The result is the same: the right size, even without millimeter precision.
Where it matters most is between sizes. If a customer is right on the boundary between two sizes, a small prediction error could tip the recommendation one way or the other. But in these cases, both sizes would genuinely fit — which is why SmartSize also shows an alternative size suggestion when the fit is close.
In practice, the combination of prediction models + range-based matching + weighted scoring means that small measurement differences rarely change the final recommendation.
Why This Approach Works
Low friction — customers answer 4-5 simple questions, not 10+ with a measuring tape
Gender-specific models — separate models for men and women improve accuracy
Body shape context — shape selectors capture how weight is distributed, not just total weight
Self-improving — the models are periodically retrained on new data
For Merchants: What You Need to Know
You don't need to configure the prediction models — they work automatically.
The quiz questions are pre-configured based on your quiz's gender and product family.
Your role is to provide an accurate size chart with body measurement ranges — the prediction side is handled for you.
If your customers tend to know their measurements already, they can skip the quiz entirely and enter values directly.