The pressure is real
US retailers processed $849.9 billion of returns in 2025 at a 15.8% return rate, per NRF and Happy Returns. The 2024 figure was 16.9% and $890 billion.
That is a board-level number, and it is the main reason measurement technology is finally appearing in product roadmaps.
Where AI measurement is actually being used
- Fit prediction at checkout, using purchase history and brand grading.
- Virtual try-on, translating body data into a visual preview.
- Size recommendation on product pages.
- Supply chain, using aggregate measurement data to inform grading.
The honest limitation
Most of these run on historical purchase data keyed on size labels, not on measured bodies. That reproduces historical sizing behaviour, including its errors.
If a brand has graded inconsistently for years, training on those labels teaches the model the inconsistency.
What genuinely works
Publishing finished garment measurements. A brand knows its garment chest to the millimetre. That single disclosure removes more uncertainty than any recommendation engine, and it costs nothing.
Research on over 1,000 garments across 50 UK brands found only 64% matched their own chart within 1 cm, which shows how much is being withheld.
Grading and supply chain
Aggregate measurement data is genuinely valuable here. If a company understands the real distribution of its customers' shapes, it can grade blocks better than inherited charts allow.
The last comprehensive US anthropometric survey for apparel was conducted in 1941. Grading built on that data does not describe modern bodies.
What retailers should measure
Returns are usually described by reason. Those reasons are not published in aggregate. A retailer that could break returns down by cause, by size and by category, would know precisely where the problem sits.
Coresight Research, surveying 100 US apparel decision-makers in 2023, attributed 53% of returns to fit. McKinsey's figure reached 70% when poor fit and style were combined.
Note the NRF publishes total return rates and fraud share but not a fit-versus-other-reason breakdown. Most industry commentary about fit shares comes from vendor surveys.
The honest scorecard
Measurement technology is helping most where the problem is already well understood, which is standard sizing in mainstream apparel. It helps least where the problem is garment construction or inaccurate grading.
And it has a ceiling: it measures the customer, not the product.
Further reading in why adoption has been slow and the returns economics.