40% of online apparel returns cite fit as the reason. Standard sizing grids are statistical tools used as personal recommendations. Built from population measurements, often surveys or anthropometric studies from decades ago, they set sizes as ranges around mean values for selected body dimensions. They suit shoppers near the middle of those ranges, but perform poorly when even one dimension differs from that center.
The issue is not poor grid design. Human proportions do not follow one axis. Dividing a population into ten buckets using one measurement cannot accurately describe every body's three-dimensional geometry. Add a second dimension and the grid weakens further, since chest-to-waist and shoulder-to-hip relationships vary. People can differ from the mean on one dimension independently of another.
The problem of proportions across dimensions
Take a grid based on chest circumference: S is 34-36 inches, M is 36-38 inches, and so on. As a one-dimension classification, it works. Add shoulder width. Someone with a 37-inch chest may average 16.5 inches across the shoulders. The grid is built around that expected proportion.
A shopper with a 37-inch chest and 17.5-inch shoulders, one inch broader than average for that chest, shares a bucket with someone measuring 16.5 inches at the shoulders. The garment has a 16.5-inch shoulder seam. On the broader body, it sits inside the shoulder joint, pulls at the armhole, and misdirects the sleeve. The grid selected the correct size, yet the garment misses the shopper's shoulder geometry.
This is common. Shoulder-to-chest proportion varies substantially, as do waist-to-hip ratio, torso length relative to height, and thigh circumference relative to hip. When any ratio differs from the mean, assigning a size by the dominant dimension can create poor fit elsewhere.
Averaging compounds across dimensions
A one-dimension deviation creates a problem at that point. Deviations across two or three dimensions can make every chosen size problematic. Sizing up for the larger dimension leaves excess fabric where the body is closer to average, so the garment fits none of those dimensions well.
Return data shows this among shoppers who repeatedly return items across brands and categories. They did not necessarily choose the wrong grid size. The grid lacks a bucket for their actual proportions. For them, a more accurate bucket recommendation changes little, since the issue is the bucket system's inability to describe their geometry.
We are not saying most shoppers have this issue. A grid works adequately for bodies near the population mean across most dimensions. Still, more shoppers differ meaningfully on at least one relevant dimension than brands may infer from returns. Aggregate rates conceal this by combining shoppers who fit the grid and never send items back.
How 3D body measurement exposes grid limits
A 3D body model built from two photos captures several dimensions at once, including ones standard grids do not classify. Shoulder width, shoulder slope, torso length, and waist-to-hip ratio can be measured separately, then compared with the garment's construction at each matching point.
The result is not a bucket. It compares the shopper's geometry with the garment's construction at every fit-relevant dimension. Someone with broader shoulders and an average chest can receive a recommendation that identifies shoulders as the key factor in a structured garment, then suggests sizing up for shoulder room instead of using chest alone.
This does not remove the grading issue. If a larger size improves shoulder fit but adds excess chest fabric, the body-geometry recommendation should state both effects. If no available size suits the shopper's proportions, the honest answer is that none will fit well. We say so. Sending someone toward a size that still fits poorly is not a good recommendation. It is a return in 10 days.
What sizing grids still provide
Sizing grids should not be discarded. They remain useful as a starting point and shared language. When a brand says, "this garment runs small, we recommend sizing up," it communicates how grid assignment relates to fit, which helps. When a shopper says they are "normally a size 10," that vocabulary supports a customer service conversation.
The problem comes when the grid is treated as enough for fit advice. A shopper who is "normally a medium" knows something about their body, but not how those proportions match a garment's construction. Fit intelligence fills that gap by adding the dimensional detail the grid lacks, without replacing the grid convention.
How proportion affects return patterns
Viewing returns through proportion deviation changes the picture. At Thinx3D, we see the highest rates not necessarily at extreme sizes, where fit problems might be expected, but among shoppers with proportion mismatches: average-size bodies that differ on one or two dimensions from the grid's expected relationships.
Purchase history alone rarely identifies these shoppers, since repeat returns can resemble preference changes or category exploration. Body measurements reveal the pattern: a shopper's waist-to-hip ratio exceeds the average for their chest, and each woven trouser fits the waist but binds at the hip. The pattern is systematic. The answer is not another grid size, but advice that treats hip circumference separately from waist.
