A garment labelled a given size can differ by several inches between brands, and occasionally between items from the same brand. This is not a quality failure. There is no standard being violated.
Why there is no standard
Historical and commercial rather than technical.
Sizing systems originated from surveys of body measurements conducted decades ago, on populations that were not representative and are not current.
Those systems were voluntary and were adapted by manufacturers to suit their own customers.
Over time each brand developed its own block — the base pattern from which garments are cut — fitted to an idealised customer specific to that brand.
Which means a size is a statement about that brand's block rather than about a body, and there is no reason for two brands to agree.
Vanity sizing
The phenomenon that has made it worse over time.
Nominal sizes have drifted larger, so that a garment labelled a given number today corresponds to considerably larger measurements than the same number decades ago.
This is documented by comparing brand size charts over time, and the drift is substantial.
The commercial logic is straightforward. A buyer who fits into a smaller number feels better about the purchase, and brands compete on that.
The consequence is that the numbers have detached from measurement entirely and now convey almost no information.
The grading problem
A more technical issue that explains why fit degrades at the ends of a range.
Patterns are graded — scaled up and down from a base size — using rules about how much to add at each point.
Bodies do not scale proportionally. Proportions change with size, and a linear grade produces garments that fit reasonably near the base size and progressively worse away from it.
Which is why the extremes of a size range frequently fit poorly, and why brands with wide ranges that grade properly use multiple blocks rather than one.
Doing that costs more, which is why it is uncommon.
What actually helps when buying
Practical approaches, since the system will not improve.
Measure yourself once, properly, and write the numbers down. Bust, waist, hips, inside leg, shoulder width. This takes ten minutes and is the only fixed reference you will have.
Use brand size charts rather than the label, since most publish actual measurements and those are informative in a way the number is not.
Look for garment measurements rather than body measurements where they are given, since those tell you the actual dimensions of the item.
And note your size per brand, because it is stable within a brand even though it varies between them.
The returns consequence
The economic effect, which is substantial.
Return rates in online clothing are very high, and fit is the dominant reason.
Which is expensive for retailers, environmentally costly through transport and through returned goods that are not resold, and irritating for buyers.
Several approaches have been tried — fit prediction from previous purchases, body scanning, virtual fitting — with mixed results, and none has resolved it.
The obvious solution of standardised measurement labelling has been proposed repeatedly and has not been adopted, because no individual brand benefits from being the one that reveals its garments run small.
Alteration is the underused answer
The practical remedy that most people do not use.
Most garments can be adjusted at moderate cost, and the difference between something that approximately fits and something that fits is larger than the difference between price tiers.
The alterations with the best return are taking in the waist, shortening sleeves and hems, and adjusting straps.
Anything involving shoulders is difficult and expensive, which is why shoulder fit is the thing to prioritise when buying.
What would actually fix it
Mandatory measurement labelling, which exists in some technical clothing categories and works.
Alternatively, brands publishing garment measurements as standard, which some already do and which costs nothing.
Neither requires a common sizing system, which is the thing that is repeatedly proposed and repeatedly fails because it would require agreement among competitors with no incentive to agree.
Menswear is not exempt
Worth noting since it is frequently presented as the better system.
Trousers and shirts sold by measurement are more informative in principle, and the stated measurements still vary between brands because of how they are taken and because of finishing.
Jacket sizing is nominally a chest measurement and varies as much as anything else.
Which means the same advice applies — record your own measurements, use brand charts, and note your size per brand rather than expecting consistency.
The children's clothing exception
Worth noting because it demonstrates the alternative works.
Children's sizing is generally by age and height, and height is an actual measurement.
It is not perfectly consistent between brands and it is considerably more informative than adult sizing, because the underlying reference is a physical dimension rather than an arbitrary number.
Which is a reasonable demonstration that measurement-based labelling is workable, and that its absence in adult clothing is a choice rather than a technical limitation.