| Document Title |
|---|
| Fashion sizing crisis explained: returns, vanity sizing, body scans, and data-driven design | |
|
|---|
| Sizing inconsistency drives massive returns. Tech can help at checkout, but the bigger fix is upstream: using returns and feedback to adjust patterns before production. | |
| Title Attribute |
|---|
| oEmbed (JSON) | |
| oEmbed (XML) | |
| JSON | |
| View all posts by Admin | |
| Liquid cooling is becoming the bottleneck tech for AI data centres | |
| Fire-blocking materials are being reinvented — because the old flame retardants were toxic | |
| Page Content |
|---|
| Fashion sizing crisis explained: returns, vanity sizing, body scans, and data-driven design | |
| Nature | |
| Climate | |
| Can technology fix fashion sizing? The real issue is incentives, not measurements | |
| / | |
| Technology | |
| / By | |
| Admin | |
| Sizing in fashion is broken in a very particular way: it’s not just that the labels are “wrong”, it’s that they’re | |
| inconsistent by design | |
| . A size 10 in one brand can map to a size 14 in another, and even within the same brand the fit can drift across seasons and factories. The result is a predictable consumer behaviour loop — order multiple sizes, keep one, return the rest — that turns returns into a business-as-usual logistics system. | |
| The interesting question isn’t whether technology can build a better virtual fitting room. It’s whether data and AI can push the industry toward a more honest, measurable definition of “fit” — while still allowing for style, preference, and the uncomfortable truth of “vanity sizing.” | |
| The returns engine: why the sizing problem became expensive | |
| The BBC puts a number on the consequence: fashion returns are estimated to cost retailers | |
| £190bn a year | |
| globally. | |
| Returns aren’t just a cost line. They are a structural force that shapes the industry: | |
| Inventory risk | |
| : a size run that doesn’t fit becomes dead stock | |
| Logistics complexity | |
| : reverse shipping, inspection, repackaging | |
| Margins | |
| : “free returns” are rarely free; they’re baked into prices | |
| Waste | |
| : damaged or unsellable returns can end up discarded | |
| So sizing has moved from “annoying customer experience issue” to something boards and sustainability teams pay attention to. | |
| Why a “size” is a bad measurement | |
| A label is a single number, but fit depends on a stack of variables: | |
| body measurements (multiple dimensions) | |
| body shape (distribution of measurements) | |
| fabric stretch and recovery | |
| pattern grading assumptions | |
| intended style (skinny vs relaxed) | |
| the wearer’s preference (“tight at the waist is fine” vs “never”) | |
| That’s why the UK Fashion and Textile Association’s Paul Alger (quoted by the BBC) is basically right: people aren’t mannequins, and fit is subjective. | |
| But “subjective” doesn’t mean “unsolvable.” It means the industry is trying to compress a multi-dimensional reality into an overly simple label. | |
| The industry’s dirty secret: brand sizing is a marketing choice | |
| One of the most important points in the BBC piece is | |
| vanity sizing | |
| (Alger calls it “emotional sizing”). | |
| Brands can (and do) decide to make their size labels more generous because: | |
| shoppers like the feeling of fitting into a smaller number | |
| it increases conversion and reduces friction | |
| Once a brand establishes its internal sizing norms, it often sticks with them season-to-season. So the inconsistency isn’t a bug. It’s part of brand identity. | |
| This is the part technology can’t “fix” unless the incentives change. | |
| Two places tech can intervene: checkout vs manufacturing | |
| The BBC describes a growing ecosystem of sizing tech. It helps to separate it into two intervention points: | |
| 1) Checkout-stage tools (help you choose a size) | |
| Examples mentioned include 3DLook, True Fit, EasySize and virtual try-on systems. | |
| The promise: | |
| reduce uncertainty online | |
| reduce “buy three sizes, return two” | |
| How they generally work: | |
| ask the shopper for measurements and/or smartphone photos | |
| map those signals to brand-specific fit data | |
| recommend a size for that specific garment | |
| The risk: | |
| privacy concerns (body scans/photos) | |
| calibration and bias (does it work equally well for different bodies?) | |
| confidence illusions (a recommendation is not a guarantee) | |
| 2) Manufacturing-stage tools (stop bad fit from being produced) | |
| This is where the BBC’s profile of Fit Collective is interesting, because it’s upstream. | |
| Instead of saying “help the shopper adapt to broken sizing,” the idea is: | |
| use return reasons, sales data, and customer feedback to adjust patterns and materials before production | |
| That is arguably closer to a real fix: fewer poorly-fitting garments exist in the first place. | |
| Fit Collective’s approach: turn returns into design feedback | |
| According to the BBC: | |
| Fit Collective analyses returns, sales figures and customer emails | |
| it produces advice for design and production teams | |
| example: reduce garment length by a few centimetres to cut return rates | |
| This is essentially an ML-assisted quality loop: | |
| customers signal pain (returns, complaints) | |
| the system aggregates signals across products and cohorts | |
| designers adjust patterns/materials | |
| fewer returns happen | |
| If you’ve ever worked in software, this is just product analytics. Fashion hasn’t done it at scale because the data is messy and siloed. | |
| The hard part: returns data is noisy and often dishonest | |
| There’s a reason returns data isn’t automatically a perfect truth source: | |
| shoppers choose “didn’t fit” when the real reason is “didn’t like it” | |
| some return reasons are constrained by UX options | |
| “fit” might mean length, waist, shoulders, or just vibe | |
| So for a system like this to work, it needs more than classification — it needs a model that can infer: | |
| which return reasons correlate with measurements and pattern issues | |
| which correlate with styling/presentation issues | |
| That’s doable, but it’s not a magic spreadsheet trick. | |
| What “good” sizing tech would actually output | |
| Most people imagine sizing tech as: “tell me if I’m a 10 or a 12.” | |
| The more useful output is something like: | |
| “This garment runs tight in the hips; if you size up the waist will be loose.” | |
| “The fabric has low stretch; if you prefer comfort, consider X.” | |
| “Length tends to be long for your height range.” | |
| Notice how these are not single-number answers. They are trade-offs. | |
| That’s why the best systems will likely feel less like a calculator and more like a | |
| fit explanation engine | |
| . | |
| Why virtual try-on is seductive (and where it breaks) | |
| Virtual fitting rooms are popular because they address a different problem: | |
| confidence about how something will | |
| look | |
| , not just fit | |
| But they have two limitations: | |
| visual realism is hard (lighting, drape, body movement) | |
| “looks good” and “feels good” diverge | |
| A realistic future is a hybrid: | |
| size prediction for fit | |
| try-on for style | |
| clear uncertainty indicators (“high confidence” vs “low confidence”) | |
| The incentive shift that could make this real | |
| Sophie De Salis at the British Retail Consortium (quoted by the BBC) frames sizing tech as a lever to reduce returns and support sustainability goals. | |
| That framing matters because it links fit to money: | |
| returns cost money | |
| returns create waste | |
| When a problem becomes a boardroom issue, it becomes budgeted. | |
| The best sign that sizing tech is becoming real is not more avatars — it’s retailers treating: | |
| fit analytics | |
| pattern adjustment | |
| measurement standardisation | |
| as core operations. | |
| Privacy: the quiet deal-breaker | |
| Any approach that relies on body scans or photos has to confront privacy head-on. | |
| If users feel: | |
| scanned | |
| profiled | |
| or that their images could leak | |
| they won’t opt in. | |
| So “privacy-respecting by design” should be a competitive advantage: | |
| minimal data retention | |
| on-device processing where possible | |
| clear opt-out and deletion | |
| transparency about what is stored | |
| The sizing industry has a chance to learn from ad tech’s mistakes. | |
| A practical take: what would reduce returns fastest? | |
| If you care about immediate impact, the likely high-ROI steps are boring: | |
| publish accurate | |
| garment measurements | |
| (not just size labels) | |
| standardise “fit notes” (“runs small”, “relaxed”, “high-stretch”) with consistent definitions | |
| improve product photos and drape information | |
| Then layer AI on top to: | |
| personalise recommendations | |
| close the manufacturing feedback loop | |
| AI helps most when the fundamentals are clean. | |
| Bottom line | |
| Technology can absolutely reduce sizing pain — but only if it targets the right layer. | |
| Checkout-stage sizing tools can lower uncertainty and returns, but they don’t change the underlying chaos. The more meaningful shift is upstream: using real fit signals (returns + feedback) to adjust patterns before production. | |
| And the final truth remains: if brands keep treating sizing as marketing (“emotional sizing”), inconsistency will persist. The winners will be the brands that combine honest fit communication with data-driven design — and make it feel like a service, not surveillance. | |
| Sources | |
| BBC News (Technology): | |
| https://www.bbc.com/news/articles/cjekg1pd9j4o?at_medium=RSS&at_campaign=rss | |
| ← | |
| Previous Post | |
| Next Post | |
| → | |
| oEmbed (JSON) | |
| oEmbed (XML) | |
| JSON | |
| View all posts by Admin | |
| Liquid cooling is becoming the bottleneck tech for AI data centres | |
| Fire-blocking materials are being reinvented — because the old flame retardants were toxic | |
| Sizing inconsistency drives massive returns. Tech can help at checkout, but the bigger fix is upstream: using returns and feedback to adjust patterns before production. | |
| |