Can technology fix fashion sizing? The real issue is incentives, not measurements

/ Technology / By

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:

  1. customers signal pain (returns, complaints)
  2. the system aggregates signals across products and cohorts
  3. designers adjust patterns/materials
  4. 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:

  1. visual realism is hard (lighting, drape, body movement)
  2. “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

n English