Fashion Sizing Failure: The Definitive Structural Diagnosis

Fashion keeps diagnosing sizing as a chart problem. A 38 in one system becomes a 40 somewhere else; a medium at one brand fits like a large at another. That diagnosis is familiar and partly true. But it is not the deepest one. The root cause is not that numbers disagree. It is that the knowledge built up during every fitting, alteration, and return is discarded the moment a transaction ends.

Why Does Fashion Sizing Keep Failing Despite Decades of Standardisation?

Sizing standards have existed since the 1950s. ISO 8559-1 defines global garment construction norms. The EU has pushed harmonised sizing tables since the 1990s. Yet online fashion returns still run at roughly 30% — with fit cited as the primary driver in most major studies. The reason standardisation has made so little difference is that it solves the wrong problem.

A universal size chart improves the starting assumption for an anonymous transaction. It gives a new customer a slightly better prior before she clicks 'add to cart'. What it does not do is carry forward anything that was already learned. It does not remember that a given person needs more room through the shoulder, prefers a specific trouser break, or consistently alters ready-to-wear at the hem. A better chart is still a blank slate.

Research published on ScienceDirect on digital product fitting shows that even sophisticated virtual try-on tools reduce returns by roughly 28% when they are built on accurate individual body data — but performance drops sharply when that data cannot be reused across sessions. The return reduction belongs to the data, not to the fitting interface itself.

What Standardisation Cannot Solve: The Three Structural Gaps

Standardisation addresses one variable — the relationship between a label and a body measurement range. There are at least three structural gaps it cannot touch. Each one means that fit knowledge resets, and the wearer pays the cost of that reset in the form of a bad purchase or a return.

• Prior garment outcomes are forgotten. When a customer returns a jacket at retailer A because the sleeve pitch was wrong, retailer B has no access to that diagnosis. The mistake will be repeated.
• Historical fit judgment cannot move between operators. A tailor who has fitted a client for three years holds that knowledge in a private system or in memory. It cannot be transferred to a made-to-measure brand the client tries next.
• Prediction quality resets whenever the operator changes. Machine learning models for size recommendation are only as good as the longitudinal data behind them. A new brand starts with a cold-start problem regardless of how good its chart is.

How Fragmented Fit Data Drives the €800 Billion Return Problem

The economics of fit failure are large enough to restructure supply chains. European fashion returns cost the industry an estimated €800 billion in reverse logistics, reprocessing, and lost inventory value annually, according to Statista and McKinsey State of Fashion data. Approximately half of those returns are fit-related. The cost is not born by a single bad chart. It accumulates through millions of repeated cold-start problems — each one a transaction where no prior fit knowledge was available.

In practice, when measuring the cost per return for mid-market European brands, the figure runs between €15 and €25 per item after factoring in shipping both ways, restocking labour, and markdown risk on returned stock. A brand processing 100,000 online orders per year at a 25% return rate, with 60% of those fit-related, is absorbing roughly €2.25 million annually in avoidable fit-failure cost. That is not a chart problem. It is an infrastructure problem.

The goal is not to make every brand use the same numbers. The goal is to let what is already known about a person remain usable when the commercial context changes. — Caprice Bespoke, fit infrastructure design principle

What Would a Portable Fit-Data Layer Actually Require?

A portable fit-data layer must do something fundamentally different from a size chart. It must persist across commercial relationships, be controlled by the wearer rather than any single operator, and degrade gracefully when partial data is all that is available. The architecture has to solve four problems simultaneously.

• Wearer ownership. Fit records must belong to the person, not to the brand that collected them. Without ownership, portability is a feature the operator can revoke.
• Operator-neutral encoding. Measurements and garment outcomes must be stored in a format that any downstream system can interpret — not in a proprietary schema that requires a specific vendor's API.
• Provenance and context. A measurement taken by a trained tailor carries different weight than a self-reported chest measurement. A portable layer needs to encode who measured, when, under what protocol, and with what confidence.
• Outcome linkage. Body measurements alone are insufficient. The layer must also carry which garments fit, which did not, and what alterations were made — because that is where the predictive value lives.

The GS1 Digital Link standard and the EU's Ecodesign for Sustainable Products Regulation (2024/1781) are both moving in this direction at the product level — creating persistent, machine-readable identifiers for garments. The next step is the same capability at the wearer level: a persistent, portable, wearer-controlled record that travels across every commercial relationship.

Why the EU Regulatory Environment Is Accelerating the Shift

The EU Ecodesign for Sustainable Products Regulation, which entered into force in 2024, mandates Digital Product Passports for textiles by 2027. That requirement creates a machine-readable garment record at scale for the first time. It is primarily a sustainability instrument — tracking material composition, repairability, and end-of-life data. But the infrastructure it creates is directly reusable for fit purposes.

When every garment has a persistent GS1 Digital Link identifier encoding its construction measurements, a portable fit record can link to actual garment data rather than relying on size labels. A wearer who knows that EU garment DPP reference 4902-xxxx fits correctly in the shoulders but not through the waist has created a fit outcome record that any operator — tailor, brand, platform — could read if given permission. The regulatory push toward product-side data portability makes the parallel wearer-side infrastructure significantly easier to build.

The Structural Diagnosis: Infrastructure, Not Charts

Sizing failure is an infrastructure problem with three layers. The first is the chart layer — label inconsistency between brands. Standardisation partially addresses this. The second is the session layer — fit knowledge collected during one transaction that is not retained for the next. No current standard addresses this. The third is the portability layer — the technical and legal mechanism by which wearer-controlled fit data moves between operators. This layer does not yet exist at scale in fashion.

The solution therefore has to reach below sizing language and into data architecture. Fit improves when continuity compounds: when each fitting, alteration, and return enriches a persistent record that the wearer can bring to the next relationship. It does not improve when every operator maintains a slightly better isolated chart. The compounding effect of accumulated fit knowledge — measured in reduced cold-starts, fewer misfits, and lower return rates — is the economic case for building the portability layer.

• Layer 1 (Chart) — Standardisation of label-to-measurement mapping. Partially solved by ISO 8559-1 and EU harmonisation efforts.
• Layer 2 (Session) — Retention of fit outcomes within a single operator relationship. Partially solved by CRM and recommendation systems, but siloed.
• Layer 3 (Portability) — Wearer-controlled transfer of fit data across operators. Not yet solved at scale. This is the structural gap.

Frequently Asked Questions

Why do size charts differ so much between fashion brands?

Size chart variation exists because fashion brands developed their fits independently, often targeting different body proportions for their core demographic. An Italian luxury brand patterns for a narrower shoulder than a Scandinavian outdoor brand. Chart variation is real, but it is only the surface symptom. The deeper issue is that even a perfectly standardised chart cannot carry forward the fit knowledge that has already been built up about a specific wearer.

Does virtual try-on technology solve the sizing problem?

Virtual try-on reduces returns when it is built on accurate individual body data. Research on digital product fitting shows a return reduction of roughly 28% in controlled conditions. But those gains depend on the accuracy and recency of the body data used. When a customer must re-enter measurements every session, or when the try-on tool uses generic avatar proxies, the reduction is much smaller. The technology amplifies good data infrastructure; it does not replace it.

What is measurement portability and why does it matter?

Measurement portability is the wearer's right and technical ability to carry their fit data — measurements, garment outcomes, alteration history — from one operator to another. It matters because most fit knowledge is currently locked inside individual brand or platform databases. When a wearer changes brand, that accumulated context is abandoned, and the prediction quality for the new brand resets to zero. Portability would allow that knowledge to compound rather than reset.

What role does the EU Digital Product Passport play in sizing?

The EU's Ecodesign for Sustainable Products Regulation mandates Digital Product Passports for textiles by 2027. These create machine-readable records at the garment level, encoding construction dimensions in a standardised, portable format. While primarily a sustainability instrument, the same infrastructure can link garment-side construction data to wearer-side fit outcomes — making it possible to build a portable fit record that references actual garment data rather than label approximations.

How much do fashion returns actually cost the industry?

European fashion returns run at approximately 30% of online sales volume. The cost per returned item for mid-market brands is typically between €15 and €25 after accounting for reverse logistics, restocking labour, and markdown risk. McKinsey State of Fashion data and Statista industry figures consistently show fit failure as the primary driver of returns — ahead of colour mismatch or product quality concerns. The aggregate cost to the European fashion industry runs into hundreds of billions of euros annually.