A well-fitted image is not evidence that the garment fits

A base-size sample demonstrates that one realised garment can satisfy the agreed conditions. It does not demonstrate that the rules used to create every other size preserve fit, construction or appearance. Grading changes related dimensions according to a rule set; it is not a uniform enlargement or reduction of the whole pattern. The technician must inspect how those rules affect each point of measure and how connected features continue to work together.

The smallest and largest planned sizes test the boundaries of the grading rules. An intermediate size can reveal a discontinuity or an implementation error that endpoints alone would not explain. For the illustrative shirt, the reviewer should check chest and shoulder dimensions, armhole and sleeve relationships, neck and collar closure, pocket and placket placement, and hem balance. The exact list follows the design, while the principle is consistent: measure the released artefact, not just the base-size intention.

Nested patterns are useful for visual review, but a technician also needs the numerical point-of-measure table and seam checks. A notch that remains geometrically present can move to an unhelpful position. A sleeve can retain its stated length while its cap no longer matches the armhole as intended. A pocket can remain attached to the front panel yet become poorly placed relative to the closure or wearer. These are proposed inspection categories, not reported defect frequencies.

Tolerance belongs to this review, but it must be agreed for the product and point of measure. NIST's framework explicitly leaves risk tolerance to the organisation and its context rather than prescribing a universal threshold.[S19] GFT therefore recommends that the maker and buyer approve dimensional tolerances before seeing pilot outcomes. A generated table should never fill an unknown tolerance with a plausible industry-sounding number.

CLO's documentation lists Standard DXF, AAMA and ASTM options, a selectable size group, units, grading export choices, notch controls and cutting-line or sewing-line handling.[S22] Those controls make exchange settings inspectable. They do not prove that a particular receiving system will reconstruct every size, annotation and relationship correctly. The maker must test the actual source and receiving releases rather than treating a format label as an acceptance result.

Exhibit 29 sets out a proposed review grid. The size labels are deliberately generic because the teaching example has no authorised product range. Each row links the fit evidence to the export evidence, so a physical fit pass cannot conceal a broken receiving file and a successful file opening cannot conceal a failed garment.

The readback in Exhibit 29 should compare measurements in the receiving workflow with the authorised source table. It should also verify that the selected export contains all intended sizes, that units are unchanged, that notches and seam allowances retain their meaning, and that size labels map to the correct geometry.[S22] A technician should record pass, failure or an explained exception against the pattern revision. Bulk release then uses the readback-approved file, not an untracked attachment sent alongside it.

A neutral standing pose can show balance and silhouette, but it cannot represent every demand placed on a working garment. EASE studies pose adaptation that redistributes local allowance to relieve strain in high-stretch regions, while retaining physics-based simulation for final visualisation.[S27] This makes pose an explicit design and evaluation variable. It does not establish guaranteed comfort, safety or freedom of movement for a physical wearer.

For an illustrative uniform-shirt review, the product team might agree reaching, sitting and fastening checks with the technician and wearer panel. These are proposed teaching activities, not a universal test suite. Each activity needs a defined starting position, movement, observation point and acceptance condition. A reviewer should know whether the concern is hem lift, closure gaping, sleeve restriction, pocket access or another product-specific outcome before a result is classified.

The digital protocol should retain the same activity identifiers as the physical protocol. If the avatar performs a reach that the physical panel cannot reproduce, the two outcomes are not directly comparable. If the physical wearer uses a different movement because the digital brief was ambiguous, the deviation belongs in the record. The purpose is not theatrical realism. It is to ask the same fit question at both boundaries.

Simulation can help inspect predicted contact, deformation and local strain under declared material and pose assumptions.[S27] A wearer can report pressure, restriction, thermal sensation or preference. Those observations may disagree. A garment can show no visible digital contact problem yet feel restrictive, or show an unattractive fold that does not impede the task. The release record should preserve both forms of evidence rather than collapsing them into one score.

Material identity is equally important. A pose adaptation computed with one material preset should not be transferred silently to a different physical cloth. The material chapter's measured-swatch rule therefore remains active during movement review. A changed fabric or finish can require a new digital run and a repeated physical activity check even when the pattern coordinates are unchanged.

Fit work can require photographs, body scans, dimensions, avatars and subjective wearer comments. These records concern identifiable people and may be more intrusive than the design files that a generative system normally handles. NIST's Generative AI Profile identifies privacy risks involving sensitive information, unauthorised use, disclosure and inference, and recommends lifecycle actions rather than a single privacy switch.[S20] A fit workflow therefore needs a purpose and permission record before collection begins.

GFT proposes a data inventory that separates the fit task from optional secondary uses. The core record should contain only the measurements and activity observations needed for the defined product review. A body photograph should not become the default when a tape measurement and posture note will answer the question. Where a scan or image is necessary, its identifier, purpose, access group and retention rule should travel with it.

Consent for a fit session is not consent to train a model, publish an image or reuse an avatar for another product. Those destinations need separate, explicit permission decisions. A withdrawn or expired permission should be visible to the system and to the responsible coordinator, with affected derivatives identified for review. This is a governance design, not a conclusion about the legal effect of any consent form in Cameroon, Hong Kong SAR or another jurisdiction.

The measurement protocol should also minimise unnecessary identification. The pattern technician can work with an approved body representation and dimensions without receiving a wearer's name, contact details or full image set. The panel coordinator can hold the participation and permission mapping under a separate access rule. Role separation reduces exposure while preserving the link needed to investigate a measurement or approval dispute.

Before body data enters an external service, the maker should record the provider, model or service version, processing location disclosed by the provider, retention setting, training setting and export or deletion route. Unknown terms remain unresolved procurement items. The NIST profile's attention to third-party components and lifecycle risk supports checking the whole service chain rather than only the visible application.[S20]

Retention should follow the declared decision. A long-lived approved block may need a dimensional validation summary, but that does not automatically justify retaining every source photograph or raw scan. The team can preserve the approved measurements, protocol, reviewer and result while deleting source material under the agreed rule. Any exception should have an owner and a reason rather than becoming permanent storage by default.

Access reviews should include exported avatars and test datasets, not only the original database. A synthetic-looking avatar can remain traceable to a person if its identifiers or source files reveal the connection. Training or benchmark use requires a fresh destination check, and public presentation requires a separate publication decision. This chapter does not provide legal advice; applicable privacy, employment and buyer obligations require jurisdiction-specific review before implementation.