If you’re working with anthropometric data — whether building a model from survey data, evaluating a third-party dataset, or assessing the provenance claims of a body measurement API — you’ll encounter ISO 15535. Understanding what it specifies, and what it doesn’t, helps you evaluate data quality claims more precisely.
What ISO 15535 is
ISO 15535:2012 (General requirements for establishing anthropometric databases) is an international standard that specifies:
- The minimum metadata required for an anthropometric database to be considered scientifically documented
- Requirements for how measurements should be defined and described
- Statistical reporting requirements for each measured dimension
- Requirements for describing the survey population (the “reference group”)
It does not specify which body dimensions must be measured. That’s ISO 7250-1 (Basic human body measurements for technological design). ISO 15535 governs how any dataset of body measurements — whatever dimensions were collected — should be documented and reported.
The relationship: ISO 7250-1 defines the measurements; ISO 15535 defines how to build and document a database of those measurements.
Why it exists
Before ISO 15535, anthropometric surveys were conducted with highly variable documentation practices. A survey might report means and standard deviations for each dimension, or might report only percentile tables, or might not specify which anatomical landmark was used for a given measurement at all.
This made it impossible to compare data across surveys. A “waist circumference” measurement from a 1988 US survey might use a different anatomical landmark than a 2005 Korean survey. Without documentation of the landmark, you couldn’t know whether the numbers were comparable.
ISO 15535 mandates standardized metadata so that any two ISO 15535-compliant datasets can be meaningfully compared.
What ISO 15535 requires
1. Population description
The standard requires documentation of:
- Geographic region and nationality of subjects
- Age range (and whether children vs. adults)
- Occupational or other selection criteria (e.g., military, civilian, student population)
- Exclusion criteria (health conditions, pregnancy, etc.)
- Sample size and sampling strategy (random, convenience, stratified)
- Time period of data collection
This is critical for evaluating applicability. A database collected from Korean military personnel in 1995 has known limitations for application to contemporary Korean civilians — but only if the metadata exists to reveal those limitations.
2. Measurement definitions
For each measured dimension:
- Reference to the ISO 7250-1 dimension code (if applicable)
- Anatomical landmarks used (with photographs or diagrams recommended)
- Measurement instrument and technique
- Body position during measurement (standing, seated, arm position, etc.)
This is what prevents the waist circumference problem described above. If both surveys document that waist circumference was measured at ISO 7250-1 dimension 4.3.12 (midpoint between lowest rib and iliac crest, mid-axillary line), the measurements are comparable.
3. Statistical reporting
The standard requires, at minimum, for each dimension:
- Sample size (n)
- Mean
- Standard deviation
- At least the 5th and 95th percentiles
Optionally but recommended:
- Full percentile table (P1, P5, P10, P25, P50, P75, P90, P95, P99)
- Skewness and kurtosis
- Regression equations relating dimensions to key predictors (height, weight, age)
4. Quality control documentation
The standard requires documentation of:
- Inter-observer reliability (how consistent were measurements across different technicians?)
- Training and certification requirements for measurers
- Measurement error assessment
Checking ISO 15535 compliance in practice
When evaluating a dataset or a vendor claiming their model is “trained on validated anthropometric data,” ISO 15535 provides a checklist:
ISO_15535_CHECKLIST = {
"population": [
"geographic_region_documented",
"age_range_documented",
"sex_stratification_documented",
"sample_size_reported",
"sampling_strategy_described",
"exclusion_criteria_stated",
"collection_period_stated"
],
"measurements": [
"iso_7250_1_codes_referenced_where_applicable",
"anatomical_landmarks_described",
"measurement_technique_documented",
"body_position_specified",
"instrument_type_recorded"
],
"statistics": [
"mean_reported",
"standard_deviation_reported",
"p5_and_p95_reported",
"sample_size_per_dimension_reported"
],
"quality": [
"inter_observer_reliability_assessed",
"measurement_error_quantified"
]
}
def assess_dataset_compliance(metadata: dict) -> dict:
"""
Assess a dataset's ISO 15535 compliance given its metadata.
Returns a compliance report with missing elements.
"""
missing = []
for category, requirements in ISO_15535_CHECKLIST.items():
for req in requirements:
if not metadata.get(category, {}).get(req, False):
missing.append(f"{category}.{req}")
total_checks = sum(len(v) for v in ISO_15535_CHECKLIST.values())
passed = total_checks - len(missing)
return {
"total_checks": total_checks,
"passed": passed,
"failed": len(missing),
"compliance_pct": round(passed / total_checks * 100, 1),
"missing_elements": missing
}
How the major public datasets score
ANSUR II (2012): Extensively documented. Population described in detail (US Army personnel, age 17–58, 93 dimensions each with landmark photographs). ISO 7250-1 codes referenced. Percentile tables published. Inter-observer reliability assessed. Effectively ISO 15535 compliant.
NHANES: Strong on population documentation (nationally representative US civilian sample, age range, race/ethnicity stratification). Measurement protocols published and standardized. Percentile tables available. Limited on the full set of ISO dimension codes — NHANES measures a smaller dimension set (primarily height, weight, waist, hip circumference) rather than the full 130+ ISO 7250-1 range.
SIZE UK (2004): ISO 15535 compliant, published through SizeUK consortium. 3D body scanning protocol documented. 130 dimensions. Population: UK civilians, 11,000 subjects. Limitation: 20+ years old.
Informal or proprietary datasets: Often fail on population documentation (vague geographic description, undefined exclusion criteria) and quality control (no inter-observer reliability data). Compliance claims are often unverifiable.
What this means for body measurement APIs
When a body measurement API claims its predictions are “validated” or “based on scientific anthropometric data,” ISO 15535 gives you a framework for asking the right questions:
“Which datasets were used?” Named, citable datasets (ANSUR II, NHANES, SIZE KOREA) are publicly documented. You can read the original publications and assess the population yourself.
“How old is the data?” Body proportions shift over time. ISO 15535-compliant datasets document the collection period, allowing you to assess temporal validity.
“Which populations are covered?” If the datasets used are all US or European, accuracy for East Asian, South Asian, or African users is unvalidated — regardless of what accuracy claims are made.
“What were the measurement landmarks?” Without ISO 7250-1 references, you can’t know whether the “waist circumference” in the training data uses the same landmark as the API’s output.
A provider who can answer these questions specifically — citing ISO 15535-compliant datasets by name and version — is making verifiable, assessable claims. A provider who gives vague answers about “proprietary validated data” is not.
ISO 15535 and reproducibility
For applications where predictions may be cited in technical documentation — ergonomic certifications, product safety compliance, medical screening protocols — the scientific provenance of the underlying data matters beyond just accuracy. A prediction from a model trained on ISO 15535-documented, publicly citable datasets is reproducible: someone else can examine the training data, replicate the model, and verify the accuracy claims.
A prediction from a black-box model with undocumented training data cannot be verified, replicated, or cited in scientific or regulatory contexts.
ISO 15535 is technical infrastructure — not glamorous, not visible to users, but foundational to the scientific credibility of any body measurement system. For developers building applications where accuracy claims matter, understanding this standard gives you the vocabulary to evaluate the underlying data quality of the tools you integrate.