Accuracy and honesty
The useful answer includes what we do not know.
The desktop web app is available. Its camera-based scan produces estimates when a supported live scan provider is configured. Quality can change by measurement zone, device, clothing, pose, light, and capture conditions. Those limits stay attached to the result.
Provenance explorer
One shoulder zone. Four honest states.
Select the evidence state. The zone stays fixed so you can see what changes in the value, label, explanation, and next action.
Right shoulder46.2 cm
Observed ยท Strong
Visible shoulder landmarks support this zone more directly.
What changedDirect landmarks are available, so a value can be shown with stronger support.
Concept illustration. Fictional measurement and evidence states.
Capture playground
What would make this estimate stronger?
Try the three capture conditions. This is a simplified teaching aid, not a live scan or a prediction of product performance.
Improve all three conditions before relying on this teaching example.
Fictional confidence concept. Not a measured accuracy score.
It is designed to
- Estimate the measurement zones it supports, when capture conditions allow
- Show provenance, confidence, and scan date on every available value
- Compare your measurements with reviewed garment data and show the evidence used
- Decline to answer when the evidence cannot support a responsible recommendation
It will not
- Replace a tailor's tape or an in-person fitting
- Promise how a garment will feel after movement, wear, or washing
- Invent garment data a brand has not published
- Hide the effect of capture conditions or provider limits
How each value is described
- Estimated
- Calculated from the scan's geometry and the provider's current method. It is not a direct tape measurement.
- Observed
- Supported more directly by landmarks visible in the capture. Observed does not mean error-free.
- Inferred
- Derived using more assumptions or indirect evidence. Inferred values may be marked limited or unavailable.
- Unavailable
- The source evidence is not strong enough to present a value. Retrying under better conditions may help.
Confidence is part of the measurement
A measurement is shown with its source type, confidence, interval where available, provider and scan version, and capture date. A low-confidence value is not silently promoted into a strong size recommendation. If key garment or body zones are missing, the result explains the gap or declines to recommend.
Why there is no published accuracy number
A meaningful accuracy claim requires a defined reference protocol, diverse sample, supported device and condition matrix, repeatability testing, zone-level error distributions, confidence calibration, subgroup analysis, acceptance thresholds, and formal claim review. Until that substantiation program produces suitable evidence, no accuracy number or comparative claim is published.
Words we don't use
The following phrases are listed here only to make the boundary public. They are not approved marketing claims.
- "100% accurate"
- "exact"
- "tailor-grade"
- "fit-guaranteed"
- "digital twin"
- "knows your body"
- "perfect fit"
- "Meet your body"
- "Your body, decoded"
- "Powered by intelligence"
- Any numeric accuracy claim before substantiation and approval
When a result changes
Provider updates, a later scan, changed capture conditions, or updated garment data can change a result. Versions stay visible so a new estimate does not quietly overwrite the record that produced an earlier recommendation.