Testing & Evidence Library
See what was actually tested, on which sample or batch, by which method, when, and what the result does—and does not—prove.
The purpose is not to make every product look perfect. The purpose is to make important product facts traceable.
What Kind of Evidence Is This?
The label is part of the evidence. It tells the reader how the record was produced and what it cannot claim.
Physical Measurement
A real sample was measured under a defined method.
Documented Observation
Visual defect, smell, texture, adhesion, handling or another recorded observation.
Laboratory Test
A lab performed the test; exact lab/report/method scope still matters.
Traceable Record
A fact is supported by an identifiable document.
Supplier-Provided
Useful, but not automatically independently verified.
External Research
Background evidence, not a SanYuan sample result.
Simulation / Estimate
Useful for planning, but not a physical test or measured result.
No Qualifying Test
The correct state when a real test does not exist.
Find Evidence
Filter by product, evidence type, result status and date. Empty categories stay empty rather than receiving fake results.
Every Record Must Answer the Same Core Questions
A result without sample identity, method, conditions, limits and responsible people is not a useful public evidence record.
Evidence ID
Product, SKU/model, batch/lot and sample ID.
Why Test?
The defined question and test purpose.
How?
Standard/SOP, instrument and procedure.
Where / Under What Setup?
Environment and test configuration.
How Many?
Sample count and repeat structure.
Acceptance Rule
Predefined when a pass/fail claim is made.
What Happened?
Raw or summarized result with units.
What Does It Support?
Only the scope the test actually addressed.
What Did It Not Prove?
Uncertainty, missing dimensions and scope.
Who Did / Reviewed It?
Operator/lab, reviewer, version and date.
What We May Test or Document
Each category has its own meaningful evidence dimensions. A test in one category cannot be promoted into a different claim.
Identity & Material
Botanical identity, plant part, photos, supplier docs and advanced identification only when genuinely performed.
Defined Test Panel
A COA covers only the tests it actually contains; moisture/microbiology does not prove heavy metals, pesticides or allergens.
Food Evidence
Identity, caffeine, nutrition, microbiology, pH/Brix, preparation, packaging and sensory where tested.
Texture & Stability
Water activity, viscosity, Brix, seal and storage/stability—not tonic efficacy.
Heat & Adhesion
Temperature curve, warm-up, duration, peel/residue and wear observation—not pain relief or anti-inflammatory proof.
Formula & Use
Weight, identity, seal, steep appearance and sensory—not “detox” from water color.
Combustion & Handling
Material, burn, smoke, PM/VOC if measured, heat, ash, dimensions and extinguishing.
Technical Evidence
Wavelength, output, irradiance, geometry, electrical/laser/EMC/thermal evidence; P099 owns regulatory status.
Ten Questions Before a Result Goes Public
Traceability is the method discipline behind the library.
What?
Product/SKU/model/batch/sample
Why?
Defined question
How?
Method / standard / SOP
Conditions?
Environment / setup
How Many?
Sample size / replicates
Acceptance?
Predefined rule
What Happened?
Raw / summary result
Who Did It?
Operator / lab
Who Reviewed?
Reviewer
Limits?
Scope / uncertainty
Raw Data, Media, Failures & Corrections
Evidence credibility comes from preserving inconvenient information, not just attractive results.
Raw Data
Publish readings, timestamps, units and replicate labels where practical; otherwise explain restrictions.
Photos & Videos
Show real samples/setups. Crop, exposure and annotation may be used; defects, smoke or missing evidence must not be fabricated or removed.
Third-Party Labs
Record lab, report number, date, sample, method and accreditation scope only when verified.
Supplier Documents
Label them supplier-provided. Do not call them independent lab verification unless true.
Synthetic / Predicted
Must be visibly labeled “not a physical measurement” and cannot support “tested”, “measured” or “passed physical test.”
Negative / Inconclusive
Meaningful failures, unexpected and inconclusive results should remain versioned rather than being hidden.
Frequently Asked Questions
Short answers keep the library from becoming a blanket quality or safety certificate.
Does every SanYuan product have a COA?
No automatic claim.
Are all products lab tested?
Only say so for products/tests that meet the stated evidence definition.
What is first-party evidence?
Evidence generated or directly commissioned and traceably managed by SanYuan for a defined product/sample.
Is supplier data first-party?
No.
Is an AI simulation a test?
No.
Can synthetic data be useful?
Yes for planning, but it must be labeled predicted and cannot replace measurement.
Does a passing test mean the product is safe?
Only within the scope of the test; no universal safety conclusion.
Does one batch prove every batch?
No.
Are failed tests published?
Meaningful failures/inconclusive results should be retained and versioned rather than hidden.
Can I see raw data?
For public records, yes where practical; otherwise the record should explain restrictions.
Are test photos edited?
Only non-substantive presentation edits should be allowed; evidence content must not be altered.
Where are device approvals/licences?
Versioning & Public Dataset Policy
P100 can be dataset-like without pretending a collection of webpages is a formal public Dataset.
No Orphan Evidence
Every record needs Evidence ID, product/SKU/model, batch/sample, evidence status, method, date, result, limitations, owner, reviewer and version.
If invalidated, mark it Withdrawn / Superseded instead of silently overwriting.
Dataset Schema Only When Real
Use Dataset structured data only when SanYuan actually publishes a structured dataset with identifiable name, creator, variables/fields, scope, update date, access/download location and applicable terms/licence.
A page with several test results is not automatically a Dataset.