This is an operations autopsy, not a brochure. Method v2 is not "better testing." The category-of-one is a checkable published number: a government floor or a disclosed read-across, certified at 100 percent of that number.
What Method v1 was
Method v1 was rigorous in intent. It tried to set every subcategory and analyte cell from Heavy Metal Index occurrence literature: pool the findings for that product form, pick a percentile, publish a number. The government-limit lookup that Method v2 uses as the default did not govern those cells.
v1 also carried a two-class analyte status ladder and an overage metric, MaxPoL, with status bands above the published limit. A product could sit over a published number and still hold a transitional status. That architecture is retired. It is named here so a reader can see what was killed, not so it can be used.
Why it was unacceptable
The occurrence pool was too thin. Many subcategory and analyte cells had missing or single-survey data. The generator still had to emit a number. Calculation mistakes, thin pools, and confidence scores that could not carry a live certification decision produced limits that moved when the corpus or the code moved.
That volatility is unacceptable for published limits and for a mark a retailer or a regulator will treat as a fact. A certification program needs numbers a hostile reader can re-derive from a named instrument, not from a pool that changes under them.
Reproducibility failed the legal-defensibility test in the Governing Principles. Every published number has to trace to a named sovereign maximum or a disclosed read-across. Incomplete claim-surface cells are not cert-ready. Speed suffered for the same reason: the program could not ship a cell it could not defend.
What Method v2 changed
Method v2.0 was ratified on 2026-08-03 as a government-limit default. Later locks on the same edition gutted the remaining brittle pieces: claim cells, status, and mint gates.
| Axis | Method v1 | Method v2 |
|---|---|---|
| Limits | Occurrence pool of Index findings for every subcategory and analyte | Government floor or disclosed read-across only. Blank, monitored, or subsumed cells are HOLD, not claim cells |
| Status | MaxPoL overage ladder and status bands above the published limit | 100% of the published limit. Four statuses: Certified, Remediation, Suspension, Revocation. No transitional overage |
| Classes | A two-class analyte status ladder, including a 150% transitional overage on the second class | Every claim analyte is equal at the gate. The labels Tier 1 and Tier 2 no longer drive certification status |
| Testing | Panel plus pathway work, entangled with status bands | The same eight-total panel. Speciation is a pathway reflex. Species never replace the parent |
| Provenance | Index-derived occurrence as the limit | A named sovereign instrument. The Index is cited one way, outbound only. Occurrence carries no compliance weight |
House PDFs are react-pdf only. Mint gates refuse an incomplete claim surface. This page does not mint a DOI and does not publish a new claim limit.
What stayed
v2 gutted the load-bearing brittle pieces. It did not throw away the program around them.
How a limit is looked up, how a lot is tested, and what happens when a lot fails still live on program structure. Occurrence still characterises the market on the two pathways. It does not invent a claim cell.
How to cite
Method v2.0 is current. Method v1 is retired and did not certify a brand. Cite v2. This page is the version-change record. It has no DOI. Do not invent one.
The Method v2 governing instrument for Category 1 is Manual v3, house DOI 10.83138/hmtc.manual.infant-and-child-foods. The live subcategory pages remain the derivation of each cell.
Pendergrass, K. (2026). Method. Institute of Contaminant Standards. https://heavymetalcertified.com/program/method
© 2026 Institute of Contaminant Standards / Heavy Metal Certified. All rights reserved.