MIHEN's Transparent Supplement Comparison Method

Define the method's narrow purpose

MIHEN's comparison method organizes information that can be checked directly: package identity, label declarations, document availability, named certification records, seller terms, and observed prices. It does not decide what a person should use, interpret symptoms, select an amount or schedule, or predict outcomes. Establishing that boundary is part of the method, not a disclaimer added afterward. Topics that require clinical judgment remain outside the publication workflow unless an appropriately qualified reviewer is formally responsible for that specific content.

Each comparison begins with a written question and inclusion rule. The question identifies the product class, market, evidence date, and fields that will be compared. The inclusion rule is applied before results are known. MIHEN does not add a product merely because it has an attractive price or remove one because information is incomplete. Incomplete records can appear with a visible status or in an excluded appendix, preserving the fact that the product was considered and showing why a full comparison was not possible.

Separate evidence layers instead of blending them

The first layer is verbatim package evidence: product identifiers, serving fields, ingredient wording, other ingredients, warnings, storage instructions, and manufacturer contact details. The second layer is external documentation, such as an official regulator record, certificate issuer directory, or batch document. The third layer is transaction evidence from a named seller on a stated date. Editorial text forms a fourth layer and explains how to read the fields. The interface should label these layers so an explanation is never mistaken for a label claim.

Source authority depends on the question. A regulator is the source for the meaning and status of its own record. A certification body is the source for its current directory. The package is the source for what that package declares. A retailer is the source for its current offer and policies. No single source is authoritative for every field. Where sources conflict, MIHEN records the conflict, checks package identifiers and dates, and avoids selecting a convenient value merely to complete a row.

Use explicit states for verification and missing data

A binary badge can overstate certainty, so MIHEN uses descriptive states. Examples include label image matched to package identifier, current issuer-directory listing found on a date, manufacturer document available but batch match not established, seller claim not independently located, source conflict, not applicable, and unknown. Every state has a definition in the methodology. Reviewers record the search path and source URL, allowing another person to repeat the check instead of trusting an unexplained checkmark.

Unknown is not a failing grade and does not become a numerical zero. A missing certificate, for example, can mean that none exists, that it was not published, that a directory uses another company name, or that research was incomplete. MIHEN reports only what the evidence supports. Calculated fields remain unavailable when a required input is missing or incompatible. This rule prevents averages, guessed conversions, and silent defaults from creating false precision or systematically favoring products with less transparent data.

Calculate price fields without converting them into rankings

Price observations record exact package, seller, currency, purchase condition, fees included, and date. Cost per labeled serving can be calculated when the panel supplies a serving count. Cost per declared ingredient unit is calculated only when compared labels use a confirmed common ingredient basis and compatible measurement unit. The formula and inputs remain visible. Subscription discounts, coupons, shipping thresholds, and tax treatment are separate fields rather than hidden adjustments.

MIHEN does not label the lowest price as the highest-quality choice. A price calculation describes a transaction under stated conditions. Likewise, a larger declared quantity does not receive an automatic advantage. Tables can be sorted by transparent factual columns, but default ordering is editorially neutral, such as alphabetical or by documentation completeness with a written rule. If a summary mentions a price difference, it repeats the observation date and the exact comparison basis so the sentence cannot drift into a general market claim.

Version, review, and correct every published record

Every guide and comparison carries a first-publication date, a material-update date, a source-check date, and a scheduled review date. A sitewide deployment date is not substituted for page review. Material changes include a new label, changed package size, altered document status, revised methodology, or corrected source interpretation. Price observations can expire sooner than stable label explanations. The record's status determines whether it appears in the sitemap; drafts, unreviewed translations, and pages with critical validation failures remain unpublished or noindex.

Corrections accept direct evidence through a published contact route. The editor verifies the package and source before changing a record, notes the material correction, and retains enough history to explain the change. Automated checks test JSON structure, unique identifiers, source URLs, calculation inputs, canonical links, language alternates, and internal links. Human review checks meaning, scope, source support, and plain language. Neither layer claims to eliminate error. The method earns trust by making errors discoverable, corrections traceable, and unsupported conclusions structurally difficult to publish.

• Publish the question and inclusion rule for each comparison.
• Keep package, document, transaction, and editorial evidence separate.
• Define every verification and missing-data state.
• Show calculation inputs, formulas, conditions, and dates.
• Avoid general rankings and unsupported equivalence claims.
• Assign real publication, source-check, update, and review dates per page.
• Keep drafts and unreviewed translations out of the index.
• Maintain an evidence-based correction log.