How we score.

We score sustainability claims across six weighted areas, using three tiers of evidence. A claim with nothing behind it does not raise the score. This page shows how it works and what sits behind the number on your record.

A note on GreenSpecs, our shelf scanner: it does not run this rubric. A camera pointed at a label can only read the label, so it scores what is printed there — ingredients, packaging, carbon, and how much the brand discloses. Supply-chain depth and circularity are not on a package and cannot be scanned. The two numbers answer different questions, and the distance between them is roughly the work a company has done that never reaches the shelf.

SCORING ENGINE v1.0 · UPDATED APRIL 2026

Two platforms, two different readings

They answer different questions and they are not the same score. GreenSpecs reads a label through a camera, so it can only score what is printed on the package. YKO assembles the record behind the package — filings, certification registries, supplier disclosures — and scores that. A label cannot show tier-two suppliers or a take-back program, so those dimensions exist on this side and not on that one.

GreenSpecs.app

Point a camera at any packaging. our vision model reads every visible sustainability claim, checks it against known certification records, and marks unsupported claims as it goes.

The score comes from what the label actually claims, not from a marketing page. A product that claims nothing scores neutrally, not negatively. We do not penalize a quiet label; we flag claims that cannot be backed up.

YKO

Brand-level scores draw from regulatory filings, third-party certification records, published sustainability reports, supplier disclosures, and GreenSpecs scan data as a live consumer demand signal.

Every data point carries a source tier, so the score reflects the quality of the evidence, not just the presence of a claim. A strong claim on weak evidence lands below a modest claim on strong evidence.

How much a claim counts depends on its evidence

A “carbon neutral” line on a homepage is not the same as a third-party verified emissions report. We tag every data point with its source tier, and the tier sets the confidence level on the score.

Self-reportedLOW

From the brand’s own site, marketing, or packaging, with no independent check. We include it because transparency matters even here, but it carries low weight in confidence, and we flag it where it conflicts with third-party evidence.

Document-backedMEDIUM

Supported by published documents — CDP disclosures, SBTi target letters, supplier codes of conduct, reports with specific metrics. These claims age: a 2019 commitment with no update is read at lower confidence.

Third-party verifiedFULL

Validated by an independent certifier (B Corp, USDA Organic, Science Based Targets initiative, Rainforest Alliance, Fair Trade) or confirmed by an NGO audit, regulatory filing, or peer-reviewed supply-chain mapping. Only this tier counts as verified, and we assess each certifier on its own merits, because not all certs are equal.

A 0–100 score across six weighted areas

We score each dimension 0–10, then combine them by the weights below into the 0–100 YKO total. We always show the six sub-scores beside the total, because the total alone hides where a brand is strong and where the gaps are. The weights sum to 100.

Certifications20

Which recognized bodies have independently assessed this brand or product? We weight certifications by rigor, auditor independence, enforcement mechanism, and geographic scope. Holding a cert is not automatic credit — what it covers, and what it omits, both count.

Raises: B Corp, Science Based Targets, USDA Organic, Fair Trade, Rainforest Alliance. Context-only: 1% for the Planet, Non-GMO Project.

Supply Chain20

How much does the brand actually know about — and disclose about — its supply chain? Named suppliers, published Tier 1 and Tier 2 lists, third-party audits, and supplier codes of conduct all raise the score.

Raises: public Tier 1 list, published third-party audit, supplier code of conduct, partial Tier 2 disclosure. Lowers: an active NGO controversy. Counts for nothing: “we work with ethical suppliers.”

Carbon20

Are emissions disclosed — including Scope 3, not only Scope 1 and 2? Is there a decarbonization commitment with a published roadmap, or only a vague net-zero line? SBTi-approved targets sit above pledges with no pathway.

Raises: SBTi target approved, Scope 3 disclosed, net zero with a roadmap, carbon-neutral claim independently verified. Lowers: carbon-neutral asserted without methodology; emissions rising year over year with no corrective plan.

Packaging15

What share of packaging uses recycled content? Is it curbside recyclable, certified compostable, or part of a refill program? Is there a reduction target? This dimension is visible and verifiable from the label — a core GreenSpecs scan signal.

Raises: high recycled content, curbside recyclable, certified compostable, refillable program, a stated reduction target. Lowers: single-use plastic with no reduction program.

Narrative Integrity15

Are the claims backed by evidence, or just marketing language? This dimension starts at a neutral midpoint and moves either way. Verified, specific claims raise it; a high ratio of marketing language to data lowers it. We reward brands that name their own gaps.

Raises: verified specific claims, transparency about unsolved gaps. Lowers: a high unverified-claim count; a greenwashing finding from a credible source.

Circularity10

Is there a take-back or repair program? Is the product designed to last or to be replaced? Durable construction, lifetime warranties, take-back schemes, and design-for-disassembly all contribute. Most relevant to apparel, electronics, and hard goods; food carries the same weight but typically scores lower across the category.

Raises: take-back or repair program, lifetime warranty, designed for disassembly, biodegradable end-of-life, durable materials.

Commodity adjustment

For high-risk commodity ingredients — coffee, cocoa, palm oil, soy, oat milk, olive oil, and others — we raise the Certifications and Supply Chain weights relative to Packaging and Circularity. In those categories, sourcing certifications and named-origin traceability are the real differentiators; packaging claims are near-universal and carry less signal. The 0–100 scale and every tier threshold stay the same.

How a camera scan produces a score

GreenSpecs does not depend on a product database or barcode lookup. It reads what is on the label, so it works on any product in any market — including store-brand and private-label goods that appear in no existing database.

  1. Image capture. The camera points at the packaging; the app grabs a still at enough resolution to read label text and iconography.
  2. AI vision extraction — our vision model. the vision model reads the label and pulls every sustainability claim: certifications (logos and text), environmental claims (“biodegradable,” “carbon neutral,” “100% recycled”), sourcing claims (“organic,” named farms or regions), and packaging statements. Each is tagged by type and confidence.
  3. Claim verification lookup. Recognized certifications are checked against the YKO certification database. A Rainforest Alliance logo validates as a real cert; a bare “sustainably sourced” line is tagged self-reported. The brand is matched to the YKO entity to pull any existing brand-level score.
  4. Six-area scoring. Extracted and verified claims map to the six areas using the same rubric as brand-level scores, weighted by what the label claims. A missing dimension scores neutrally — an absent claim is not penalized unless the category baseline makes the omission notable.
  5. Scan signal recorded. The event is logged anonymously — no identity, no location beyond a regional approximation. This is the live shelf-demand signal that feeds brand dashboards: which products get scanned, in which markets, and how their claims score at the moment of decision.
  6. Result displayed. Score (0–100), tier, confidence, and a breakdown of what the label claims versus what was verifiable. Flags are shown in plain English, not jargon — a decision-useful answer in under 10 seconds.

Every score carries a confidence level

A 74 backed by three verified certifications and a published Scope 3 report means something different from a 74 drawn from marketing copy and a single self-reported line. Confidence sits beside every score, because a score without it is incomplete information.

High confidence

Built mostly on third-party verified data — multiple certifications confirmed, supply-chain disclosures published, emissions independently validated. Further research is unlikely to move it much.

Medium confidence

A mix of document-backed and self-reported data. A reasonable read of current evidence that could shift meaningfully once more third-party data lands.

Low confidence

Mostly self-reported or inferred from thin public information. The best read of what is available, but directional rather than settled. A low-confidence mark is not an accusation — it is a data-quality signal.

Five fixed tiers, 0 to 100

Thresholds are fixed, not relative — a brand scores well because its practices and disclosures meet the criteria, not because it beats its competitors. The labels are plain and non-moralistic. “Needs Work” means the data is sparse or the practices are early; it is not a verdict on intent.

Benchmark Leader90–100

Comprehensive third-party verification across dimensions — Scope 3 published, multiple rigorous certifications held, full supply-chain transparency, no active NGO controversy. No company on the records we have read reaches it yet.

Committed75–89

Strong practices with substantive evidence — at least one rigorous certification, published emissions, meaningful supply-chain disclosure. Gaps exist but are acknowledged. Patagonia (85.5) and Guayakí (79.0) sit here.

Progressing60–74

Real activity with meaningful gaps in disclosure or verification — some third-party validation, incomplete. Cotopaxi (66.3) and Clover Sonoma (63.0) are current examples, often brands mid-improvement.

Early Stage40–59

Limited activity or sparse documentation — some claims, few verified. Organic Valley (51.5) holds this tier despite meaningful practices, largely on limited public disclosure.

Needs Work0–39

Little verifiable practice, or a high ratio of unverified claims to evidence. Primal Kitchen (22.5) and Bachan’s (21.5) are current examples. A low score reflects what is publicly verifiable, not necessarily what exists.

What triggers a flag

Flags are specific findings we show beside a score. Most are already reflected in the dimension scoring, so they are not extra penalties — they are signals that warrant a closer read. Each is plain English, and each is explained.

carbon neutral — unverified
The brand claims neutrality but has published no methodology and had no third party confirm it. The claim may be accurate — this flag asks for the evidence.

emissions rising
Year-over-year emissions increased with no disclosure of why and no corrective roadmap.

greenwashing flag
A credible third party — NGO, regulator, or independent research body — has published findings that a specific claim misleads. The highest-weight flag; it applies a direct deduction.

carbon not disclosed
No emissions data is publicly available — not even Scope 1 and 2. Above a certain revenue line, that is a significant disclosure gap.

single-use plastic — no program
Primary packaging is single-use plastic with no stated reduction target, transition program, or alternative-material commitment.

active NGO controversy
Within the past two years a credible NGO has linked the brand to supply-chain labor or environmental violations that remain unaddressed.

What does not improve a score

What we leave out matters as much as what we measure. The following are common in brand communications but carry no weight here, because they carry no verifiable information.

What money does to a score: nothing

We sell records, supplier data collection and reports. We do not sell score movement, and the way that is enforced is worth stating precisely rather than promising in general.

Evidence moves a score. Payment never does.

A document-backed certification counts the same whether the company sent it to us or we found it in a public registry ourselves. There is no plan check, no account check and no payment check anywhere in the code that reads evidence — the endpoint that serves it to our scanner is public and unauthenticated, so anyone can call it and see exactly what we see. A company that has never paid us and never will is read identically to one that does.

Uploading a document cannot make it verified.

Documents a company sends us reach document-backed, and no further. The verified tier requires an independent certifier, an audit or a regulatory filing, and it is set by hand. No automated step, and no amount of paperwork from the company itself, promotes a claim into it.

Every lifted fact shows its tier and its date.

Where GreenSpecs shows evidence from a company’s record, each line carries what tier it sits at and when it was dated.

A fact with no visible reason for appearing cannot be checked against rule 1. So the provenance travels with the fact, rather than being available on request.

One thing this does not yet do: evidence from the record is shown beside a shelf scan, but it does not change the scan score. When that changes, the scan will say what it was read from — label only, or label plus record — because a number that quietly widened its inputs with nothing on screen saying so is exactly the kind of claim we exist to catch. A dimension we could not read is never scored zero: unread is not absent, and a small label is not a small effort.

There is no pass mark

A score out of 100 invites a question it cannot answer: is 50 good? On its own it is neither good nor bad, and the tier word beside it is the same problem wearing a label. A score is a position among the records we can check — nothing more, and we would rather say so than let the number imply a verdict.

So every score now carries its reference: the range and the middle of its category, and where this company sits in it. We only compare within a category once we hold at least 5 scored companies in it; below that the comparison widens to every company we have read, and the page says so. A median of two companies is not a median, and presenting one as a benchmark would be inventing precision we do not have.

The reference improves as the set grows, which means an early score is a rougher reading than a later one. That is a fact about our coverage, not about the company being read.

A modest claim with strong evidence scores higher than a bold claim with weak evidence. We reward honesty about the gaps, not marketing volume.

When emissions data is not public

Most brands under $1B in revenue have not published Scope 3. For them, we estimate the likely Scope 3 intensity of the product category from industry-average emission factors. We always label the estimate “estimated — industry average” and never present it as brand-specific.

Factors come from published lifecycle-assessment databases for material categories including conventional beef, organic and conventional dairy, cotton and recycled-polyester apparel, glass and aluminum packaging, yerba maté, coffee, and soy. The full factor table is versioned alongside the engine.

A brand that publishes its own Scope 3 — even where that number runs above the industry average — scores higher on this dimension than one that discloses nothing. We reward transparency regardless of the figure.

SOURCES — ecoinvent · USDA LCA Digital Commons · peer-reviewed lifecycle literature

How the rubric evolves

Dimensions and weights are versioned. We record any change here and apply it retroactively to every scored brand, so scores stay comparable over time. We never alter the logic silently.

Initial engine deployedAPR 2026

v1.0 — six areas, commodity weight modifier, origin-specificity tie-break, three-tier source architecture. Baseline brands scored: Patagonia, Guayakí, Cotopaxi, Clover Sonoma, Straus Family Creamery, Organic Valley, Mate Factor, Eco Teas, Primal Kitchen, Bachan’s.

Certification intelligencePLANNED

A rigor score (1–5), auditor-independence score (1–5), and enforcement score (1–5) added to each certifier — adjusting how much weight a certification carries within the Certifications dimension.

Travel Conservation dimension setPLANNED

A parallel rubric for hotels, tour operators, and airlines. Mapped to the same six categories but reweighted for the sector — energy, water, biodiversity, and community economic impact take the place of packaging and circularity as primary dimensions.

Correcting a score

Questions about a specific score or source reach us at scores@yko.earth. We explain any score in detail and correct errors when we find them — accurate data over defensible data, every time.