The Susty Score · Methodology v1
The Susty Score, measured honestly.
How fragmented public disclosure becomes one comparable 0–100 rating — what we measure, how we source it, how we grade real performance over claims, and why every penalty traces back to evidence.
Weights: People 40 · Planet 40 · Practices 20 — no deduction without a dated, documented record.
Principles
What every score is held to
The Susty Score answers one consumer question — “is this company worth my purchase?” To answer it credibly, every rating obeys four commitments. They are the standard we audit against, and the standard we invite anyone to hold us to.
Evidence sets the limits
A penalty is only as good as the record behind it.
Nothing is marked down on a rumor, a headline, or someone’s characterization of a company. A deduction requires a discrete, dated, documented event — a regulator’s order, a court record, a primary filing. Where the record on a topic is thin, we say so, rather than inventing a finding in either direction.
Performance over disclosure
We grade the answer, not the act of answering.
A company earns nothing for simply publishing a report. We assess how good the disclosed conduct actually is — the quality of the target, the credibility of the plan, the direction of the trend — against an industry-specific baseline.
Conduct over claims
An independent layer reads the record.
Layer0, our controversy screen, evaluates documented conduct — fines, findings, violations — independent of what a company says about itself. A polished report cannot outrun a regulator’s ruling.
Recency-weighted, never erased
Today’s behavior counts most; history stays visible.
Recent conduct best predicts current behavior, so it weighs most. A serious historical event fades in both the points it costs and the ceiling it sets, and old events stop compounding — but a finding is never erased from the record behind the score.
The score
One score, three pillars
Each pillar is graded against industry-specific baselines — a solar developer is never measured against a bank — so the number reflects how a company performs relative to what its sector makes possible.
People
Social impact & labor
What we grade
- Worker rights & pay equity
- Supply-chain labor standards
- Community investment
- Diversity, equity & inclusion
- Human-rights compliance
Planet
Environmental footprint
What we grade
- Emissions & net-zero credibility
- Water & resource intensity
- Waste & circular economy
- Biodiversity & ecosystem impact
- Renewable energy transition
Practices
Governance & ethics
What we grade
- Board independence & oversight
- Executive comp alignment
- Transparency & disclosure quality
- Corruption & bribery controls
- Tax integrity & lobbying
A weighting is a statement of materiality, not popularity. People and Planet carry the heaviest weight because they are where a company's real-world impact concentrates — and where claims most often outrun conduct. These weights are current and applied uniformly across industries; as the evidence base grows they shift toward industry-calibrated models, so Planet counts for more at an oil and gas company than at a professional-services firm.
How a score is built
From a thousand pages to one number
An AI pipeline reads dense, fragmented disclosure at scale and structures it into comparable inputs — but structuring is not scoring. Each input is graded for quality, screened for conduct, and published with the source attached.
Assemble
The public record on a company — filings, regulator and court records, audited disclosures, dated third-party reporting — is organized into comparable inputs, one for each question in that industry’s question set.
Grade
Each input is scored against an industry baseline for how good the conduct is. A weak answer scores low even when fully disclosed.
Screen
Layer0 applies sourced deductions and pillar caps for documented conduct — independent of the company’s own narrative.
Verify
Identity matching is biased toward precision. A sparse record is marked as limited data, and below an evidence floor we publish no rating at all; company-provided data is marked as such.
Publish
Every published score shows how it is composed and names the specific findings pushing it up or down, in plain language — not a number on its own. A link from each finding to the document behind it is the next thing we are building; until it ships, we don’t claim it.
Defensible — drives a deduction
“Fined $X by [agency] on [date], per [docket].”
A verifiable event with a citable primary record.
Excluded — never moves a score
“The company misled the public.”
A characterization stated as fact, with no primary source.
We grade the food, not the menu.
A thick report full of weak commitments scores below a lean one full of strong ones.
Performance, not disclosure
Two pledges, thirty points apart
The common failure of a disclosure score is to reward the act of reporting: points for publishing a number, regardless of whether the number is good. Susty asks a different set of questions.
A disclosure-only score asks
- ? Did they publish a net-zero pledge?
- ? Is there an emissions figure on the page?
- ? Does a policy document exist?
- ? Length and detail of the report.
The Susty Score asks
- → Is the target absolute or only intensity-based? Near-term or distant?
- → Is it Scope 3-inclusive, externally validated, and on track vs. trajectory?
- → Is capital allocation aligned with the stated plan?
- → How does the outcome compare to the sector baseline?
Worked example — two net-zero pledges, same disclosure, different scores
Vague long-dated pledge
“Net-zero by 2050.” No interim targets, no Scope 3, no validation, no capex link. Fully disclosed — and graded weak.
Credible near-term plan
SBTi-validated near-term target, Scope 3 included, interim milestones, capex aligned, trajectory on track to date.
No published plan
No disclosure on the topic. Scored on absence of evidence — and flagged as thin evidence, not penalized as proven misconduct.
Two companies can disclose the very same thing and land thirty points apart. That gap is the whole point — it is where “did they report?” becomes “is it any good?”
Layer0 · Controversy screening
Where conduct overrides claims
Layer0 applies sourced deductions and pillar caps for documented controversies — independent of a company's own narrative. Every entry is split into three parts, and only the sourced fact drives a deduction.
What, who, when, how much — a discrete, verifiable event. E.g. a $Xm penalty assessed by a named regulator on a specific date.
Drives the deductionThe citable record — a court docket, regulatory order, or primary filing. It is required before a finding can deduct anything.
No source → no deductionOptional context only. Any characterization of meaning or intent is stored separately and clearly labeled.
Never deducts on its ownTier 1 — Primary
Adjudicated
Court rulings, regulatory penalties, official findings. Required for the most serious categories.
Full weightTier 2 — Corroborated
Documented
Multiple independent, dated, named reports of an event not yet adjudicated.
Scaled by confidenceTier 3 — Unverified
Held
Single-source or low-confidence signals. Routed to review — surfaced only once corroborated.
No deductionThe heaviest categories — fraud, criminal conduct, forced labor, fatalities — require a citable primary source from a court or regulator; absent that, the trigger routes to analyst review rather than auto-deducting.
Time, decay & memory
Recent conduct weighs the most
Weighting recent conduct more heavily is standard practice in credit and risk scoring: recent behavior is the strongest predictor of current behavior. So controversy weight decays with age — on a published schedule, with firm limits.
Age of event → relative weight applied
What the schedule does not soften
- Repeat conduct — another violation in the same area on the record (environmental, labor, financial misconduct…) holds an older event at 75% instead of 50%.
- Active consent decrees — full weight for as long as the order is in force, whenever it was entered.
- Undated events — the deduction is charged in full until the date is established. We forgive nothing we cannot date.
Where age does bring relief
- Pillar caps — a controversy also sets a ceiling on the pillar, and that ceiling fades on the same schedule as the points. A recent event caps hard; a 15-year-old one caps lightly, so a company that has since reformed is no longer held down by it for ever.
- Old findings stop compounding — however many a company has, events older than five years take at most 15 points off a pillar between them. One past mistake counts; a long-ago record cannot multiply into a near-zero score. Recent findings are the exception — they are charged in full and they do stack.
The floor
Age alone never takes a deduction below half weight — no matter how old the event, and whatever the category.
Time can soften a deduction; it cannot erase it.
And the finding stays on the record regardless of its points. Even after a deduction has decayed, the event is still what the score was built from — a discount is defensible only because the thing being discounted is still there. Time can change what a finding costs; it never deletes it.
Confidence & coverage
When evidence is thin, we say so
No scoring system has perfect information, and pretending otherwise is its own form of greenwashing. Susty separates a low score driven by poor performance from a low score driven by thin disclosure — and tells the user which is which.
The limited-data flag
Limited dataWhen the underlying record is sparse, the score carries a limited-data marker — it is never silently inflated. This protects the company that simply discloses little from being mistaken for one with a documented bad record. And below an evidence floor we publish no rating at all: a number built on too little says more about what we failed to find than about the company.
A quiet company is not a guilty one.
Coverage & refresh
A company becomes eligible for reassessment 90 days after its last one, and a material event can pull that forward. Scores move when the evidence moves — not on a calendar. Identity matching is biased toward precision: when we cannot confidently attribute an event to the right entity, we hold it rather than risk penalizing the wrong company.
companies rated and published so far, in 40+ countries
Before a company is eligible to be re-scored, plus news-triggered re-screening on material events
Identity-matching bias — an event is held rather than mis-attributed
Governance & contestation
Owned, auditable, and challengeable
A credible methodology has an owner, a paper trail, and a door a company can knock on. Editorial independence is what makes the rating worth trusting — and a clear contestation path is what keeps that independence honest.
A named owner
The framework, decay schedule, and evidence tiers each have a single accountable owner. Material changes are dated and attributed — never made silently.
Published & versioned
The methodology is public and carries a version number. Every revision is logged, so a score can always be read against the rules that produced it.
A path to challenge
Any rated company may dispute a factual input. A challenge is reviewed against the same evidence bar that created the entry; an unsupported deduction is corrected or removed.
Roadmap: sector benchmarks
A planned set of high-materiality, sector-banded benchmark questions per industry — deterministic bands that push the score further toward “is it actually good?”
Independence is not the absence of accountability — it is accountability to the evidence rather than to the rated.
Sustainability, transparent by default
No deduction without a record. Every score explainable. Performance graded over disclosure, conduct over claims — and nothing serious ever quietly disappears.