How the Integrity Index is calculated
Join Worthy Integrity Index 1.0 combines editors’ ratings of resolved evidence across seven lenses into one number from 0 to 100, and shows its working for every company.
What it measures
Documented conduct: what courts, regulators, independent assessors and our own checks have established about how a company behaves. It does not measure the good a company’s products do, and it does not describe any particular team.
Which evidence counts
| Kind of finding | In the score |
|---|---|
| Commitment | Not scored until independently verified; can raise the top of the range |
| Allegation | Not scored until resolved; can lower the bottom of the range |
| Proceeding | Not scored until resolved; can lower the bottom of the range |
| Established outcome | Scored, once editors rate it |
| Independent assessment | Scored, once editors rate it |
A matter is scored from its latest established outcome or independent assessment. A settlement or order counts as resolved; a decision under appeal is scored, and the top of the range shows the score if it were overturned.
How editors rate a matter
Editors rate each published matter once: adverse, favourable or context only, and how severe, with a published reason. A minor or moderate rating counts at once. A major or severe rating counts only after a second editor confirms it; today that is a different editor of the same company’s desk. Until a matter is rated it is listed as awaiting a rating and moves neither the score nor the range. Context-only matters, such as revenue mix or a lawful contract, are recorded but never scored. A rating stays with its matter: when the matter is corrected or withdrawn, the next calculation uses what is published then.
| Severity | Points |
|---|---|
| Minor | 3 |
| Moderate | 8 |
| Major | 15 |
| Severe | 25 |
| Age of the evidence | Counts for |
|---|---|
| Within 3 years | ×1 |
| 3 to 6 years old | ×0.6 |
| More than 6 years old | ×0.3 |
From matters to a lens score
A matter linked to several lenses is rated once, and its points are split equally between those lenses, so it counts once overall. Each lens adds its matters’ points and divides the total by the company’s size factor, rounded to one decimal, giving S, then:
- With nothing scorable, a reviewed lens scores 60.
- Favourable evidence: 60 + 40 × tanh(S ÷ 30), approaching 100.
- Adverse evidence: 60 − 60 × tanh(|S| ÷ 30), approaching 0.
So the first serious matter moves a lens a long way, and the twentieth moves it little. For a company of up to 1,000 people, one major adverse outcome from this year takes a lens from 60 to about 32.
Company size
A company with more people, products and customers has more chances to be investigated, sued or reported on, so it accumulates more documented matters even when it behaves no worse. Each lens’s summed points are therefore divided by a size factor based on employee headcount: 1 + 0.7 × log₁₀(headcount ÷ 1,000), rounded to one decimal and never less than 1.
| Employees | Points divided by |
|---|---|
| 1,000 or fewer | 1.0 |
| 3,000 | 1.3 |
| 10,000 | 1.7 |
| 30,000 | 2.0 |
| 100,000 | 2.4 |
| 300,000 | 2.7 |
The adjustment moderates how matters add up; it never makes one matter less severe. A $1 billion fine is still rated by what it established, and a severe harm at a large company still moves its score a long way. Headcount is the latest disclosed figure; where only a range is published, its midpoint, or the lower bound of an open range. Each company’s working shows the published size, its source and date. When no headcount is published, or none is recorded, points are not adjusted for size; that is not a claim of 1,000 people. It does not correct for scrutiny: a heavily watched company can still have more documented matters than an equally large, quieter one.
From lenses to the index
The published index is the plain average of reviewed lens scores: every lens counts equally. A company needs at least 5 of 7 lenses reviewed; until then only its lens scores are published. A lens is reviewed only when editors complete and date a coverage review of it, even if they found nothing to publish. Publishing a matter does not review a lens. Where two whole-number scores would tie, one decimal is shown, and tied companies are listed by name.
Readers can choose their own weights (ignore, half, standard or double each lens) and a minimum score in any lens. That view is calculated in the browser, labelled as the reader’s own, kept only in that browser and never published.
The range
The range is a sensitivity range, not a statistical confidence interval. Its bottom assumes every unresolved matter (one with no outcome or independent assessment yet) is established at the severity editors expect. Its top assumes unverified commitments are verified and appealed decisions overturned. Each bound starts from the same rounded lens total as the score. A later unresolved step inside a matter that is already scored does not move the range until editors rate it again. An unreviewed lens is assumed to fall between 40 and 80, for the range only; reviewed lenses run from 0 to about 80, so for thinly covered companies the range can understate the uncertainty.
The Pareto line and the shaded quadrant
On the pay chart, a company is on the Pareto line when no other company shown is at least as good on both index and pay and better on one. Companies with exactly the same index and pay share one mark and are on the line or below it together. The shaded quadrant covers companies at or above the median of those shown on both measures. Both use point values; ranges are drawn so you can judge how close a call is. With only one company there is no comparison, so no line or shading is drawn. Companies whose index is withheld are never plotted.
Known limitations
- The size adjustment uses headcount, not scrutiny or how much research a company has received; heavily watched companies may still carry more documented matters.
- Ratings are editorial judgements. They are published with a reason on every matter.
- Version 1.0 parameters are first settings, to be revisited with real data: the severity points, the age factors, the baseline of 60, the scale of 30, splitting multi-lens matters, equal default weights, the 5-of-7 coverage minimum, the 40–80 assumption for unreviewed lenses, the size factor, and the editorial review steps.
- Every company and figure in this local preview is fictional.
Versions
The method has a version number, and every score says which version produced it and the date it was calculated as of. Scores are recalculated on each request from what is published then, so a correction, withdrawal or new rating shows on the next page load; age is measured from each rated finding to that date. A change to weights, points or rules is a new version, announced with its effect on every company. Version 1.0, calculated as of .