The arithmetic.
Every number published on a ranking page is produced by the operations on this page. Each measure is scored on a fixed scale, combined by a stated formula, and rounded by a stated rule. The worked examples use entries you can open and check.
01Scales and rounding
One scale governs everything. Every measure and every composite is expressed on a 0 to 100 scale.
| Quantity | Range | Precision |
|---|---|---|
| Individual measure | 0 to 100 | One decimal place |
| Composite score | 0 to 100 | Whole number, or one decimal where the pool requires it to separate ranks |
| Confidence | 0 to 100 | Whole number |
| Source reliability | 1 to 5 | Integer |
Rounding is half-up and applied once, at the end. Measures are rounded for display only. The composite is calculated from unrounded measures, so a composite recomputed from the displayed one-decimal values can differ from the published composite by up to 0.5 without either being wrong.
02Evidence qualification
An evidence item enters scoring only if it passes all three tests. Items that fail are excluded and logged with the reason.
Independence is tested before recurrence is counted. Where two citations trace to a single origin, such as an article quoting another article, they count once. This means recurrence counts unique origins, not unique URLs.
Two categories are excluded before any test runs: material published by the subject about itself, and material that was paid for. A dimension with no qualifying evidence is marked INSUFFICIENT, scored 0, and still carries its full weight in the composite.
03Scoring bands
Each measure is scored against fixed evidence bands. The bands hold the scale steady across editions and across researchers.
| Band | Independent origins | Pattern across the period |
|---|---|---|
| 90 to 98 | 6 or more | Recurring across the full period, including at least one source outside the subject's home market |
| 75 to 89 | 4 to 5 | Recurring and consistent across the period |
| 60 to 74 | 2 to 3 | Recurrence confirmed |
| 40 to 59 | 2 | Limited recurrence, or strong evidence confined to a short part of the period |
| Below 40 | 2 | Minimum qualifying evidence only |
| 0 | 0 to 1 | INSUFFICIENT |
04Composite formulas
People and companies use different formulas because the evidence behind them is different.
People
Five measures, weighted, normalized to 100.
| Measure | Weight |
|---|---|
| Trust | 0.25 |
| Reach | 0.20 |
| Practice | 0.20 |
| Influence | 0.20 |
| Results | 0.15 |
| Sum | 1.00 |
Worked example, Leading Global 100 Professionals, rank 2:
Companies
Five measures, equal weight. The composite is their arithmetic mean.
Worked example, Leading Global 100 Companies, rank 1:
These formulas describe the global editions. The Leading Global 100 Professionals and the Leading Global 100 Companies are computed as stated above. The country editions were drawn on separate dates under earlier scoring generations, and their composites are not reproducible from the measures displayed beside them. They are being re-derived onto these formulas.
05Normalization
Measures are normalized within the pool being ranked, not against an external scale.
Two constraints apply after normalization.
- No value is clipped. If a transformation produces a value above 100, the transformation is wrong and is re-derived rather than truncated.
- The top of the pool lands below the ceiling on its own. A pool whose leaders all sit at exactly 100.0 has been compressed, and compression removes the separation the ranking exists to show.
Because normalization is pool-relative, a measure of 88 in one edition and 88 in another are two different quantities. Section 08 states what this permits and forbids.
06Confidence
Confidence is calculated separately from the composite and never enters it.
A public record is never complete, so confidence is not reported at the top of the scale. Observed ranges in the current editions:
| Edition | Minimum | Maximum |
|---|---|---|
| Leading Global 100 Professionals | 31 | 75 |
| Leading 25 Professionals, by country | 57 | 89 |
| Company editions | Not currently calculated | |
Confidence qualifies a reading. It does not reorder one. Two entries with the same composite and different confidence hold the same rank.
07Measures outside the composite
The Leading Global 100 Companies displays four quantities that are not inputs to the composite. They are shown as context and carry a weight of zero.
| Measure | Reads | In composite |
|---|---|---|
| Change | Year over year movement in standing | No |
| Belief | Whether belief is moving up, holding, or slipping | No |
| Exposure | How directly daily interactions connect to revenue and reputation | No |
| Risk | How far a strong or weak experience travels publicly | No |
Removing any of these from the calculation leaves the composite unchanged. Adding any of them to it would change every rank on the page.
08Comparability
Because measures are normalized within a pool and each edition is drawn on its own date, scores carry meaning only inside their own edition.
| Comparison | Valid |
|---|---|
| Two entries in the same edition | Yes |
| One entry across consecutive editions of the same program | Yes |
| An entry in one program against an entry in another | No |
| An edition against an edition of a different program | No |
| Two non-consecutive editions of the same program | No |
A composite of 88 on a company ranking and a composite of 88 on a people ranking are not the same quantity, are not produced by the same formula, and do not describe the same thing.
09Recomputation
Every published composite is recomputed from its measures before an edition is released. An entry that does not reconcile is re-derived from its evidence, never adjusted to fit.
The evidence log for each edition is archived against its publication date, so any challenge to a score can be answered from the record that produced it rather than from the page that displays it.
