Quoting HumanToll
A figure from HumanToll is an estimate with a range, attributed under stated assumptions. Quoted that way, it is accurate. Quoted any other way, it is not.
Quote every figure with its interval, at two significant figures, and say that it is attributed to the supply chain of a product, not caused by it. Say that it counts the production phase only and that it is a floor: harm the model cannot attribute is not counted. Name the scenario and the run. HumanToll names no company, brand or person, and neither should a sentence that quotes it.
Where it stands
No figure is published yet. The method, the sources and the harm ledger are public; the figures are withheld until a validated run is published. A number attributed to HumanToll today did not come from it.
Say this, not that
The phrasings HumanToll itself uses, and the ones that would misstate it.
| Say | Not |
|---|---|
| “attributed to the supply chain of” | “caused by” |
| “an estimated 0.4–3.5 deaths per million units” | “1.2 deaths per million units” — a central value without its range |
| “under the economic-embedded scenario” | a figure with no scenario named |
| “production phase only; excludes harm from using the product” | silence about the phase |
| “this is a floor: harm the model cannot attribute is not counted” | “the total human cost” |
| “the model estimates” | “the truth”, “the real cost” |
Six readings that are wrong
Does HumanToll say a product kills people?
No. It attributes, statistically and under stated assumptions, a share of deaths and injuries counted in conflicts, workplaces and deprivation to the supply chains of products. It is not a causal claim about any death or injury, and no figure says that buying a product caused one.
Is this the total human cost of a product?
No, and by construction. Every figure is a floor: harm the model cannot attribute to a supply chain is not counted, the material route counts only the raw materials a bill of materials traces, and every figure is the production phase only — harm from using a product is not in it.
Is product A deadlier than product B?
Only where HumanToll says so. Two products are compared on draws taken together, and one is called larger only when the whole interval of their ratio lies on one side of one. Otherwise the model cannot tell them apart, and the comparison page says exactly that. Compare per unit only for like units; per kilogram, the page says which.
Which companies are responsible?
HumanToll names none, and a figure from it is not a statement about any company, brand or person. The unit of analysis is an archetype — “a midsize electric car” — and its supply chain.
Can I round the figure, or quote the central value?
Two significant figures is the most the model has, and the interval is part of the finding: quote it with the figure, every time. A figure graded INDICATIVE has no central value at all — only its range may be quoted.
Is the country where the harm was counted to blame?
No. A country page shows where harm was counted and where the products it reaches were made and consumed; the two are usually not the same people. How much of a conflict’s toll reaches the economy is a scenario — the default routes all of it — and every page lets a reader switch the scenario and see the figure move.
Citing a figure
Cite the run and the page the figure came from, as each page’s own citation block gives it:
HumanToll <year>, run <id>, methodology v<version>. <the page’s address>
What every figure is
HumanToll estimates human harm statistically attributed to supply chains under an explicit set of modelling assumptions. It is not a causal claim about any individual death or injury, not a statement about any company, and not a complete accounting — harm that cannot be attributed is not counted, so every figure is a lower bound. Figures are production-phase only and exclude harm arising from use of the product. Read the methodology before quoting a number, and quote its range with it.
Questions, and errors
A question about a figure or the method goes to info@quiescence.eu