This site shows impact estimates: browse a charity's profile and you'll see something like "your $100 ≈ 50 bed nets." Donors love these conversions, and for good reason — they turn an abstract act into a concrete one. But a conversion is only as honest as its exchange rate. In July 2026 we sat down and audited every cost-per-outcome figure on this site against its source. This article is the full account of what we found, because we think a ratings site that asks charities to be transparent should hold itself to the standard it applies to others.
The policy we adopted
That sentence now governs every estimate here. It sounds obvious. Following it required deleting or rewriting more than half of our figures.
What the audit found
We checked each figure against the charity's own website, FAQ, annual report, or published evaluator data. Three kinds of problems surfaced:
1. Overstated impact. Our worst error: we listed the World Food Programme at $0.06 per emergency food ration, implying $100 fed a small crowd for weeks. WFP's actual published figure — via its ShareTheMeal program — is $0.80 to feed one child for one day. Our number overstated impact roughly thirteen-fold. We also had Feeding America at five times its published "$1 = 10 meals" claim, and deworming at five times Evidence Action's reported cost. In one case we'd priced an "eye surgery" at fifty cents; Sightsavers' published figures are $0.23 for a trachoma treatment and about $41 for a cataract surgery — two different things, and we'd blurred them.
2. Understated impact. Errors ran the other way too. We showed Against Malaria Foundation at $5.50 per bednet; AMF's own figure is about $2, with public donations going entirely to nets. Room to Read's published "$50 funds a year of a child's literacy program" replaced a vaguer, less generous figure of ours. Honesty sometimes means reporting more impact, not less.
3. Numbers that shouldn't have been numbers. For thirteen organisations — including Médecins Sans Frontières, UNICEF USA, Save the Children, and World Central Kitchen — we had displayed per-unit costs that no published source supports. WCK, for instance, we'd listed at $0.50 per meal, when its published all-in meal costs run $3–10 depending on the operation. Those organisations do vital work; they simply don't publish a unit price, and it wasn't ours to invent. Their profiles now describe what a donation supports, with no number attached.
What changed, concretely
- Of 31 audited figures, 2 were kept unchanged, 16 were corrected to the organisation's published figure, and 13 were removed in favour of descriptive text.
- Every remaining figure traces to a claim the organisation makes about itself — a donation-page statement, an FAQ, a gift catalog price, or a published program cost.
- Where an estimate comes from a third party rather than the charity (one case: Founders Pledge's cost-per-tonne estimate for the Clean Air Task Force), the attribution is in the label itself.
- The disclaimer under every impact panel now states the policy exactly, and the changelog records these corrections permanently.
Why publish this?
Because the alternative is the quiet edit — and the quiet edit is exactly the behaviour that makes charity data untrustworthy in the first place. The evaluators we most admire publish their mistakes; it is, counterintuitively, the strongest trust signal there is. We would rather tell you our meal math was wrong and is now right than have you assume every number here was always perfect.
Two things are true at once: impact figures are the most useful tool a donor has, and impact figures are easy to get wrong — by scammers deliberately, and by well-meaning sites like this one through sloppiness. The remedy for both is the same: trace every number to its source, and say plainly when there isn't one.
If you spot a figure on this site that doesn't match what an organisation currently publishes, tell us. We'll check it, fix it, and log the correction — publicly.