For more than a decade, GiveWell described every program’s value the same way: as a multiple of giving cash to very poor people. A program at “10x cash” produced ten times the benefit per dollar of an unconditional cash transfer. It was intuitive, and it had a flaw.
The flaw
The benchmark was tied to a specific charity’s specific program. When GiveDirectly changed how much it transferred, where it worked or what it cost to deliver, the yardstick moved, and every other program’s score moved with it for reasons that had nothing to do with those programs.
The fix
GiveWell’s November 2025 model update defines the benchmark directly: the value of doubling consumption for one year for a person living at the $2.15-a-day international poverty line. That is a fixed unit of welfare. It no longer depends on GiveDirectly’s operations.
GiveWell says the value of the new benchmark is close to the old one, so estimates before and after the change are comparable. A program at “10x” in 2024 is roughly a program at “10x” now.
The bar
The same update states the threshold for funding plainly: “As of November 2025, our bar for funding programs is 8x our benchmark.”
The bar is a clearing price, not a quality grade. GiveWell has a pool of money and a ranked list; the bar is where the money runs out. It was around 10x for several years. A move to 8x means GiveWell expected to fund a little deeper into its list than before, which happens when money grows faster than the supply of the very best opportunities, or when the aid cuts of 2025 opened gaps that are cost-effective but not extraordinary to fill.
What did not change
- The top charities: Helen Keller Intl (vitamin A), Malaria Consortium (seasonal malaria chemoprevention), New Incentives (vaccine incentives), Against Malaria Foundation (nets). Their estimated cost per life saved, averaged over 2022 to 2024, stays in the $3,500 to $5,500 range.
- The moral weights: how GiveWell trades off a death averted against income gained is a separate set of assumptions, published and unchanged by this update.
- The habit of publishing the spreadsheet. The November 2025 model is downloadable, as every version has been.
Why a donor should care
If you give through GiveWell, nothing changes for you. If you read cost-effectiveness numbers from anyone, the change is a reminder of what they are: ratios against a yardstick, and the yardstick is a choice. GiveWell’s new one is more stable and easier to state. It is still a choice, and a program at 8x is “eight times as good as doubling a very poor person’s consumption for a year,” not eight times as good as anything else you might value.
Method notes
Quotations are from GiveWell’s November 2025 cost-effectiveness page. The account of the old benchmark’s flaw is GiveWell’s own stated reasoning, paraphrased.
Sources
Figures come from the cited sources as of the dates shown. Cost-effectiveness estimates change as programs and evidence change; check the source before giving. We are not financial advisers and receive nothing from any charity named here.