GiveWell's criteria page asks a question that has nothing to do with charity ratings: "What will the donation allow to happen that wouldn't have happened otherwise?" Ask that about a forest and most nature giving goes quiet.
That silence is not the fault of the people doing the work. It is a unit problem. Global health has a unit — a net over a bed, a child who did not get malaria. "The environment" is not a unit. It is a shelf holding everything from a lobbying budget to a peat bog.
How to give to nature: name a place instead of a cause, identify the living stock it belongs to, ask what happens to it if your money never arrives, find out who already benefits and whether they pay, define what one unit of the gift actually holds, and agree in advance on the evidence that would show it worked. Six questions, asked before the money moves.
what GiveWell actually does — and what it doesn't
Four criteria run everything GiveWell recommends: evidence of effectiveness, cost-effectiveness, room for more funding, and transparency. Since July 2022 the bar for a funding recommendation has been roughly ten times as cost-effective as unconditional cash transfers. The reviews are published, including the parts that make their own picks look worse.
Now notice what GiveWell does not do. It does not rate every charity. It narrowed to global health and development deliberately, because that is where the evidence base was strongest, and it says so in public. It also treats judgment calls as unavoidable and writes them down rather than hiding them inside a number.
So here is the honest version of this post: the method is portable, the metric is not. Nobody here can tell you that a peatland clears a dollars-per-life-saved bar. Anyone who does is selling you a decimal point they invented. What travels is the discipline — counterfactual, room for more funding, published reasoning — pointed at an object solid enough to hold it. GiveWell is a cousin method, not a scoreboard we claim to top.
Effective altruism forgot the factory argues why the living substrate deserves that discipline in the first place. This post is the checklist.
six tests before you fund a place
| test | the question | a good answer sounds like | walk away if |
|---|---|---|---|
| named place | Which place, exactly? | "Trail Ridge, on the eastern rim of the Okefenokee, in Charlton County, Georgia." | The answer is a region, a biome, or a brand |
| living stock | What kind of living system is it, and what does it produce? | "Inland wetlands. Peat holding water, and a water table the whole swamp depends on." | Nobody can name what would be lost, in what units |
| additionality | What happens to this place if my money never arrives? | "A permit gets finalized." / "The easement lapses." / "Monitoring stops and the encroachment isn't caught." | The honest answer is "the same thing that would happen anyway" |
| who already pays | Who benefits downstream, and are they paying for it? | "A city, a utility, two rivers, an insurer — and no, none of them pay a dime toward it yet." | The beneficiary is a large firm that could pay and your gift lets it keep not paying |
| what one unit holds | What does one increment of my gift actually hold? | "An acre under a recorded restriction, monitored annually." / "Retired mineral rights." / "A stewardship season." | The unit is "awareness," "support," or an unaudited average |
| how you know | What evidence would show this worked, and what would show it failed? | "Water table depth and canopy at these points, published yearly, reviewed with you each January." | Success is defined as the grant being spent |
The tests are ordered on purpose. Each one only makes sense once the one above it has an answer.
Four of them are GiveWell's criteria in different clothes. Additionality and who-already-pays are both room for more funding — what extra money enables, and whether someone else should be paying it. What one unit holds is cost-effectiveness with an honest denominator. How you know is evidence of effectiveness plus transparency. The other two are what the living world adds — a place, and the kind of system that place is.
the tests, one at a time
1. is there a named place?
A theme cannot fail. A place can. That asymmetry is the whole reason to insist on the second one.
"Freshwater" is a category. "The Okefenokee" is nearly half a million acres of blackwater swamp with a specific water table, a specific fire history, and a specific set of people who show up when something goes wrong. You can drive to it. You can be wrong about it in ways someone will correct.
If a fund cannot tell you the place, it is asking you to fund its judgment rather than an outcome. That may be a good bet — GiveWell itself is a bet on somebody else's judgment — but price it as what it is.
2. what living stock is it?
Living systems come in kinds, and the kind decides what can be measured. Our accounting sorts them into fifteen ecosystem stocks — inland wetlands, temperate forests, grasslands, rivers and lakes, coastal systems, and so on — each producing a set of flows: clean water, water abundance, risk resilience, habitat, healthy soils, and the rest of the nineteen.
Naming the stock turns a sentiment into a measurable claim. Inland wetlands is a stock; peat depth, water table, and outflow are its instruments. The Okefenokee is one instance of it — a peat raft that rises and settles with water, feeding the Suwannee and the St. Marys. "Wetlands are important" is a mood. "This peat holds this water table" is a claim you can be wrong about.
3. what happens if you don't fund it?
This is additionality, and it is the test that most conservation appeals quietly skip.
The Okefenokee gives you a clean worked answer. Twin Pines Minerals had sought permits since 2019 to mine heavy mineral sands on Trail Ridge, the sandy formation along the swamp's eastern edge that experts have described as the rim of the bowl holding the swamp's water in. Draft state permits were close to final. In June 2025, The Conservation Fund agreed to buy roughly 7,700 acres from the company for about $60 million, mineral rights included, closing by the end of July. Ten to fifteen major private donors were funding the purchase, with more expected.
That is what additionality looks like when you can see it: a named counterfactual (a mine approaching a state permit), a price (about $60 million), and a deadline (a permit nearly signed). It is also what room for more funding looks like — Georgia Rivers has said more than 25,000 acres along Trail Ridge remain exposed to mining. The project stopped; the question did not disappear, it relocated.
One caution, because the test is easy to abuse. The absence of a bulldozer is not the absence of additionality. Most living systems are not lost to a single permit but to slow subtraction: an easement nobody monitors, a ditch nobody plugs, a fire regime nobody funds, a family that sells because the tax bill arrived and the stewardship grant did not. Slow loss is still loss, and preventing it is still counterfactual impact. It is just harder to photograph.
4. who already pays — and who should?
Every living system has beneficiaries. Ask who they are before you decide your gift is the right instrument.
A city that draws its water below a wetland benefits. So does the utility, the insurer writing policies downstream, the farm irrigating from the same aquifer, and the data center that needs cooling water in August. Sometimes one of them already pays, through a payment for ecosystem services arrangement — a contract to keep a flow coming. What payment for ecosystem services actually is works through that mechanism and its limits; this post will not repeat it.
What matters for your decision is which of two situations you are in.
A beneficiary could pay and doesn't. Your grant may be absorbing an operating cost a well-capitalized party should carry. That can still be the right move as a bridge, especially if the gift buys the assessment that makes the beneficiary's obligation legible. Just do it with your eyes open, and ask what changes after the bridge.
No beneficiary can pay, and none is coming. Existence, legacy, habitat, and research value rarely have a payer at all. This is where philanthropy is not a subsidy for someone else's balance sheet — it is the only instrument available. That is a feature, not an apology.
The mirror image of this test is written for the person who owns the exposure: what the first check actually buys is about a dependent funding the condition their own premium already prices. If that describes you, that post is the more useful one.
5. what does one unit of the gift hold?
GiveWell can say what a dollar buys because the intervention has a unit. Conservation has units too; they are just less standardized, so people skip them and write "support."
Real units look like this: an acre under a recorded conservation easement, monitored at least once a calendar year, which is what Land Trust Alliance accreditation standards require of the land trusts that hold them. Mineral rights retired in a deed. Acre-feet still in a wet meadow in August. A stewardship season with a named crew. US land trusts were involved in about 61 million acres as of the 2020 census — about 20 million under easement and 8.5 million owned outright, the rest reconveyed to agencies or protected other ways. The easement acres are the ones a land trust must walk at least once a calendar year.
Our own accounting engine, RealValue, prices the annual flows a specific parcel produces against the cost of securing it, which is how a premium on a natural asset gets set. Read that number as what it is: a bridge that makes a living system legible to capital. It is not a statement of what the swamp is worth. The price exists so the protection can be funded. It never stands in for the thing.
6. how would you know it worked?
Write the evidence down before the money moves, including what failure would look like.
Condition is measurable now, and not only by satellite: water table depth, streamflow, soil moisture, canopy cover, forage, species counts, permits withdrawn, restrictions recorded. Pick a handful. Set the points, the frequency, and the date you and the steward will sit down and read them together. Agree that "the grant was fully spent" is not a result.
Then fund the measuring. Asking a land trust for annual monitoring data while restricting your grant to fieldwork is asking for a report nobody is paid to write.
the same six tests, run on one place
Honest answers, including the parts we cannot answer.
| test | Trail Ridge / Okefenokee |
|---|---|
| named place | Trail Ridge, eastern rim of the Okefenokee National Wildlife Refuge, southeast Georgia |
| living stock | Inland wetlands — peat, water table, blackwater outflow to the Suwannee and St. Marys |
| additionality | Answered in June 2025 for the mine site: about $60 million, mineral rights retired, the state permit never finalized, the sale ended that project. Still open for the 25,000+ acres along the ridge that Georgia Rivers says remain exposed |
| who already pays | Two river systems, downstream communities, a refuge budget, and a tourism economy benefit. None of them paid for the acquisition — a national nonprofit's revolving fund and ten to fifteen private donors did |
| what one unit holds | Acres bought outright plus mineral rights extinguished — the strongest unit in conservation, because it removes an option rather than restricting it |
| how you know | The deed and the halted permit path are the proof for the mine site. Whether the water table holds through the next drought is a longer measurement, and nobody has a five-year answer yet |
That last row is the point. A rubric that only accepts settled answers will fund nothing that matters.
where this rubric goes wrong
Applied badly, these six tests become a way to say no while feeling rigorous. Four failure modes worth naming, because land stewards see all of them.
Additionality as a cudgel. If you only fund the newest, most threatened thing, you starve the boring work that keeps already-protected land protected — monitoring, fencing, fire, invasive species, staff. An easement with nobody watching it is a promise decaying quietly.
Unfunded measurement. Demanding evidence without paying for it transfers your due diligence onto a four-person nonprofit. Fund the monitoring you are asking to read.
Novelty over duration. Living systems run on decades. Grant cycles run on twelve months. If your rubric rewards new projects over renewed commitments, you are optimizing against the thing you claim to value.
Treating the steward as a vendor. The people who have tended a place — ranchers, tribes, land trusts, watershed councils, district foresters — knew what it was worth before anyone priced it. Sovereignty and local authority are not line items in your framework. Ask questions; do not audit people who were there first.
Used well, this rubric is a conversation opener with a steward who will usually be relieved someone finally asked. Used badly, it is a spreadsheet that funds nothing.
one limit, stated plainly: a grant is not a holding
If you are giving through a donor-advised fund, be clear about what a grant can and cannot do.
Once assets go to a DAF sponsor, the sponsor has legal control; you hold advisory privileges. A grant recommendation cannot come back to you as a benefit. Internal Revenue Code §4967 puts an excise tax on a donor or advisor who recommends a distribution producing more than an incidental benefit to themselves — which means a DAF grant cannot buy you an asset, a security, or a return you hold. A certificate — a record tied to a named place — can make the relationship legible, but it is minted to and held by the recipient (the place's account or the land trust). What the donor gets is the open, updating view of that record, not the token. It is not consideration, not a return benefit, and not contingent on the grant.
If part of what you want is to hold something with duration, that is a different pocket — a taxable account, not the DAF — and a later post in this series takes it up. The mechanics of moving a grant out to a place are in how to grant from a DAF to a place. None of this is tax, legal, or investment advice; your sponsor and your advisors decide their own parts.
where ensurance fits
The natural asset is the thing. Watersheds, peat, forage, soils, and species exist whether or not anyone funds them. Ensurance is how a gift can sit on one of them — and, from a different pocket, a hold. Named place, measured condition, where the money went, a receipt that keeps updating in the open. It funds the living thing; it is not the living thing.
Two glosses, once, late: an agent is an onchain account that stands for a place, a group of people, or a purpose, and can receive and route funds. inland-wetlands.ensurance is the stock-level door; the grant lands on a named place inside it, such as the Okefenokee. A certificate is a record tied to that named asset, held by the recipient — which is how a gift on a specific place becomes something you can still point at years later.
We are early and we say so. Taking a grant directly from a DAF sponsor is a design target, not a shipped button; today that route usually runs through a land trust or fiscal-sponsor cousin, and we will tell you which one rather than pretend otherwise.
take the next step
The question "how to give to nature" has a boring, durable answer: pick one place, ask the six questions, and stay with it long enough to read the evidence twice.
Run the tests on a place you already care about. If the answers come back thin, that is information — take it to the steward before you take it as a verdict.
- Explore named places already on the map — start with a stock, end with a place
- What payment for ecosystem services actually is — the who-already-pays test, worked all the way through
- Start a giving conversation — bring the place name, the stock, and the one sentence you want to still be true in ten years
A later post in this series takes up what you can still see after the grant clears, which is the sixth test with a longer memory.