A biodiversity offset is the fourth step of a four-step sequence. By the time you are pricing one, three better options have already been exhausted.
That is not an accusation. It is the definition. The International Finance Corporation's Performance Standard 6 — the document a large share of global project finance is underwritten against — defines a biodiversity offset as a measurable conservation outcome designed to compensate for "significant residual adverse biodiversity impacts... persisting after appropriate avoidance, minimization and restoration measures have been taken."
Residual. Persisting. After.
what a biodiversity offset actually is
A biodiversity offset is a measurable conservation gain, produced somewhere else, credited against a loss you are about to cause here.
The gain is usually real work: fencing, seeding, hydrological repair, a management plan with a funded term. The problem is rarely the work. The problem is the word against.
Offsets exist because permitting systems need a way to say yes to development without saying yes to unlimited loss. That is a legitimate job, and no-net-loss policy is a serious attempt at it rather than a marketing invention — the mitigation hierarchy it rests on was formalized by the Convention on Biological Diversity in 1992.
The useful question is not offsets or no offsets. It is where in the sequence the money enters.
the hierarchy is the product. the offset is the residual.
Four steps, ordered on purpose.
| step | what it does | what it costs you | what it protects |
|---|---|---|---|
| avoid | don't put the footprint there | design time, siting flexibility, sometimes the scheme | the actual place |
| minimize | shrink the footprint, change the method, time the works | engineering cost, schedule | most of the actual place |
| restore on-site | repair what you disturbed, where you disturbed it | capex plus a maintenance obligation | the site's own recovery |
| offset | fund an equivalent gain elsewhere | a purchase price | somewhere else |
Only the first three protect the site in front of you. The fourth moves the accounting.
The part that matters if you sign permits: If avoidance was misjudged — the survey missed a species, the hydrology was more connected than the model said — you can still revisit a design. If the offset was misjudged, the site is already gone, and what you hold is a management contract somewhere else and a hole where the thing used to be.
The offset is the only step that begins by conceding the loss. Restoration can fail just as permanently. The difference is that the offset spends the loss first, then hopes the gain holds.
like-for-like, and the honest name for the rest
Performance Standard 6 requires an offset to follow the "like-for-like or better" principle: conserve the same biodiversity values the project is impacting. Chalk grassland for chalk grassland. An in-kind swap.
The same standard permits an out-of-kind offset, known as trading up — where the impacted biodiversity is neither a national nor a local priority, and higher-priority biodiversity elsewhere is under imminent threat, a low-priority loss may be offset with a high-priority gain.
Trading up is defensible ecology. It is also where the instrument stops being an accounting identity and becomes a judgment about which living systems matter more — normally commissioned by the party that needs the permit. Not fraud. A structural conflict of interest that deserves naming rather than an appendix.
And underneath both variants: like-for-like is a claim about ecological equivalence, and equivalence between two distinct places is an approximation, not a fact. Soil biota, hydrological position, and forty years of undisturbed succession do not transfer. A hectare is not a hectare. The metric that says otherwise is a convention, adopted because a consenting authority needs a number by Friday.
Then there is leakage: protecting one parcel often displaces the pressure instead of removing it, so the grazing, clearing, or drainage relocates and the ledger books a gain the landscape never received.
You may be reading this as the position of someone who has never had to consent a road. Fair — here is the concession. Avoidance is not always on the table. Sometimes the alignment is fixed by geology, the grid connection has one viable corridor, and the residual impact is genuinely residual. There, an offset is the least-bad instrument available, and refusing to use one is not principle. It is paralysis.
So the objection is not that the fourth step exists. It is that the fourth step is now reachable first — priced, brokered, and purchasable before anyone has argued seriously about the first three.
what the corporate standard-setter did with the offset step
If you expect the strongest institutional objection to offsets to come from campaigners, check where it actually came from.
The Science Based Targets Network is the nature-side sibling of the emissions-focused SBTi — the body large companies use to make nature targets credible to investors. Its framework, AR3T (avoid, reduce, regenerate and restore, transform), is the mitigation hierarchy ported into corporate target-setting, built explicitly on IFC Performance Standard 6. It does not accept biodiversity offsets or credits as a substitute for the avoid-and-reduce steps.
When the guidance landed in 2023, technical director Varsha Vijay framed the reason operationally rather than morally: "What we're trying to incentivize on the ground is not possible with the use of offsets." Credits, she added, still lack global standards, regulation, or substantial safeguards.
SBTN tightened this in its 2025 guidance on remediating converted land: companies must remediate within the landscape, ecosystem, or sub-national jurisdiction where the conversion occurred, and buying credits to satisfy that requirement — outside a multi-stakeholder, landscape-scale program — is not permitted.
Read that as a design instruction, not a scolding. The geography of the harm is part of the obligation: a loss in one watershed is not discharged by a gain in another.
the constituency most permitting files never read
On 2 October 2024, ahead of COP16 in Cali, more than 270 civil society organizations and academics signed a joint statement opposing biodiversity offsets and credits outright. Signatories included Third World Network, Friends of the Earth International, the Global Forest Coalition, Rainforest Foundation UK, and the Indigenous Environmental Network. Their language is not hedged: offsets and credits are "false solutions to a false problem."
We do not sign that statement, and this is not a dunk we are joining. Several instruments it rejects — statutory habitat banking, in-lieu fee programs, compliance registers — do real work, and the people running them are not villains.
But if you are structuring an offset program and have not read it, you are missing the argument made by, and on behalf of, people who end up hosting offset sites. The statement's load-bearing claims are political, not procedural: Global North buyers purchasing cheap credits from the Global South; Indigenous Peoples and local communities receiving a fraction of the proceeds; land grabbing and displacement. Those are facts in the file. They are also, incidentally, the most complete risk register on an offset — written by people who did not write it for you.
Name the constituency; do not caricature it. It has the numbers, standing at the Convention on Biological Diversity, and a decade of carbon-market case files.
contribution: money that does not buy permission
The most useful move in the offset debate did not come from the abolition side. Barbara Haya, who runs the Carbon Trading Project at UC Berkeley, argues that poor credit quality is inherent to the incentive structure rather than a symptom of bad actors — every participant in a credit market benefits from more credits existing. Her exit is not a better unit. It is a different claim.
A contribution claim says: we funded this work. It does not say: therefore our loss is cancelled.
That single deletion changes the economics. Once the money no longer buys permission, the buyer stops needing the seller to certify equivalence, and the apparatus of counterfactual baselines, ecological exchange rates, and permanence guarantees becomes optional rather than load-bearing. The fence gets built either way. What disappears is the second product — the license.
Isn't that just an offset that gave up its best feature? In accounting terms, yes — and that feature is the one carrying every failure mode in the instrument.
| what you buy | what you may claim | what can break it | |
|---|---|---|---|
| offset | a certified unit of gain elsewhere | your residual loss is compensated | equivalence, baseline, permanence, leakage, consent |
| contribution | funded conservation work | you paid for the work | little — the claim is what happened |
| certificate of ensurance | a hold on one named place | that place is funded now, by you | the place's condition, measured and public |
The third row is not a better offset. It is not an offset at all.
where ensurance sits in this
ensurance is contribution architecture attached to a named place.
The mechanics, plainly: each place, community, or purpose has an onchain account we call an agent. A certificate of specific ensurance is issued 1:1 against one of those agents. Buying it funds the present condition of that named place, and proceeds route to the stewards responsible for it. There is no second unit — nothing a third party can present to net off a loss somewhere else.
That constraint costs us. A fungible unit is a far easier sell: market price, secondary bid, a compliance use case finance departments already budget for. The certificate has none of those. We are not building toward them, and we are honest that volumes are small — general ensurance coins are a protocol currency, not a claim of gain; specific certificates are live, not at scale.
What we will not do is print the thing that makes the sale easy. A certificate of ensurance is not a license to clear the next parcel. The moment it becomes one, the permission is the product and the place is the packaging, and every criticism in this post applies to us.
Nor are we a replacement for the hierarchy or the regulator. Avoidance remains step one, and statutory regimes — including the UK's biodiversity net gain duty, covered separately in demand by statute — do work no voluntary instrument can do. Ensurance is money for the residual place, entering after regulation has done what regulation can do. Not instead of it.
if you sign the permits
Two moves are inside a consenting authority's control, and neither requires abolishing offsets.
Make step one expensive to skip. Hierarchies fail not because offsets exist but because avoidance is never priced. If the alternatives assessment runs three paragraphs and the offset arrives as a spreadsheet with a delivery schedule, you have told the applicant which one you take seriously. Then make "like-for-like or better" testable on the record before consent: which values, measured how, against what baseline, funded for how long.
Separate the money from the permission. A jurisdiction can welcome funded stewardship on a named place while declining to treat it as compensation for a loss elsewhere. That move belongs to regulators as much as buyers: it keeps the capital and removes the license.
the bottom line
Keep the hierarchy. Fund the residual place. Do not sell anyone the right to skip step one.
A biodiversity offset converts a certain loss here into a probable gain there, then records the two as equal. Sometimes that is the best trade on the table. It is never free, and it is never the first thing to reach for.
what this looks like for a permitting authority →
talk about credits, offsets, and what you actually hold →
Adjacent reading: how to transition from nature-negative to nature-positive — the same argument from the corporate strategy side.