The Cost of Knowing

Every carbon project model you will ever be shown starts at the project start date. Hectares, cost per hectare, a planting schedule, a monitoring line, a buffer deduction, an issuance curve. It reads cleanly because it begins at the moment the answer is already yes.

The work that produces the yes sits entirely before that date, and almost nobody publishes what it costs. Not mainly out of secrecy. It is because the stage before a project exists has no project to charge to.

What follows is a description of that stage: what it is made of, what actually drives the cost, how long it honestly takes, and why the numbers are so hard to find. There are no invoices here. Site-specific costs are commercially confidential, and no single project's invoice would generalise anyway. What can be set out precisely is the structure, and the structure is where the surprises live.

The standard specifies the work and does not fund it

The VCS Standard, v5.0 was issued on 16 December 2025. Verra's stated effective date for most of the program updates bearing on project design and implementation is 1 January 2027; changes to Verra's own processes and registry procedures took effect immediately, and some safeguards and stakeholder-engagement requirements carry a longer transition for existing projects (Verra, VCS Version 5 program notice). Read as a cost document rather than a compliance one, v5.0 is unusually explicit about what must already be true before a project starts.

• Section 3.6.1 requires the proponent to hold the right to operate and the right to reductions and removals at the project start date and throughout implementation, substantiated by an analysis of applicable laws, regulations and legal instruments.
• Section 3.6.3 requires that where that analysis determines a project is likely to affect land or resource rights, a land or resource rights analysis is conducted, informed by input received through stakeholder engagement, before the project start date.
• Section 3.6.6 sets out what that analysis must contain: existing tenure categorisation (private, state, communal, open access); existing customary rights, with a list of community lands and territories and their locations; the presence or absence of overlapping claims, competing claims and violent conflict — and, where competing claims exist, the claimants, the locations, the process and an estimated timeline to secure the rights or resolve the dispute.
• Section 3.6.4 requires the team doing that analysis to have in-jurisdiction knowledge of land tenure regulation, customary rights and related social and gender dynamics, decision-making systems, international human rights law, and existing and potential conflicts. Section 3.6.5 requires a proponent without that expertise to partner with external organisations that have it.
• Section 3.17.2 requires a stakeholder engagement plan, and Section 3.18.1 an environmental, social and governance risk assessment on Verra's template, both before the project start date.
• Section 3.17.11 requires that free, prior and informed consent be obtained from customary rights holders and Indigenous Peoples before the project start date, where the stated conditions apply.

Then the clock. A pipeline listing request is due within one year of the initial crediting period start date (Section 3.8.2). The initial registration request is due within five years for AFOLU projects, extended to eight years for afforestation, reforestation and revegetation, restoring wetland ecosystems and improved forest management projects of any size — logged-to-protected-forest excepted — and for AFOLU projects with an ex-ante annual estimate of 20,000 tCO2e or less (Section 3.8.3, Table 7).

Verra's own charges are published and, against the whole, modest: USD 1,500 for a VCS pipeline listing request, USD 3,750 for a registration review, USD 5,000 for a verification review, of which half is a prepayment credited against future issuances, plus USD 750 to open a registry account and USD 750 annually to keep it (Verra Program Fee Schedule, v1.2, 15 September 2026). One activity carries an area-scaled charge before anything is issued: projects applying VMD0055, avoided unplanned deforestation, pay USD 10,000 per project activity data allocation request plus USD 0.25 per hectare of submitted KML, capped at USD 150,000 per project and payable at the time of the request. The consequential line in that document is not a rate. It is the rule that request review fees are not refundable where a project is rejected or a request is returned, and are payable again on resubmission unless Verra determines otherwise.

So the standard now specifies the pre-start work in detail, sets a clock that begins at the crediting period start date, and says nothing whatsoever about who finances the years before it. That silence is the subject of this piece.

The stack

Six workstreams sit between a candidate landscape and a registration-ready file. They are usually costed as line items. They behave more like schedules with cost attached.

Workstream	What it must produce	What actually drives the cost	Honest elapsed time
Tenure and carbon-rights diligence	A legal analysis naming the instrument that grants the right to the reductions, plus a map of competing and overlapping claims.	The number of overlapping instruments, not the area. Records that are undigitised, unindexed, and occasionally inconsistent between the provincial and the national copy.	Months, not weeks; materially longer where a tenure instrument is pending rather than issued.
Consent, run at community pace	A documented process, not a signature: assemblies, notices, minutes, attendance, and a decision reached under the community's own rules.	Calendar and attendance. The number of decision-making bodies. Planting, harvest and ceremonial seasons. Repeat travel to places a vehicle cannot reach.	From first contact to a documented decision, most of a year is normal; longer where leadership is contested.
Boundary reconciliation	One polygon that survives being walked, and a written statement of what was excluded and why.	The size of the gap between the mapped layer and the ground: cloud, canopy, fragmentation, handset accuracy, and disputed edges.	Weeks of field time per landscape, then comparable desk time to reconcile it.
Site viability and hydrological screening	An ordered set of pass or fail tests applied on the ground, and a record of every disqualification.	Tide and season windows. Access. Above all, the number of candidate sites you must walk in order to find the ones that work.	Seasonally constrained; a missed window costs a year, not a month.
Biodiversity baselining	A baseline a verifier and a biodiversity buyer would both accept: survey design, camera traps, acoustic loggers, identifiable taxa.	Deployment and retrieval cycles rather than analysis. Seasonality. Data volume and the identification effort behind it.	One full seasonal cycle as a minimum; two if a year-on-year comparison is wanted.
Registry and assurance readiness	Project description, land and resource rights analysis, stakeholder engagement plan, environmental, social and governance risk assessment, and a validation body with capacity to take you.	Iteration. Every returned request restarts a review and repeats a fee. Validator availability is a queue problem before it is a price problem.	Two to four quarters, of which a meaningful share is waiting rather than working.

The last column does not add up, because several of these run in parallel. It also does not compress to a single season. On our own reading of the work, two to three years from first site visit to a registration-ready file is an ordinary outcome for a first project in a jurisdiction where tenure is communal and consent is statutory. A six-month estimate describes either a repeat project in a landscape the developer already knows, or a schedule that has not yet met the calendar.

Consent is a calendar cost, not a consultancy cost

The Philippines is a useful worked example, because the process is codified rather than improvised. Under NCIP Administrative Order No. 3, s. 2012, Section 19 lists carbon trading and related activities among the extractive, intrusive and large-scale undertakings, alongside large-scale agricultural and forestry management projects, the issuance of land tenure or resource use instruments, and the exploration, development and utilisation of natural resources — the last of which expressly carries the feasibility studies related to it. Section 19 then closes by directing everything on that list to the Section 22 process. For carbon work, in other words, the question of which consent track applies is not open. The instrument settles it, and it settles it on the long track.

• The full track (Section 22) requires two community assemblies. The first is held once the FPIC fee is paid and released, after written notices are posted in and around the community seven days beforehand and formal notices are personally served on the ancestral domain representatives named in the CADT Recognition Book or ADSDPP Book, also seven days beforehand. Decision-making and consensus-building must begin not less than ten days after the second assembly and conclude within not more than two months.
• The short track exists and carbon cannot use it. Section 24's non-extractive, small-scale process runs on two meetings with the council of elders and leaders, with consensus-building beginning not less than ten days after the first meeting and completed within thirty days. Section 19 puts carbon outside it.
• The field-based investigation (Section 13) commences within ten days of the fee being deposited or paid and must be completed within ten working days of actual commencement, absent force majeure.
• Under Section 12, the proponent funds a work and financial plan agreed at the pre-conference — food and snacks, lodging, transport, documentation — deposited into a trust account established by the Regional Office and computed at rates applicable in the area where the work is done.
• Under Section 29, inaction by the applicant for six months from the last FPIC activity, without justifiable reason, is a ground for the Regional Director to declare the process terminated.

Two things follow. First, the day-rate view of consent is the wrong model. The cash is real but bounded, and partly statutory; the binding constraint is elapsed calendar and the obligation to keep turning up. A six-month lapse in your own attention can end the process you have already paid for.

Second, cost tracks the number of counterparties as well as the number of hectares — though the published modelling is more restrained on this than practitioners tend to be. Cacho and Lipper's model, reported in Pearson et al., put a 1,000-hectare project with 625 participating actors at USD 7.39 per tonne in transaction costs against USD 5.34 for the same area with 200. Pearson and colleagues describe that band as relatively modest, and on a per-tonne basis it is. The mechanism that hurts at the pre-project stage is not the per-tonne arithmetic but the calendar behind it: the statutory clocks above run per decision-making body, not per project, so every additional body adds assemblies, notices and repeat travel to a schedule that is already the binding constraint.

Budget against the current instrument rather than the one you read last year, and check which instrument is actually in force. A revised set of FPIC guidelines intended to replace the 2012 order was approved by the NCIP on 1 May 2026 according to Human Rights Watch (4 August 2026) and the Philippine Daily Inquirer (7 August 2026). On those accounts, assemblies must be convened within ten days of notification, with thirty days for consensus-building on large-scale projects and four days to assembly and seven days to consensus on small-scale ones — against no stringent assembly deadline and up to two months for consensus under the 2012 order. Amnesty International Philippines has publicly criticised the compression on the grounds that fixed day counts are laid over decision processes the guidelines themselves say should follow customary law. Shorter statutory clocks do not reduce the cost of consent. They relocate the risk: the work still has to be done properly, and there is now less room to do it inside the rule. Treat the status as open, though: as of August 2026 several rights organisations were still describing these guidelines as proposed and under consultation, and we have not been able to confirm an administrative order number or an effectivity date against NCIP's own issuance. Confirm both before budgeting against the compressed timelines, and keep the 2012 clocks in the schedule until you have.

The hectarage is the finding, not the input

Screening budgets are set against the area a developer hopes to build on. They are consumed by the area that has to be eliminated.

Global layers are good science and are not parcel boundaries. Global Mangrove Watch v3, the most widely used global reference for mangrove extent, reports an overall map accuracy of 87.4% (95% confidence interval 86.2–88.6%). The accuracies of its individual gain and loss change classes are 58.1% and 60.6%. The authors name the sources of error plainly: misregistration in the underlying radar mosaics, which could only be partially corrected, and confusion in fragmented areas of mangrove — specifically, around aquaculture ponds (Bunting et al., Remote Sensing, 2022). Aquaculture ponds are exactly where abandoned-fishpond restoration screening happens. In the one land-cover transition such a project depends on, the best available global change layer is not far from a coin toss.

This is why the walk is the line item rather than the overhead, and why the cost sits on the wrong side of the finding. In one coastal feasibility assessment of abandoned fishponds the screen was paid for across the whole candidate area and roughly thirty times as much went in as came out — the method and the ordering are set out in The 39-Hectare Answer. What the money actually bought was not the surviving area. It was the documented reason every other block failed, which is the only part of a screen that is useful to whoever underwrites the next site. A screen that cannot produce those reasons has not been run. It has been assumed.

Why nobody publishes it

• It reads as waste. Transaction costs look like inefficiency and are treated accordingly. Phan, Brouwer and Davidson, surveying seventeen forestry carbon projects across Latin America, Asia and Africa, start from the position that these costs are "typically unknown or unquantified and therefore often neglected" (Ecological Economics, 2017). Their own measured range for search, design and negotiation costs was zero to USD 1.738 per tonne, and the characteristic with the largest effect was institutional design.
• The denominator disappears. Registries list projects that got listed. Sites that failed the screen never enter a registry, never appear in a pipeline count and never carry a published cost. Every figure in circulation is drawn from survivors.
• It has nowhere to sit in the accounts. Pre-registration spend is development expense against an asset that does not yet exist. It is not capitalised, not amortised across a crediting period, and not separable in most reporting.
• The empirical literature is thin and dated. Pearson and colleagues remain among the better public accounts: four tropical case studies with transaction costs of USD 0.38 million to USD 27 million, or USD 0.09 to USD 7.71 per tonne, equal to between 0.3% and 270% of anticipated project income. For the voluntary-market projects in that sample, insurance accounted for 41–89% of transaction costs, with monitoring at 3–42% and regulatory approval 8–50%. Their global analysis concluded that most existing estimates of the marginal cost of carbon sequestration are understated by up to 30% because transaction costs were not included. That work was published online in 2013, and the VCS Standard has been through two major version changes since.

The number worth carrying from that work is the spread: 0.3% to 270% of anticipated income. It looks like a useless range and it is the honest one. On a large, institutionally simple site, the cost of knowing rounds to nothing. On a small, fragmented site with many counterparties, it exceeds everything the project would ever earn. The entire purpose of the screen is to establish which of those two you are standing in, before you commit to the second.

How to budget it anyway

• Fund the screen as a decision, not a deliverable. Define the kill criteria before the first site visit, and name who is authorised to invoke them. A screen with no authority to fail is marketing with a field budget.
• Sequence the cheapest disqualifier first. Ordering the tests by cost to run rather than by scientific interest is the largest single saving available at this stage. Tenure conflict and hydrological connectivity disqualify more candidate sites than anything a laboratory will tell you, and both can be established early and cheaply.
• Price consent by calendar, not by day rate. Put the notice periods, assemblies and consensus-building windows into the schedule as fixed objects. Budget for repeat travel, and treat a gap in your own attendance as a live termination risk rather than a saving.
• Hold hectarage provisional until it is walked. Carry two numbers — screened and qualified — and never let the first travel alone into a term sheet.
• Assume re-collection. First-pass field data fails: points outside the boundary, duplicated plots, handsets without signal, an edge the community disputes. Budget the second pass rather than discovering you need one.
• Separate the cost of knowing from the cost of building, and expect not to recover it. It is spent before there is a project and it is unrecoverable if the answer is no. Any structure that implies otherwise is worth reading twice.

One diagnostic question is worth more than a feasibility line item. Ask a developer not what the feasibility study cost, but how many hectares were screened to arrive at the number in front of you, and what disqualified the rest. A developer who can answer has paid the cost of knowing. One who cannot has paid for a document.