Impact Value methodology

Impact Value is Regen Bazaar's own relative index for comparing reports from small NGOs, community groups and individual creators. It is not a certification, not a carbon or biodiversity credit, and it does not let a buyer claim any emissions result. Weights marked "assumption" are platform judgement awaiting an expert review. Projects certified by a registry can declare their serial number, which prevents double claiming.

1. How much impact. Report → AI lists the stated actions → quantities become physical units → score per impact area → one Impact Value.
2. How sure we are. Proof links and media → automatic checks raise flags → a validator sets the proof level P0 to P4.

Screens lead with the domain score and physical units; the single IV is used for price, staking and the overall ranking.

A language model reads the report and lists actions (type, quantity, unit). The creator checks and corrects that list; the validator sees both. The score is a pure function of the numbers and the published tables. Report text is treated as data, so instructions hidden in it are ignored.

A unit that cannot be converted (for example "bags" for kg) scores 0 and is flagged for the validator.

Environment is counted in tCO₂e-equivalent points (1 point = 1 tCO₂e). Every weight has a card with its definition, anchor, calculation, sensitivity and open questions. Status: sourced (a cited table), derived (our arithmetic on a cited table), assumption (needs checking).

SM, area factor (area-based ecosystem actions only, on the report's total area for that action): from 1 ha ×1 · from 0.5 ha ×0.9 · from 0 ha ×0.75; ×1.0 when the site joins existing habitat. Small isolated patches hold less biodiversity (Haddad et al. 2015); bigger sites get no bonus per hectare. All other actions: ×1.0.

ESM, environmental sensitivity (1 to 1.3, environment only): a validator checks the site coordinates on open maps (mangroves, reefs, forest) and sets it. Until then, 1.0.

S, survival: coral ×0.65 (Boström-Einarsson et al. 2020); mangroves by the measured surviving share, ×0.72 by default (proxy from Bourgeois et al. 2024).

The platform does not compare impact across categories at this stage: each category is scored in its own physical units, and the six domain coefficients are all 1.

The server saves a copy of each link (hash and date), and automatic checks compare dates, numbers and place with the report. They only raise flags. A person sets the level; it never changes the impact score. P factors are a pilot hypothesis.

The Impact Value sets the price. The rate is $1 per IV point for now. P is the proof level factor. E is the cost coefficient: the group can declare what the work took (volunteer hours, the value of an hour, money spent on materials, transport, equipment, food and water, paid services), and that raises the price of its impact by at most 50% (E from 1.0 to 1.5). Costs are never added as dollars and never replace the impact: more impact means a higher price at the same cost.

A validator sees the country's statutory minimum hourly wage next to the declared value of an hour, as a reference frame, not a limit. The 50% cap is set for this methodology version and is not calibrated yet.

First sales settle in a dollar stablecoin (USDG) at that price; resale prices are set by the seller. On the Celo test network test CELO is counted as $1 until a stablecoin is added before mainnet. v0.1 reports keep IV × 0.5 in the sale currency.

• Lines of the same action are added up before scoring, so splitting a report changes nothing.
• Area and tree count of one planting: the area counts, the count is evidence.
• Work already registered with a carbon standard (declared serial number) scores no carbon here, so it is not claimed twice.
• Recycled plastic adds only its carbon benefit when collection is also reported.
• Workshops score 0 when participants are counted; meals and family support for the same families count once.
• Volunteers are an input and score 0. Wildlife releases need justification and monitoring (IUCN/SSC 2013).
• One action is capped at 1,000,000 units; carbon at 20 monitored years.
• Nothing the submitter types raises a multiplier: sensitivity and proof level are set by a validator.

Unit definitions are mapped to IRIS+ metric IDs (IRIS+ by the GIIN, cited with attribution) and structured around the Five Dimensions of Impact. SDG tags mean "contributes to"; SDG indicators are country-level.

Every score is stamped with the methodology that produced it and is never rescored. Reports scored before v0.2 keep their v0.1 value (v0.1-seed-2026-06: IV = Σ AW × SM × TBV × ESM × PIM × ACDM, seed weights). The two scales differ, so v0.1 and v0.2 numbers should not be compared directly.

What changed and why: TBV (time bonus) and PIM (population density) removed; ACDM moved to price; SM changed from a size bonus to a small-patch discount; weights moved to physical units with sources; proof levels and double-counting rules added.

• Bernal, Murray, Pearson 2018, Carbon Balance and Management: planted mangrove sequestration, Table 2
• IPCC 2019 Refinement, Vol. 4 Ch. 4 Forest Land, Table 4.10; IPCC 2006 Guidelines Table 4.3
• IPCC 2013 Wetlands Supplement, Ch. 4 Coastal Wetlands (cross-check)
• US EPA WARM v16 (Dec 2023), plastics, Exhibit 5-3
• IFI Default Grid Factors v3.0 (UNFCCC, Dec 2021)
• Gold Standard Safe Drinking Water Supply methodology
• Boström-Einarsson et al. 2020, PLoS ONE: coral restoration survival
• Bourgeois et al. 2024, Science Advances: planted mangrove biomass
• Haddad et al. 2015, Science Advances: habitat fragmentation
• WOAH Terrestrial Code Ch. 7.7, dog population management
• IUCN/SSC Guidelines for Reintroductions 2013
• WHO/UNICEF JMP drinking water service ladder (SDG 6.1.1)

Open maps validators use to check the site (an automatic lookup is planned):

• Global Mangrove Watch (CC BY 4.0)
• Allen Coral Atlas (CC BY 4.0)
• RESOLVE Ecoregions 2017 (CC BY 4.0)
• Hansen / UMD Global Forest Change (CC BY 4.0)
• ESA WorldCover (CC BY 4.0)
• WRI Aqueduct (CC BY 4.0)
• WorldPop (CC BY 4.0)
• OpenStreetMap contributors (own extract) (ODbL)

Protected-area and species databases whose terms forbid commercial use (WDPA, IUCN Red List, KBA, IBAT) are not queried.