Impact Value methodology
How we turn an NGO's report into a single, comparable Impact Value (IV). The method is deterministic, auditable, and versioned — the same input and table version always produce the same score.
0. Extraction (LLM) vs scoring (deterministic)
A language model reads the free-text report and extracts a structured list of actions (type, quantity, unit). That is all the model does. It never assigns value. The score below is a pure function of the extracted numbers and the published tables, so it is reproducible and cannot be talked up by clever wording. If the model is unavailable, a deterministic keyword parser is used instead.
AW — Action Weight
Each recognised action maps to a base weight and a set of framework tags (UN SDGs, plus the Ecological Benefits Framework for environmental actions). Unknown actions score zero until added to the table. Current seed table:
| trees planted | per trees | 0.1 | SDG-13 · SDG-15 · EBF:carbon · EBF:biodiversity |
| mangroves planted | per trees | 0.15 | SDG-13 · SDG-14 · SDG-15 · EBF:carbon · EBF:biodiversity · EBF:water |
| coral planted | per fragments | 0.4 | SDG-14 · EBF:biodiversity · EBF:water |
| waste collected kg | per kg | 0.05 | SDG-12 · SDG-14 · EBF:soil · EBF:water |
| plastic recycled kg | per kg | 0.06 | SDG-12 · EBF:soil · EBF:water |
| co2 offset ton | per tCO2e | 1 | SDG-13 · EBF:carbon |
| hectares restored | per ha | 2 | SDG-15 · EBF:biodiversity · EBF:soil |
| water purified liters | per liters | 0.001 | SDG-6 · EBF:water |
| renewable energy kwh | per kWh | 0.002 | SDG-7 · SDG-13 · EBF:carbon · EBF:air |
| animals rescued | per animals | 0.5 | SDG-15 |
| animals sterilized | per animals | 0.3 | SDG-15 |
| animals treated | per animals | 0.2 | SDG-15 |
| animals adopted | per animals | 0.6 | SDG-15 |
| wildlife released | per animals | 0.8 | SDG-15 |
| students taught | per students | 0.2 | SDG-4 |
| workshops held | per workshops | 0.05 | SDG-4 |
| scholarships granted | per scholarships | 1 | SDG-4 |
| teachers trained | per teachers | 0.5 | SDG-4 |
| books distributed | per books | 0.02 | SDG-4 |
| schools built | per schools | 50 | SDG-4 · SDG-9 |
| meals provided | per meals | 0.02 | SDG-1 · SDG-2 |
| people housed | per people | 1 | SDG-1 |
| microloans issued | per loans | 0.5 | SDG-1 · SDG-8 |
| jobs created | per jobs | 1.5 | SDG-1 · SDG-8 |
| families supported | per families | 0.4 | SDG-1 |
| clean water access people | per people | 0.3 | SDG-6 · SDG-1 |
| people trained | per people | 0.1 | SDG-8 |
| volunteers mobilized | per volunteers | 0.05 | SDG-17 |
| women empowered | per women | 0.3 | SDG-5 · SDG-8 |
| community events held | per events | 0.1 | SDG-11 |
| patients treated | per patients | 0.2 | SDG-3 |
| vaccinations administered | per vaccinations | 0.05 | SDG-3 |
| medical kits distributed | per kits | 0.1 | SDG-3 |
| mental health sessions | per sessions | 0.15 | SDG-3 |
SM — Scope Multiplier
Rewards reach. Applied per action by its quantity:
- quantity ≥ 1,001 → ×2
- quantity ≥ 500 → ×1.5
- quantity ≥ 100 → ×1.2
- quantity ≥ 0 → ×1
TBV — Time-Based Value
Longer, sustained programmes count for more: 1 + 0.1 × years, capped at ×2.0. With no reporting period given, TBV = 1.0.
ESM — Environmental Sensitivity Multiplier
Applied to environmental actions only, by where the work happened (sensitive ecosystems weigh more). Beta uses a rule-based region table; later releases will use geospatial datasets (WDPA / Aqueduct). Default 1.0.
PIM — Population Impact Multiplier
How many people the work reaches, by population density of the area:
- low → ×1
- medium → ×1.2
- high → ×1.5
- very high → ×1.8
ACDM — Action Complexity & Difficulty
The average of five self-reported answers (each mapped to a multiplier). Harder, more regulated, larger-scale work scores higher:
Anti-gaming & integrity
- Each action quantity is clamped to 1,000,000 before scoring.
- Tables are versioned; a score is always stamped with the version that produced it.
- The LLM extracts but never scores; its output is validated before it reaches the formula.
- Every IV ships with a full per-factor breakdown (see any submission's detail page).
- Human verification is required before an impact is tokenised — the score only assists.
Status & honesty
These weights and multipliers are seed values (v0.1-seed-2026-06) spanning the full impact spectrum. They are platform-assessed and not third-party certified. They are meant to be calibrated with domain experts and published openly. Impact Value is a transparency and comparison signal, not a financial valuation; price is set by the creator.