A Working Standard for Impact Accounting

What this is

An accounting standard for impact: the rules by which a social or environmental outcome becomes a comparable, auditable number. It applies the discipline of financial accounting (double-entry, five account types, recognition rules, consolidation boundaries) to six capitals, so that outcomes can be recorded, compared, and improved with the same rigor the world already applies to money.

It is a standard, not software: published so anyone can read, apply, cite, or contest it. Each area sets out its method, the parameters it uses, the published sources behind it, its maturity status, and the conditions that would reopen it. The standard is open about what it has settled and what it has not.

Who it’s for

Preparers whose impact figures carry real weight: regulatory submissions, compensation, sourcing and investment decisions, LP reporting; the auditors who assure those figures; and the standard-setters and researchers building the field.

The shape of the standard

Six capitals

Every outcome is booked against one of six capitals. The classification is the established integrated-reporting model — from the International Integrated Reporting Council (IIRC) framework and the Capitals Coalition Protocol — not one this standard invented:

• Natural Stocks of natural resources — forests, fisheries, soil, minerals, the atmosphere — and the ecosystem services they provide, from clean water and pollination to climate stability and biodiversity.
• Human The knowledge, skills, health, and capacity that exist inside individual people — embodied in the person, and lost when they leave.
• Social The relationships, networks, trust, and shared norms that exist between people, and between an organization and its stakeholders — the connective tissue that makes coordinated action possible.
• Intellectual Knowledge-based intangibles that exist independently of any single person — patents, brand, software, documented processes, and institutional know-how that survive turnover.
• Manufactured Physical objects produced by people — buildings, machinery, equipment, vehicles, and infrastructure — the durable stuff that enables production and operation.
• Financial Money and money-equivalents — cash, securities, receivables, debt, and equity claims — a store of value that produces outcomes only when converted into one of the other capitals.

Five account types

Each capital carries the same five account types as financial accounting: Asset, Liability, Equity, Revenue, Expense, with standard debit/credit behavior. Six capitals × five types = thirty categories. The books balance per capital: Assets − Liabilities = Equity; Revenue − Expenses = net impact, closing to Equity each period.

Monetize or narrate — the recognition rule

An account is monetized when a published, citable methodology can price it. Otherwise it is pending — the honest default for a figure that is monetizable in principle but not yet priced — or, only where the standard can defend that a figure cannot be booked as impact value, it is held at a terminal status: disclosed in narrative (real, reported, never priced) or deliberately excluded (with the reason stated, e.g. a value already counted elsewhere). Three commitments follow:

• Nothing is priced on an invented factor. Factors are inherited from published methodologies and cited; a wrong number in an audited book is a finding.
• Nothing shows a false “pending” it will never clear.
• Activity is never impact. Money disbursed, audits performed, training delivered: volume metrics are disclosed as activity, but impact value books only on outcomes (income created, harm reduced, capability built).

One result, counted once

The standard’s core invariant, inherited from consolidation accounting: one economic result is recognized at most once across the whole chain of reporting entities. It is what lets impact aggregate up a portfolio or a multi-layer fund structure without inflating.

Valued globally, not locally

The standard prices impact on a global basis: one value for a given outcome, applied everywhere, rather than adjusting it downward for people in poorer places. A life, an injury, a unit of harm is not worth less because of where it is lived. This is a deliberate equity choice: local willingness-to-pay would price the same harm cheaper in a poorer country. It is also the position the field’s own valuation framework has moved toward (IFVI’s General Methodology 2 sets the global perspective as its default, and the water methodology was revised specifically to remove a wealth bias). Where an entity has a defensible reason to use a local value, it does so as a disclosed override, never the standard default.

This is a rule about the worth of a unit of harm, not its size. Where a methodology differentiates the physical magnitude of an impact by place — water consumed in a water-scarce country is a larger harm than the same volume in a water-rich one — the standard uses those location-specific factors, resolved by default to the reporting entity’s own country. What never varies by place is the value of a unit of human harm or wellbeing once measured.

Holdings

The standard supports investor → fund → investee chains on the pattern financial accounting already uses for investments (the equity method): the investee books each result once, on its own books, and every holder above it carries a proportional claim — in proportion to its share of the investee’s capital — computed as a rollup, never re-booked as a second posting. The share basis is disclosed, and changes to it are disclosed like any method change. That is what lets impact aggregate up a portfolio or a multi-layer fund structure without double-counting. Worked multi-layer examples are part of an engagement.

The shape of the standard, in one example

One observation, followed end to end: through recognition, correction, and the statements. The value of a statistical life is the factor the standard has adopted; the risk rates are illustrative inputs, so the arithmetic can be rerun with any rates.

The setup

A coffee farm in Guatemala employs 50 workers. During a certification inspection, the inspector records that workers handling agrochemicals have no personal protective equipment: no gloves, respirators, boots, or protective clothing. That is one observable fact.

Four certification schemes cover this farm. Each scores that fact its own way, with its own indicator code and its own classification scale. Four indicator codes, four scoring systems, no number.

Impact accounting records the fact once, on a chart of accounts, as a journal entry. The four reports are generated from that entry afterwards.

The books

Each of the six capitals carries the same five account types finance uses and produces its own balance sheet and its own profit and loss. Assets minus liabilities equals equity, per capital, every period. Cash and impact sit on one set of books, in one presentation currency, but the capitals are never summed into a single headline figure. They share a unit, not fungibility.

Pricing the observation

The health cost of missing protective equipment is priced with published health-economics factors, not a house estimate:

The $2.9M value of a statistical life is the global figure from the IFVI factor database, which the standard adopts as its default. By default every life is priced identically; there is no regional discount. Every factor a figure depends on is cited on the posting, and any override of a published factor is disclosed in the report.

Entry 1. The risk is recognized

The same shape as a provision in financial accounting: a probable obligation with a reliable estimate. The expense hits the Human capital P&L; the obligation sits on the Human capital balance sheet. No cash moved, and Financial capital is untouched. The externality is now visible on a statement.

If the farm never corrects the problem, this is where the books stay. The liability does not go away because nobody paid it.

Entry 2a. The farm buys protective equipment

The farm spends $5,000 cash on equipment. Value leaves Financial capital and enters Human capital, so the entry carries an equity transfer line in each capital to keep both balance sheets in balance:

Debits 10,000, credits 10,000. Financial: assets down 5,000, equity down 5,000. Human: assets up 5,000, equity up 5,000.

Entry 2b. The obligation is released

Workers are now protected, so the recognized obligation is reversed and the value created is booked as a gain:

What the statements show after the period

Human capital. P&L: expense 148,000, gain 148,000, net impact zero for the period. Balance sheet: protective equipment 5,000, liability nil, equity 5,000.

Financial capital. Cash down 5,000, equity down 5,000.

A reader sees what actually happened: 5,000 of cash retired a 148,000 obligation to workers. Under cash accounting only the 5,000 would ever appear. Note what is not booked. The 5,000 the farm would have “saved” by never buying the equipment is not a transaction; the standard never records savings from inaction. And the two capitals are not netted against each other into one number.

What the certification schemes get

From that one journal entry, each scheme’s report can be generated in its own format: the finding recorded against the scheme’s own indicator, the corrective action dated, the indicator closed when the equipment is in place. Compliance reporting becomes a byproduct of keeping the books, not a separate exercise per scheme.

What accumulates

Across many farms and several years, the ledger holds every observation, every intervention, and every outcome. That is what the second half of Sedoha is built for: the continuous-improvement engine is designed to compare interventions against outcomes and learn, for example, whether protective equipment alone or equipment plus safety training produces lower injury rates for farms with heavy agrochemical use. The next farm that asks what to do gets an answer drawn from evidence, not a checklist.

Where the standard comes from

The standard is an integrator: it does not invent valuation factors but inherits published methodologies and assembles them into one chart of accounts. Its environmental and workplace valuations come from the factor methodologies published by IFVI (the International Foundation for Valuing Impact), now developed and governed since late 2025 by the Impact Value Standards Board under the Capitals Coalition, through a public due process with open comment periods. Where those methodologies reach, the standard applies their factors as-is and cites them. Where they do not yet reach — most human and social outcomes — it sources or develops a method from other published research and named academic studies, drawing on three named bases: the Harvard Impact-Weighted Accounts framework; the value of a statistical life; and wellbeing valuation, anchored on the WELLBY — the wellbeing-adjusted life year used by the UK Treasury’s Green Book and the LSE wellbeing literature — as the named basis for outcomes whose value is welfare a person experiences. Every area names its source. Building on a body governed by a public standard-setter, rather than on proprietary numbers, is part of what makes the figures audit-ready.

Every factor declares its provenance

Each conversion factor in the books carries one of three provenance classes, and the reader can always see which: a published standard (a factor applied as published by a public methodology, cited); an entity override (a disclosed, justified departure from the published value); or a provisional estimate (a documented interim value used where no published methodology yet reaches, held until one does). Disclosure climbs as provenance descends: the further a figure sits from a published source, the more the books say about how it was made. This mirrors the fair-value hierarchy financial accounting has used for decades, and it gives every provisional figure a stated graduation path: when a public methodology arrives, the estimate retires in its favor.

Audit-ready, not audited

This standard is designed so an independent auditor could test the books, and is careful never to claim more than that.

Audit-ready means the books are kept in a state an auditor can examine: every figure traces to its source, every number can be recomputed, the books balance, judgments are documented, and what is not priced is labeled and explained. Audited means an independent, licensed firm has examined the books and issued a formal opinion. The first is what a preparer can build; the second only an outside auditor can grant. For monetized impact the realistic level of assurance available in the field today is limited assurance (“nothing came to our attention”), and the assurance practice for impact figures is itself still forming. So the honest description of a well-kept set of impact books is audit-ready — never “audited.” Claiming otherwise is exactly the overstatement this standard exists to prevent.

Audit-readiness is not a slogan; it is seven specific capabilities an auditor can test, each one a practice financial auditors already rely on:

A standard that supports all seven is one an auditor can actually work with. Sedoha builds to all seven: the books balance, consistent methods, and the audit trail with period close are fully in force today; trace to source, recompute, documented judgments, and labeled completeness are built in and still maturing.

Why maturity is stated openly

Each area carries its status — adopted, pending, or deferred / disclosed-not-recognized — and the conditions that would reopen it (a standard-setter publishing a factor, a source revising its figures). A standard that hides what is still provisional is not one you can trust; a disclosed quantity with a stated reason is a strength, and a confident number the evidence cannot carry is a liability. Rulings are attributed and dated.

How the standard treats each area

A companion to the sections above. For each area it gives the position, the method (the formula, the parameters the standard uses, and the account the entry books to), what the standard recognizes and what it deliberately does not, the current status, and the sources. Everything here is set out at the level an auditor could apply and check: a figure is not just traceable to its source but recomputable from the rule. What is not here is Sedoha’s product — the software and pipeline that run these rules — and its internal deliberation. The rules themselves are public, as an accounting standard’s rules must be.

Every outcome books to one or more of the six capitals — Natural, Human, Social, Intellectual, Manufactured, and Financial — and each account id below begins with the capital it belongs to (human.safety.fatalities is Human capital, financial.taxonomy.deferred_alignment_obligation is Financial, and so on). Several areas span more than one capital: financial-inclusion harm, for instance, books to three Human accounts and one Social account. Each capital carries its own color, applied to the account it names.

Two rules cut across everything below: the recognition rule — price it or narrate it — and the rule that activity is never impact, both set out in full under “Monetize or narrate” above.

Workplace safety — fatalities, injuries, illness

Human capital · human.safety.fatalities

Position. A workplace death, injury, or illness is a real cost to human capital. The standard records it as an expense in the period it occurs, the same way a business records any cost it has incurred.

Method. Each life lost is valued using the value of a statistical life (VSL), the figure economists derive from what people are collectively willing to pay to reduce a small risk of death. The standard uses IFVI’s published global VSL of $2,895,021 (2024 USD), and applies that one figure equally to every life rather than adjusting it downward for people in poorer countries. That is a deliberate equity choice: a life is not worth less because of where it is lived. A fatality is recorded as count × VSL, booked as an expense to human.safety.fatalities with a matching liability, the standard cost-incurred pattern (the applied factors are IFVI’s severity cells built on this basis: $2,896,310 per fatal injury and $2,895,201 per fatal illness). Non-fatal injuries and illnesses are valued using IFVI’s published factor for each severity level, each of which is a fraction of that same VSL, from long-term incapacity down to temporary injury or illness, with the temporary cases priced per lost workday. When an entity reports only an injury rate rather than a case count, the number of cases is derived as rate × (hours worked / 200,000), where 200,000 hours is the standard full-time base of 100 workers. When the severity of an injury is not disclosed, the standard records it at the lowest severity as a conservative floor and states plainly that the resulting figure understates the true cost.

Status: adopted. Sources: IFVI Occupational Health & Safety methodology (2026); OECD (2025) mortality-risk-valuation meta-analysis; “Impact Accounting Has an Equity Problem” (Stanford Social Innovation Review), the published critique of country-specific life-valuation.

Employment

Human capital · human.employment.turnover_cost

Position. Employment impact is valued on a published, wage-based framework: the Impact-Weighted Accounts developed at Harvard. Losing employees is a cost; creating jobs and paying well are gains. The standard recognizes each where the data supports it.

Method. The cost of turnover is the cost of replacing the people who leave: the Society for Human Resource Management (SHRM) replacement cost (67% of salary, the adopted value within SHRM’s measured 33–200% range) × turnover rate × headcount × average base salary, booked as an expense to human.employment.turnover_cost with a matching liability. Because a departure and its replacement describe the same event, the two are netted against each other so the exit is not counted twice. The value of training is measured as hours × (average salary / 2,080), recorded as an asset, an investment in human capital rather than a cost. (2,080 is the standard full-time work year; this training measure is derived by Sedoha, not a published IFVI or IWA factor, and is flagged as such.)

What is held pending, and why. Two components — the gap between wages paid and a local living wage, and the cost of diversity gaps — depend on a reliable local living-wage benchmark that the standard cannot yet source responsibly. Rather than substitute a legal minimum wage and pass it off as a living wage, the standard holds these two components pending until it has the right input. Status: adopted, with named components pending their inputs.

Sources: the Impact-Weighted Accounts employment framework (Harvard / IWA); published turnover-cost research (SHRM, 33–200% band).

EU Taxonomy alignment

Financial capital · financial.taxonomy.deferred_alignment_obligation

This treats the alignment gap only. It does not re-book the emissions of the misaligned activities, which are booked directly under GHG.

Position. The EU Taxonomy sets a benchmark for how much of a company’s activity should be aligned with the climate transition. The share that is not yet aligned stands for money the company will eventually have to commit to get there. The standard treats that as a deferred obligation: a future commitment built up gradually on the balance sheet, never charged all at once as a current cost.

Method. The obligation is based on the portion of the company’s capital or revenue that is not yet aligned with the benchmark, accrued gradually over the transition period. Each year’s addition is [(benchmark% − aligned%) ÷ 100] × scope ÷ 25, where scope is the company’s reported capital expenditure or revenue in currency, and the benchmark is the company’s disclosed taxonomy-eligible share (the standard uses 100% only where eligibility is not disclosed). Two EU terms sit behind the formula: an activity is eligible when the taxonomy covers it at all, and aligned when it also meets the taxonomy’s technical criteria, so the gap between the eligible share and the aligned share is the part the company still has to bring into line. Put plainly: take that misaligned fraction of the money and spread it evenly across the 25 years to the EU’s 2050 target. It is recorded as a credit to financial.taxonomy.deferred_alignment_obligation with a balancing debit to equity, and no expense in the period. It accretes like a long-dated provision, but deliberately departs from IAS 37: no period expense is charged; the balancing debit is to equity, building the obligation on the balance sheet without touching any period’s operating result.

Status: adopted. Sources: EU Taxonomy Regulation and Delegated Acts; the deferred-provision analogs in financial accounting (IAS 37 / ASC 410).

Gender

Social capital · social.equity.gender_leadership / gender_workforce

This addresses two specific questions — how the income effect of financing that reaches women is treated, and how structural gender gaps in leadership and workforce are treated. It is not a treatment of every gender-related impact.

Position. Value is gender-neutral: a dollar of income is worth the same no matter who earns it. The standard never prices people differently by identity, the same position it takes on the value of a life.

Method — how a loan is valued. How a loan is valued depends on what it funds, not who borrows. A loan for a productive asset — a sewing machine, a delivery bike, farm inputs — raises the borrower’s income; where that gain has been measured in field studies (a randomized trial or equivalent), the standard records the income uplift at local market prices. A loan for consumption — smoothing an income gap, covering an emergency — does not raise income, so the standard records the wellbeing benefit of steadier consumption instead. The test is the type of financing and the evidence behind it, never the borrower’s gender.

What deliberately isn’t booked. When income reaches women it tends to produce further benefits — better child health, more schooling, improved nutrition — and these are real and have been measured in the field. The standard does not yet record them as value: the dollar values (their “shadow prices”) that would put a number on them are still being checked against their primary sources, so for now these benefits are disclosed but not recognized. This follows the same rule financial accounting uses for a contingent asset: you disclose a probable future gain, but you only record it once it is virtually certain. Structural gaps in leadership and workforce (accounts social.equity.gender_leadership and gender_workforce) are treated the same way for a different reason: they are disclosed as gaps, but the cost of those gaps has not yet been derived to the evidentiary bar the rest of this standard requires, so they are held deferred rather than priced.

Grounding & status. No published framework currently monetizes gender impact. The standard states its own position and will adopt a published method if the field converges on one. Status: approach adopted; the downstream co-benefit disclosed but not recognized, under the stated evidence conditions; the structural gap-cost deferred.

Sources: the randomized-trial literature on microcredit income effects (Banerjee, Karlan & Zinman, AEJ: Applied 2015) and published randomized evidence on productive-asset financing; “Impact Accounting Has an Equity Problem” (Stanford Social Innovation Review), the critique of country-specific valuation; IFVI’s Adequate Wages methodology (checked, confirmed gender-neutral).

Avoided emissions

Natural capital · never booked (no posting)

Position and method. “Avoided” or counterfactual emissions — the emissions a product or project claims to prevent somewhere else — are never booked as value created. In a double-entry ledger, recording such a claim would net it against real emissions, cancelling actual harm with a hypothetical benefit, which every major framework prohibits. So the standard assigns no factor and makes no posting. Real flows the entity itself produces (energy generated, land conserved) are booked normally at their own value, and the avoided-emissions claim is disclosed with a stated baseline and boundary, and managed as a goal, but never added to the books. The standard’s position matches the field’s consensus.

Status: adopted (a deliberate decision not to recognize). Sources: GHG Protocol; the Science Based Targets initiative; WBCSD; PCAF; Project Frame; IFVI GHG Topic Methodology.

Financial-inclusion harm

Spans three Human accounts and one Social account

This treats specific harm channels of financial-inclusion lending, not all financial-inclusion impact; the benefit side is treated separately.

Position. When lending harms borrowers — pushing them into over-indebtedness, pulling a child out of school, forcing a distress sale of land, or triggering coercive collection — the standard books that harm, symmetrically with the benefit it books elsewhere, using published welfare valuations. The lender’s own loss from defaults or write-offs is never used as the harm figure: that is the lender’s loss, not the borrower’s.

Method (per channel).

• Over-indebtedness is valued as the loss of wellbeing over the time it lasts: −0.355 points on a 0–10 life-satisfaction scale, per person-year in distress (FCA/Simetrica 2020), converted to a local wellbeing value (about £6,746 per person-year at the source, adjusted to the local income level), booked as an expense with a matching liability, and scaled by surveyed prevalence × the number of active borrowers.
• Education interruption is valued as the present value of the earnings a child forgoes: the country-specific return to schooling (the income gain from each extra year of school, for example 4.3% per year) × GNI per capita × school-years lost, for each documented case of a child leaving school.
• Forced asset disposal is valued as the loss taken in a distress sale: fair value × a distress discount (about 20%, with the band disclosed), the realized loss only, never the full value of the asset, because under IFRS 13 a forced-sale price is not fair value.
• Coerced collection is a breach of rights, not a welfare loss, and is treated as zero-tolerance: valued on a remediation-cost basis where one can be defined, never priced on a wellbeing scale and never netted against benefits. The number of cases is disclosed regardless.

Why the figures wait. Booking these harms depends on real prevalence data: how often each one actually occurs. Until that evidence is in hand, the harm is carried as an honest disclosed gap rather than a fabricated number. Status: adopted; specific figures pending evidence.

Sources: FCA / Simetrica (2020) wellbeing-cost study; Montenegro & Patrinos (World Bank WPS7020) on returns to schooling; the IFC accountability office’s documented Cambodia case; 60 Decibels MFI Index (2025); HM Treasury Green Book wellbeing guidance; IFRS 13.

What’s ahead

This is a working standard, published as it matures. Areas in active development — further method treatments across the six capitals and broader framework mappings — publish here as they reach the bar the existing treatments meet.

Changelog

Seeded at first publication; every substantive revision lands here with its date. The full revision history is retained.

• 2026-09-06 — Added “The shape of the standard, in one example”: a farm safety finding followed through recognition, correction, and the per-capital statements, on adopted factors, so the mechanics are shown rather than described.
• 2026-07-18 — Added “Where the standard comes from” (the integrator role; factors from IFVI, now governed by the Impact Value Standards Board under the Capitals Coalition) and “Valued globally, not locally” (the global-equity pricing stance as a front-door principle). Noted the six-capital classification’s IIRC / Capitals Coalition lineage.
• 2026-07-17 — Added “Audit-ready, not audited”: the seven testable capabilities and the audit-ready-vs-audited distinction. The recognition rule now names its full grammar (monetized / pending / disclosed-in-narrative / deliberately-excluded).

Notices

These notices are published in good faith and accompany the standard. Third-party methodologies (IFVI, IWA, PCAF, the GHG Protocol, the EU Taxonomy, academic sources) are cited and referenced, never reproduced at scale.

No warranty, not advice

Published as a good-faith working document, provided “as is.” Sedoha does not warrant its accuracy or completeness, and makes no representation that applying it will satisfy any regulatory requirement or be accepted by any assurer, regulator, or counterparty.

It is not professional advice and does not replace a qualified professional’s judgment or any applicable law, regulation, or reporting framework. How to apply it is the user’s responsibility.

A living, revised standard

This is a living standard: it is revised continuously; the standard carries its last-revised date above, and the changelog records each substantive change. Reopen conditions stated in the standard (for example, that a decision will be revisited if a named external methodology is published or revised) are statements of Sedoha’s maintenance intent, not commitments or warranties to any party. Prior states of the standard are retained in the public revision history. Cite as: “A Working Standard for Impact Accounting, [section], as revised [date].”

Third-party methodologies and sources

This standard builds on, cites, and maps to methodologies and research authored by others: IFVI, the Harvard-lineage Impact-Weighted Accounts framework (IWA/IWAF), PCAF, the GHG Protocol, SBTi, WBCSD, the EU Taxonomy, and named academic sources. Those works are the property of their authors and are governed by their own terms; they are cited and referenced here, never reproduced at scale, and no endorsement by those authors is implied. Conversion factors and figures attributed to third parties remain subject to those parties’ licensing.

Authorship and license — use freely, with attribution

A Working Standard for Impact Accounting is authored and maintained by Sedoha, and published under the Creative Commons Attribution 4.0 International license (CC BY 4.0). Anyone may read, apply, implement, adapt, build upon, extend, or supersede it — including commercially — provided that use attributes the standard: “A Working Standard for Impact Accounting, Sedoha (sedoha.com/methodology), as revised [date],” with an indication of any changes made. Attribution is the only condition.

← Back to Methodology