Zero Hunger Measurement & Scoring Guide

Recommended measures methods and reporting for the nine pillars

Working version 0.1 • October 23 2026 Companion to Framework and Implementation Handbook v0.3Prepared for World Food Bank Foundation and the Institute for Applied AI & Robotics

Recommendation

Use three auditable measures for each of the nine Zero Hunger Formula (ZHF) pillars. Show the pillar achievement score, its weakest component, change over time, evidence status and any safeguard breach together. Keep observed hunger and nutrition outcomes separate from capacity scores and simulated resilience. This makes the model useful for diagnosis without allowing operational improvement to stand in for demonstrated hunger reduction.

World Food Bank Group (WFB) would coordinate field delivery. The Institute for Applied AI & Robotics (IAAIR), where AI means artificial intelligence, would maintain the technical methods and reporting. Independent reviewers should verify measurements and evaluation claims. These proposed roles do not establish agreements with governments or the organizations reviewed here.

Scientific status and scope

This guide reviews the nine pillars in the v0.3 Framework and Handbook and recommends a more explicit measurement layer. It draws on established international methods, but the proposed indicators, weights, grades and combined equation are research specifications. They have not been validated as a global standard. Repeated field tests may support, revise or reject them. All numerical examples in this guide are synthetic, not results from participating countries.

Objective measurement means reproducible observations under published rules. Indicator selection, acceptable thresholds and weights still involve judgment. No index, audit trail or digital platform can by itself establish causation. The benefits and limitations described below concern suitability for ZHF; a measure may be sound for its original purpose yet insufficient for this framework.

What changes from version 0 point 3

Keep all nine pillars and the candidate outer equation. Expand each single core indicator into three submeasures, using new codes ending in a, b and c to avoid silently redefining the earlier R01–C01 series. Preserve earlier values and run both specifications during a bridge period. The expansion adds detail; it does not add evidence that the equation is correct.

The proposed pilot remains Zambia, Zimbabwe, Mozambique, Botswana and Namibia, and Colombia, Costa Rica, Panama, Nicaragua and Honduras. Colombia is in South America, so the second group is described as Latin America. The same measurement architecture can support cities, provinces and countries, subject to matched boundaries and denominators.

1 Common scoring rules

Each proposed core measure below is a proportion of an independently defined eligible population, area or service that meets a registered standard. For sampled observations use survey or area expansion weights, not a convenience-sample percentage. Record both the raw measurement and the pass or fail decision.

Component achievement s = 100 × N / D, where D > 0 and 0 ≤ N ≤ D.N = eligible units meeting the standard; D = all eligible units in scope.Pillar achievement A = (s_a + s_b + s_c) / 3.Pillar gap = 100 − A. Weakest component = min(s_a, s_b, s_c).

N greater than D is a data error, not a value to clip. For continuous supplementary measures, use 100 × clip((x − L)/(T − L), 0, 1) for beneficial measures, or 100 × clip((U − x)/(U − T), 0, 1) for adverse measures. T is the target; L or U is the fixed worse anchor. Publish raw values because clipping conceals changes outside the anchors.

Achievement band	Proposed descriptive grade
90 to 100	A — high attainment
75 to less than 90	B — substantial attainment
60 to less than 75	C — material gaps
40 to less than 60	D — major gaps
0 to less than 40	E — severe gaps

These are proposed pilot communication bands, not internationally endorsed standards or proof of safety. Assign grades using unrounded scores. Display one decimal place. If the uncertainty interval crosses a band boundary, show the grade range. A high numerical grade with a critical breach must be labeled “safeguard failure”; it cannot receive a satisfactory overall status.

Aggregation choices and safeguards

Equal thirds are transparent starting weights, not estimates of causal importance. Arithmetic means allow trade-offs within a pillar; retain the minimum and require explicit safety gates. Test a geometric mean and a minimum rule alongside this specification. A missing component withholds the pillar composite; never redistribute its weight automatically. A true measured zero is different from missing data.

Food or drinking-water safety failures, unlawful habitat conversion, serious labor or tenure violations and unsuitable financial products trigger separate incident review under the applicable authority. Exclude noncompliant service units from qualifying numerators. Display the measured score, affected people or area and incident status; do not invent an arbitrary numerical penalty.

2 Evidence and comparability rules

Status	Meaning and release rule
V — verified	Current data under the registered protocol, with independent checks and uncertainty stated. Eligible for provisional scoring.
P — provisional	Documented administrative or sampled data with incomplete verification. Show estimates, not assured grades.
M — missing or stale	Required observation unavailable or beyond its registered refresh window. No composite release.
X — inapplicable	Denominator genuinely absent. Explain scope; no automatic zero or perfect score and no silent reweighting.

Set refresh windows by indicator before collection. Operational records are normally monthly or quarterly; ecological measurements follow seasonal and biological change rates. A soil carbon observation can remain current under a registered multi-year protocol, but must not be described as newly measured each quarter. Show observation date separately from report date.

Register before scoring

For every measure record: code and definition; geographic boundary and sourcing footprint; reference period and season; eligible denominator; exclusions; raw units; threshold and rationale; numerator; sampling frame, weights and uncertainty; collector and verifier; source record; refresh date; method version; consent and access class; revisions and appeals. Maintain a ledger of denominator additions and removals so that excluding difficult cases cannot manufacture progress.

For ecological measures, an accredited or otherwise qualified local scientific panel should set soil, habitat and water thresholds by context, using applicable standards and reference sites. Publish the rationale and test equivalent meaning across zones. Universal numerical thresholds for soil carbon or tree cover would be misleading. Until equivalence is demonstrated, compare trends within matched contexts and withhold cross-country ranks.

Sampling and verification

Use stratified probability samples across agroecological zones, farm sizes and relevant population groups. Calibrate instruments, blind laboratory duplicates where feasible and remeasure a random subset independently. Size samples around prespecified precision and clustering, rather than a universal number of farms. For a simple random proportion, n ≈ 1.96² p(1−p)/e² is an initial calculation; inflate for design effects and nonresponse, and apply finite-population correction when appropriate.

Report 95% uncertainty intervals with the design stated. For composite scores, propagate survey, measurement and model uncertainty jointly, preserving correlations. Separately test alternative weights, thresholds and aggregation rules. Distinguish statistical intervals from scenario ranges and methodological sensitivity. An operator’s complete ledger can still have measurement or exclusion errors.

A dashboard should flag the failing service, location, responsible owner, timestamp and documented reason. It may suggest hypotheses and dependencies. It should not infer blame or causation from a red score, or automatically allocate public budgets.

R Regenerative agriculture

Established practice and limitations

Food and Agriculture Organization of the United Nations (FAO) Sustainable Development Goal (SDG) indicator 2.4.1 assesses agricultural sustainability across economic, environmental and social dimensions. FAO soil carbon protocols provide a stronger basis for consistent sampling and stock measurement than practice labels alone. [2,3]

Benefit: these approaches connect land management to measurable condition. Limitation for ZHF: sustainability screening is broad, while carbon changes can be slow, costly to detect and incomplete as a measure of living soil. Certification or hectares adopting a named practice cannot demonstrate restored function.

Recommended component definitions

R.a Soil carbon condition  Eligible sourcing hectares meeting the registered soil organic carbon stock reference band / all eligible sourcing hectares. Estimate stocks with consistent depth, bulk density and coarse-fragment treatment; use equivalent soil mass where needed. Retain tonnes of carbon per hectare and uncertainty. Synthetic example: 600 / 1,000 ha = 60.

R.b Physical soil function  Eligible hectares meeting all registered infiltration and erosion criteria / eligible hectares. Use standardized infiltration tests and measured or calibrated erosion observations in matched seasons; record each criterion separately. Synthetic example: 700 / 1,000 ha = 70.

R.c Input stewardship  Eligible hectares meeting nutrient-balance and pesticide-risk requirements / eligible hectares. Verify nutrient additions, removals and locally justified ranges plus active ingredient, dose, toxicity and exposure controls. Do not equate zero fertilizer with adequate nutrition. Synthetic example: 800 / 1,000 ha = 80.

Scoring collection and verification

R = 70.0, grade C; weakest 60.0. Soil teams collect field and laboratory data; independent agronomists verify samples. Review input records each season; measure physical function annually at comparable times and carbon on a power-justified schedule, initially baseline and endline with interim checks.

Additional monitoring and safeguards

For example only, if a registered infiltration minimum is 10 millimeters per hour, a plot measured at 12 meets that criterion; it must also meet the erosion criterion to qualify for R.b. This is not a recommended universal threshold. Also monitor soil respiration or microbial indicators as research measures, yield stability, labor, net income after transition costs, nutrient density where valid food assays exist, and greenhouse gases. Test rotations, living cover, agroforestry, silvopasture and integrated pest management against local controls. Reduced chemical use earns no stand-alone bonus; verify reduced risk without soil nutrient mining or food-safety harm.

W Water

Established practice and limitations

The World Health Organization (WHO) and United Nations Children’s Fund (UNICEF) Joint Monitoring Programme (JMP) defines safely managed drinking water using source type, accessibility, availability and quality. FAO water-stress methods incorporate environmental flow requirements. [4,5]

Benefit: JMP prevents “a water source exists” from being confused with safe service; basin accounting exposes resource limits. Limitation for ZHF: national annual averages conceal seasonal shortages, unequal access and aquifer depletion. Domestic and agricultural water cannot be added as if liters have equal social meaning.

Recommended component definitions

W.a Safe domestic service  Residents using safely managed drinking-water services / all residents. Follow JMP definitions and estimation methods; use representative household and water-quality evidence, including excluded settlements. Synthetic example: 7,000 / 10,000 people = 70.

W.b Production service  Required production unit-days supplied with adequate water of suitable quality at the required time / all required unit-days. Define a unit within each production stratum; combine stratum percentages with fixed baseline weights, not irrigation volume. Synthetic example: 800 / 1,000 unit-days = 80.

W.c Sustainable source use  Registered source-months meeting withdrawal, environmental flow and groundwater constraints / all required source-months. Use abstraction meters and basin or aquifer observations; fix source boundaries and calendar before scoring. Synthetic example: 90 / 100 source-months = 90.

Scoring collection and verification

W = 80.0, grade B; weakest 70.0. Utilities and farm-water teams collect; water authorities and independent testers verify. Review supply monthly, quality at risk-based intervals and basin conditions seasonally; intensify after floods or contamination.

Additional monitoring and safeguards

Retain each basin’s abstraction volumes and stress ratio, days of supply failure, collection time, affordability, salinity and sanitation or hygiene conditions. A small source and a major basin receive equal weight only under the initial source-month rule: test population and withdrawal-weighted alternatives and disclose concentration. Any unsafe supply triggers action regardless of the mean.

B Biodiversity

Established practice and limitations

The United Nations System of Environmental Economic Accounting Ecosystem Accounting separates ecosystem extent from condition and uses ecological characteristics to describe assets. This supports a spatial baseline rather than counting planted trees as universal progress. [6]

Benefit: area and condition can be tracked consistently through time. Limitation for ZHF: remote sensing cannot establish all ecological functions, and reference conditions differ by ecosystem. Species counts alone can reward invasive species or inappropriate afforestation of native grasslands.

Recommended component definitions

B.a Habitat integrity  Mapped sourcing-landscape units meeting a registered habitat-retention and connectivity standard / all eligible units, weighted by area. Freeze the map and reference date; include a restoration requirement where the baseline is degraded. Synthetic example: 800 / 1,000 ha-equivalents = 80.

B.b Habitat condition  Eligible habitat area meeting the ecological-context condition protocol / all eligible habitat area. Use native vegetation structure, disturbance and relevant water or soil condition measures; retain each raw observation. Synthetic example: 300 / 500 habitat ha = 60.

B.c Functional biodiversity  Area-weighted monitoring plots meeting a preregistered native functional-group reference band / all eligible plots. Choose pollinator, decomposer, natural-enemy or other locally relevant groups before observing results. Standardize season, effort and detection. Synthetic example: 70 / 100 weighted plots = 70.

Scoring collection and verification

B = 70.0, grade C; weakest 60.0. Ecologists and community monitors collect; independent ecological reviewers verify. Update land-cover alerts during the year and field observations in matched seasons. Do not classify inaccessible plots as healthy.

Additional monitoring and safeguards

Also track conversion outside the contracted area, invasive species, native habitat connectivity, chemical exposure and relevant ecosystem services. City scores require a traceable sourcing footprint and disclose unknown origin. Monitor silvopasture against appropriate pasture and habitat comparators; tree planting must not replace protection of native ecosystems.

P Policy execution

Established practice and limitations

Public Expenditure and Financial Accountability (PEFA) assesses public financial management through a structured framework, including budget reliability and information relevant to service delivery. Use its evidence discipline and documentary trail as a model for ZHF obligations. [7]

Benefit: explicit requirements and evidence reduce reliance on perceptions or policy announcements. Limitation for ZHF: a well-executed budget or formally compliant process does not prove nutrition improvement. Counting minor obligations can conceal failure of one critical authorization.

Recommended component definitions

P.a Due service obligations  Institutional obligations completed to agreed quality by deadline / all obligations due. Fix a common taxonomy and eligible register; do not split easy tasks into many records. Exclude commercial payments scored under C.c. Synthetic example: 80 / 100 obligations = 80.

P.b Cross-agency handoffs  Required interagency handoffs accepted to specification by the receiving agency on time / all handoffs due. Retain sender, receiver, dependency and timestamps; do not count submission alone as acceptance. Synthetic example: 30 / 50 handoffs = 60.

P.c Verified remedies  Eligible cases due for remedy with independently confirmed resolution within the service standard / all eligible cases due. Include grievances and audit findings under a fixed taxonomy; retain severity and unresolved critical cases. Synthetic example: 14 / 20 cases = 70.

Scoring collection and verification

P = 70.0, grade C; weakest 60.0. Agency registrars collect monthly; auditors and community accountability channels independently confirm. Register duties before their deadlines and publish changes. Review access to the grievance channel so that low reporting is not mistaken for few problems.

Additional monitoring and safeguards

Also monitor budget release versus commitment, procurement duration, conflicts of interest, policy reversals and legal barriers. A period with no remedy cases is X, not 100; use a prespecified rolling window for sparse events. Report critical obligations individually. Cross-agency programs need documented legal authority and eligible use of funds, beyond these performance measures.

S Surplus redistribution

Established practice and limitations

The United Nations Environment Programme Food Waste Index establishes measurement for retail, food service and household waste, including explicit boundaries. The World Food Programme (WFP) Food Consumption Score supplies a practical household dietary-frequency measure. [8,9]

Benefit: waste accounting helps separate avoided losses from redistributed food, while consumption measures examine recipients. Limitation for ZHF: tonnage is not nutritional benefit; dietary frequency is not proof of nutrient adequacy. A system that generates less surplus should not be penalized for success.

Recommended component definitions

S.a Need coverage  Eligible unmet meal-service units fulfilled through suitable surplus / all registered unmet units in the defined population and period. Specify age-appropriate nutrition, safety, acceptability and actual receipt. Do not shrink demand to available stock. Synthetic example: 9,000 / 15,000 meal units = 60.

S.b Recovery performance  Safely redistributed edible surplus tonnes / independently estimated eligible edible surplus tonnes available for recovery. Use the same moisture basis; exclude inedible material and ordinary planned commercial supply. Synthetic example: 75 / 100 tonnes = 75.

S.c Delivery quality  Delivered meal-service units meeting safety, nutrition and recipient-acceptability requirements / all delivered units. Keep safety failures and rejected meals visible; receipt is distinct from actual consumption. Synthetic example: 9,000 / 10,000 units = 90.

Scoring collection and verification

S = 75.0, grade B; weakest 60.0. Recovery operators and recipient services record each lot; independent food-safety personnel and recipient samples verify. Reconcile monthly, with safety monitoring at every relevant handling stage.

Additional monitoring and safeguards

Record surplus generated per tonne produced, waste prevented, displacement of local sellers, household consumption and cost per compliant meal. S.a and S.c overlap: test redundancy. If surplus or unmet need is genuinely zero, mark the relevant measure X and withhold the nine-pillar composite. Test a future “food allocation and recovery” definition; do not silently broaden S or encourage excess production to obtain a score.

D Storage

Established practice and limitations

The Logistics Cluster operational guidance recommends physical inventory reconciliation and defined stock-management procedures. This provides an auditable base for measuring usable stock rather than warehouse capacity on paper. [10]

Benefit: reconciled lot records expose discrepancies and double commitments. Limitation for ZHF: a correct count does not demonstrate food safety, retained nutritional quality or readiness to meet demand. Tonnes of different commodities are not nutritionally interchangeable.

Recommended component definitions

D.a Required storage service  Required commodity tonne-days served under the registered storage standard / required tonne-days. Compute separately by commodity, cap daily qualifying service at that day’s requirement, then use fixed requirement weights. Synthetic example: 8,000 / 10,000 tonne-days = 80.

D.b Usable retention  Quality-released closing stock plus compliant issues from the tracked cohort / opening usable stock plus usable receipts for that cohort. Use a common moisture basis and mass balance; quarantined, spoiled or unexplained stock does not qualify. Synthetic example: 950 / 1,000 tonnes = 95.

D.c Reserve readiness  Required complete meal-equivalent days backed by safe, uncommitted, accessible stock / required meal-equivalent days. Use nutrition-constrained meal recipes and demand scenarios; cap at 100 and prohibit duplicate commitments. Synthetic example: 63,000 / 90,000 meal-days = 70.

Scoring collection and verification

D = 81.7, grade B; weakest 70.0. Warehouse teams record lot movements and conditions; independent stock counters and quality laboratories verify. Monitor conditions continuously where warranted, reconcile monthly and conduct unannounced checks.

Additional monitoring and safeguards

Track age profile, cold-chain excursions, aflatoxin or other commodity-specific hazards, nutrient degradation and release lead times. In the proposed Federal Emergency Management Agency (FEMA) and American Strategic Food Reserve Initiative (ASFRI) example, school meals and emergency stocks require separate commitments. Buying excess, drying and mixing it creates usable reserve only after recipe, safety, shelf-life and dispatch validation.

I Infrastructure

Established practice and limitations

The World Bank Rural Access Index measures the rural population within two kilometers of an all-season road. It offers a comparable access baseline; operating records add information on whether the food network actually performs. [11]

Benefit: geographic access identifies excluded communities. Limitation for ZHF: proximity does not establish affordability, route safety, border clearance, freight capacity or reliable delivery. Infrastructure spending and installed capacity are weak substitutes for functioning services.

Recommended component definitions

I.a Delivery reliability  Required consignments received in full, on time and to specification / all consignments due. Register order sizes and deadlines; retain quantity-weighted performance separately to prevent splitting small orders. Synthetic example: 85 / 100 consignments = 85.

I.b Critical service uptime  Required operating hours in which registered processing, energy and communication services meet capacity and quality needs / all required hours. Compute by service, then use fixed baseline requirement weights. Synthetic example: 900 / 1,000 hours = 90.

I.c Effective network access  Residents meeting a registered food-service access standard throughout the relevant season / eligible residents. Assess travel time, availability and affordability against fixed thresholds. Include urban and rural service definitions separately. Synthetic example: 6,500 / 10,000 residents = 65.

Scoring collection and verification

I = 80.0, grade B; weakest 65.0. Operators and transport teams collect monthly records; independent route checks and household sampling verify. Repeat in wet and dry seasons and after outages. Do not report I.c as the official Rural Access Index.

Additional monitoring and safeguards

Also monitor border dwell time, delivered food cost, route redundancy, energy intensity, connectivity outages and maintenance backlog. For regional systems, map mills, ports and corridors shared by multiple countries. A national mean can look healthy while a single common node remains a major failure risk.

E Education and practical capability

Established practice and limitations

The International Labour Organization describes competence through demonstrated ability and competency-based training through assessed outcomes. This supports practical performance assessment instead of counting course attendance. [12]

Benefit: observed tasks are closer to usable capability than certificates alone. Limitation for ZHF: test design and assessor bias matter, and training may not translate into practice when capital, time or equipment is unavailable. A narrow test can miss valuable local knowledge.

Recommended component definitions

E.a Demonstrated competence  Eligible people demonstrating all required task competencies / eligible people. Use a representative sample of the eligible workforce, including non-attendees; assess relevant farming, storage, food-safety or management tasks with a fixed rubric. Synthetic example: 140 / 200 people = 70.

E.b Sustained application  Eligible person-task opportunities performed to standard at follow-up / all eligible person-task opportunities in the sample. Observe workplace or farm practice after three to six months; retain reason codes for nonapplication. Synthetic example: 120 / 200 opportunities = 60.

E.c Timely advisory access  People with independently identified advisory needs receiving qualified, usable support before the relevant decision deadline / all people with those needs. Count unmet needs among people who did not submit requests. Synthetic example: 160 / 200 people = 80.

Scoring collection and verification

E = 70.0, grade C; weakest 60.0. Qualified assessors collect practical observations; independent assessors rescore a blinded subset. Record inter-rater agreement. Assess at baseline, after learning and after a production cycle; monitor advisory service quarterly.

Additional monitoring and safeguards

Report competence and access by gender, age, language, disability and other locally relevant groups. Record demonstrated traditional or tacit skills even without certificates, with consent. Failure to apply a skill is a diagnostic finding, not automatically a failure of the learner: distinguish missing inputs, inappropriate advice and time constraints.

C Capital service

Established practice and limitations

The World Bank Global Findex examines financial access, use and resilience through population surveys. It provides a useful inclusion benchmark beyond formal lending statistics. ZHF needs additional evidence on farm and enterprise cash cycles and suitable working resources. [13]

Benefit: access and resilience questions expose exclusion that bank portfolios miss. Limitation for ZHF: account ownership does not show whether capital arrived before planting or whether terms were suitable. Loan volume and repayment rates alone can reward excessive borrowing or repayment through distress sales.

Recommended component definitions

C.a Timely resource coverage  Verified eligible working-resource need met in time under suitability rules / total verified eligible need. Count grants, retained funds, equity or credit actually available; cap at need and keep one constant-price currency basis. Synthetic example: 600,000 / 1,000,000 units = 60.

C.b Inclusive access  Eligible producers or enterprises with their minimum viable resource need met suitably by the deadline / all eligible producers or enterprises with verified need. Survey non-applicants; freeze eligibility and document exclusions. Synthetic example: 70 / 100 entities = 70.

C.c Timely commercial receipts  Value of valid commercial receivables paid in full by contractual due date / total value due. Use supplier confirmations and transaction records; retain overdue age and payee counts. Exclude public obligations already scored under P. Synthetic example: 450,000 / 500,000 units = 90.

Scoring collection and verification

C = 73.3, grade C; weakest 60.0. Finance and producer teams collect monthly, with pre-planting checks; independent reviewers verify suitability, actual receipt and excluded demand. Define suitability before lending: transparent total cost, voluntary terms, cash-flow fit and applicable protections.

Additional monitoring and safeguards

Also monitor debt-service stress, distress sales, effective transaction costs, settlement time, defaults, complaints and loss protection. No bonus follows from token adoption, financing volume or a named product. If optional SYLO protection is tested, report promised and paid claims, reserve adequacy and basis risk separately; the sponsor’s cap is 20% of crop value, not 20% of loss.

3 Outcomes resilience and additional monitoring

Pillar scores describe capacity and service attainment. Retain the v0.3 outcome gap O and modeled resilience gap H as separate domains. The four outcome observations below should appear as raw prevalence estimates, with uncertainty and dates, before any composite is displayed.

Measure and source	Benefit	Limitation and ZHF response
Food insecurityFAO Food Insecurity Experience Scale [14]	Experience-based food access, with statistical validation and international calibration.	Requires representative sampling, language adaptation and scale checks. It does not measure nutrients or establish why access changed.
Healthy-diet unaffordabilityWorld Bank Food Prices for Nutrition [15]	Connects local food prices to dietary requirements and available resources.	Least-cost availability differs from actual purchasing and consumption. Preserve income, nonfood-need and price assumptions.
Stunting and wastingUNICEF WHO World Bank methods [16]	Standardized child anthropometry supports comparison and identifies serious deprivation.	Different time horizons; health and sanitation also affect growth. Use trained measurement, appropriate age groups and season; avoid quarterly extrapolation.
Dietary frequencyWFP Food Consumption Score [9]	Practical short recall period for household monitoring.	Household frequency does not establish individual nutrient adequacy. Use as supplementary monitoring, not a substitute for the above.

Candidate O = (food insecurity gap + diet unaffordability gap + nutrition gap) / 3; nutrition gap = (stunting gap + wasting gap) / 2. Normalize each with registered targets and worse anchors before combining. Do not equate raw stunting and wasting percentages with the same severity. These choices require calibration and sensitivity tests; no validated global ZHF anchors are yet available.

H = mean percentage of required nutritional-service person-days unmet across the two v0.3 reference scenarios: a 20% inflow reduction for 30 days and a seven-day outage of the largest critical node during a 30-day horizon. Define demand, season, substitution, concurrent commitments and recovery. These are proposed comparison scenarios, not estimated event probabilities. Add local drought, flood, conflict and price scenarios separately; validate simulations against drills and actual disruptions.

Cross-cutting monitoring must cover: equitable access; real net producer and household income; full delivery and transition costs; nutrition and food safety; employment and labor protections; tenure and displacement; ecosystem leakage; market prices; fiscal burden; privacy and digital exclusion. Report benefits, harms and distribution. These measures may alter decisions even when they do not enter the headline score.

4 From pillar grades to a master gap

Divide each pillar achievement A by 100 to obtain r, w, b, p, s, d, i, e and c. Retain the candidate v0.3 relationship for testing:

g₁ = (r × w × b)^(1/3)       g₂ = (p × s)^(1/2)g₃ = (d × i)^(1/2)              g₄ = (e × c)^(1/2)Q = (3g₁ + 2g₂ + 2g₃ + 2g₄) / 9Capacity gap K = 100 × (1 − Q)Candidate master gap G = (O + K + H) / 3

Multiplication represents proposed complementarity within groups; roots keep group means on a common scale. Addition permits partial trade-offs across groups. Division forms means, not a causal law. A true zero eliminates its group; missing data must not be treated as zero. The 3:2:2:2 weights yield equal marginal pillar influence when all achievements are equal and positive. None of these relationships has been empirically established for ZHF. [1]

Worked synthetic calculation

Using the nine pillar examples: R 70.0; W 80.0; B 70.0; P 70.0; S 75.0; D 81.6667; I 80.0; E 70.0; C 73.3333. Calculate with unrounded values. The group means are approximately 0.7319, 0.7246, 0.8083 and 0.7165; Q = 0.7438, so K = 25.6.

Suppose separately calibrated outcome inputs are food insecurity gap 45, diet unaffordability gap 55, stunting gap 30 and wasting gap 10. Nutrition gap = 20 and O = 40. Suppose unmet service is 18% and 22% in the reference scenarios, giving H = 20. Then G = (40 + 25.6 + 20) / 3 = 28.5. This demonstrates arithmetic only; the outcome numbers are normalized example gaps, not actual prevalence or approved anchors.

If a comparable baseline one year earlier had O = 42, K = 30 and H = 24, its G was 32.0. Gap closure is 3.5 points per year; O improves by 2, K by 4.4 and H by 4 points. Use elapsed years for unequal intervals. Do not describe these changes as statistically established without uncertainty estimates. Zero indicates the registered benchmarks, not proof that every form of hunger has disappeared.

Outcome alert: if O or any raw hunger or nutrition outcome worsens, display that change beside G even when G improves. State uncertainty and affected groups. Capacity gains cannot justify describing worsening outcomes as demonstrated hunger reduction.

Aggregation across places

Pool compatible numerators and denominators, or use the appropriate survey estimators, then recompute scores. If district A serves 700 of 1,000 residents and B serves 2,400 of 3,000, combined coverage is 3,100/4,000 = 77.5%, not 75%. Never average ranks, overlapping populations or incompatible hectares and people. For composite pillars, aggregate each submeasure first.

Use residence-based outcomes and a separately traced food-sourcing footprint for environmental measures. A city’s lack of farmland is not a zero regeneration score. Reconcile shared reserves and regional trade flows; run regional shock simulations on the network, not on averaged national resilience scores. Publish unknown sourcing shares and withhold comparisons when they are material.

5 Example executive summary report

SYNTHETIC DEMONSTRATION — no country results. Geography: Pilot Area A. Period: Year 1; baseline: Year 0. Method: proposed measurement guide v0.1. Evidence: illustrative inputs only. Field confidence intervals, assurance and ranking are not available.

Headline: candidate gap 28.5, down from 32.0 over one year. Observed-outcome gap 40.0, capacity gap 25.6 and modeled resilience gap 20.0. All three improve in this illustration. These are descriptive changes; there is no causal attribution. Field reports must replace this text with raw outcomes, uncertainty and observation dates.

Pillar	Achievement	Grade	Weakest	First diagnostic focus
R Regeneration	70.0	C	60.0	Soil carbon condition
W Water	80.0	B	70.0	Safe domestic service
B Biodiversity	70.0	C	60.0	Habitat condition
P Policy	70.0	C	60.0	Cross-agency handoffs
S Redistribution	75.0	B	60.0	Unmet meal-service demand
D Storage	81.7	B	70.0	Reserve readiness
I Infrastructure	80.0	B	65.0	Effective network access
E Education	70.0	C	60.0	Sustained skill application
C Capital	73.3	C	60.0	Timely resource coverage

Required interpretation beneath the scorecard

Safeguards: report verified incidents, affected people or area, containment status and owner. In this synthetic example, safeguard status is not assessed; do not replace that with “no incidents.” Equity: show at least the most deprived relevant subgroup, its denominator and the gap from the overall result. Missingness: report the number of current core measures out of 27 and the unsampled share of each denominator.

Diagnostic priorities: examine capital timing and interagency handoffs as possible constraints on practical skill use and reserve replenishment. This is a hypothesis, not a conclusion from the grades. Urgent safety or deprivation needs take precedence. Each proposed response should name an accountable owner, predecessor task, evidence, expected outcome, earliest feasible date, full cost and funding status. The dashboard flags; authorized people select and schedule.

Attach the audit and evaluation annex

Include all 27 component records: numerator, denominator, raw observations, score, baseline, change, uncertainty, source, observation date, verifier and version. Add outcome estimates, model assumptions, incident register, subgroup results and spending against milestones. Provide a reproducible calculation snapshot and change log. Separate observed associations from causal evaluation findings and distinguish committed, spent and still-unfunded resources.

6 Evaluation adoption and revision

Preparation and operating cadence

During the six-to-eight-month preparation period, first agree boundaries, eligible populations, service definitions and independent oversight. Map existing statistics and collect only the additional data needed. Field-test instruments in contrasting sites in both regional groups, validate translations and assessor reliability, then approve a frozen protocol and baseline. Budget for household sampling, field ecology, laboratories, independent audits and secure data management—not only data entry.

During three years of operation, reconcile operational records monthly, review the pillar profile and dependencies quarterly, and collect outcomes and seasonal ecological measures on their registered schedules. Use the final six months to complete verification, causal analysis, reproducible archives and public reporting. Some soil, biodiversity and child-growth effects may require follow-up beyond the grant period; do not force biological change to fit the reporting calendar.

Test what the measures and interventions actually establish

Measurement validation: test repeatability, assessor agreement, missingness, denominator gaming, construct coverage and equivalent meaning across locations. Estimate the cost of measurement. Check whether correlated submeasures repeatedly reward the same service, especially S.a with S.c, E.a with E.b and overlapping policy or payment obligations. Preserve the raw components when simplifying.

Causal evaluation: preregister primary outcomes and minimum meaningful effects. Where feasible randomize a phased rollout across enough independent implementation units; otherwise use a justified comparison design with baseline trends, explicit confounding assumptions, spillover analysis and robustness checks. Ten participating countries alone do not provide a well-powered cluster trial. Determine sample size from the intervention level, expected effect and intracluster correlation.

Catalytic effects: compare a bundle with its components and a credible counterfactual. For outcomes Y under a factorial design, interaction = Y_AB − Y_A − Y_B + Y_0. This interaction is not identifiable merely because two indicators rise together. Include full costs, harms, uncertainty and timing. Test whether a low pillar score predicts an actual binding constraint before prioritizing solely on its rank.

Practical models to test

Hamara use case: link grain purchase, feed conversion, animal performance, farmer net cash and household diet records. Test whether better feed access and timely payment increase net income after feed, labor, debt and production risk. FEMA and ASFRI use case: combine a lot-level mass balance, recipe constraints and competing school/emergency demand in a reserve simulation. Compare alternatives on safe meal coverage and full cost. Regional portfolio use case: optimize suitable crops and livestock against nutrition, water, habitat, labor, income and seasonal demand constraints, then validate on actual production and market access.

Revision and release: IAAIR maintains deterministic calculations and versioned methods; WFB coordinates operational correction; independent reviewers assess validity and claims. Local and Indigenous knowledge may generate hypotheses under consent and appropriate governance. Language-model agents may retrieve evidence, never manufacture observations or silently change weights. Publish null and adverse findings. Release comparative ranks only after independent review of completeness, uncertainty, equivalence and sensitivity.

7 Sources for the measurement specification

Primary institutional methods reviewed October 23 2026. Numbered references support the descriptions of established practice; the proposed ZHF adaptations, grades, examples and critique are recommendations in this guide. Use the source version archived at protocol registration. Listing an organization does not imply endorsement of ZHF.

[1] Organisation for Economic Co-operation and Development, European Union and European Commission Joint Research Centre. Handbook on Constructing Composite Indicators. 2008.https://doi.org/10.1787/9789264043466-en

[2] FAO. SDG indicator 2.4.1 Agricultural sustainability. Official indicator and methodology resources.https://www.fao.org/sustainable-development-goals-data-portal/data/indicators/Indicator2.4.1-proportion-of-agricultural-area-under-productive-and-sustainable-agriculture/

[3] FAO. Global Soil Organic Carbon monitoring reporting and verification protocol. 2020.https://www.fao.org/fileadmin/user_upload/GSP/MRV/GSP_MRV_Protocol_A4_EN_002.pdf

[4] World Health Organization and United Nations Children’s Fund Joint Monitoring Programme. Drinking water service ladder.https://washdata.org/topics/drinking-water

[5] FAO. Progress on the level of water stress. SDG indicator 6.4.2 report. 2024.https://openknowledge.fao.org/items/06b38388-ae5d-4366-97a8-8a02708e3ab9

[6] United Nations. System of Environmental Economic Accounting Ecosystem Accounting. Methodology resources.https://seea.un.org/en/methodology/ecosystem-accounting

[7] PEFA Secretariat. Public Expenditure and Financial Accountability Framework. 2016.https://www.pefa.org/resources/pefa-2016-framework

[8] United Nations Environment Programme. Food Waste Index Report 2024.https://www.unep.org/resources/publication/food-waste-index-report-2024

Sources continued

[9] World Food Programme. Food Consumption Score. Vulnerability Analysis and Mapping Resource Centre. 2026 guidance page.https://vamresources.manuals.wfp.org/docs/food-consumption-score

[10] Logistics Cluster. Monitoring and Counting. Logistics Operational Guide.https://log.logcluster.org/en/monitoring-and-counting

[11] World Bank. Rural Access Index. Data Catalog.https://datacatalog.worldbank.org/search/dataset/0038250/rural-access-index-rai

[12] International Labour Organization. Glossary of Skills and Labour Migration. Competence and competency-based training entries. 2021.https://www.ilo.org/resource/glossary-skills-and-labour-migration

[13] World Bank. Global Findex Database 2025. Report and methods.https://www.worldbank.org/en/publication/globalfindex/report

[14] FAO. Food Insecurity Experience Scale data analysis. Statistical validation and calibration resources.https://www.fao.org/measuring-hunger/access-to-food/fies-data-analysis/en

[15] World Bank. Food Prices for Nutrition DataHub. Cost and affordability methods.https://www.worldbank.org/en/brief/2022/03/01/foodpricesfornutrition

[16] United Nations Children’s Fund, World Health Organization and World Bank. Joint Child Malnutrition Estimates standard methodology. 2024.https://data.unicef.org/resources/jme-standard-methodology/

Design documents reviewed

Zero Hunger Formula Framework v0.3 Research and Pilot Edition and Zero Hunger Formula Implementation Handbook v0.3 Research and Pilot Edition. These establish the nine pillars, candidate aggregation and sponsor use cases. They are design sources, not independent evidence of program effects.