Assessing needs: data, severity and priorities

1Needs assessment in the humanitarian programme cycle

The humanitarian programme cycle (HPC) links five steps: needs assessment and analysis, strategic planning, resource mobilisation, implementation and monitoring, review and evaluation. The first collective product is the Humanitarian Needs Overview (HNO), which feeds the response plan (since 2024, most countries publish a single document, the HNRP). An NGO that designs a programme without reading its country's HNO deprives itself of the best available synthesis and risks proposing what others already cover.

The type of assessment depends on the phase of the crisis:

Three notions must be distinguished. A need is the gap between the observed situation and a standard (for example 15 litres of water per person per day according to Sphere). Vulnerability is the likelihood of suffering harm from the shock, which depends on exposure and capacities. Capacities are the resources of people, communities and institutions. A good diagnosis describes all three; a list of gaps without capacities leads to responses that replace what was already working.

Golden rule: every question in an assessment must serve a decision identified in advance. An "interesting" question that changes no decision costs respondents time and dilutes the analysis.

2Secondary data: start with what exists

The secondary data review (SDR) gathers what is already known before collecting anything: HNO, cluster reports, demographic and health surveys (DHS), IPC analyses, other NGOs' assessments, data from government technical services, displacement tracking (IOM DTM), market prices, protection reports. A well-done SDR halves the size of a primary data collection, because it lets you ask only the questions that remain unanswered.

Each source is rated on two dimensions. Source reliability: known method, described sample, author without a conflict of interest. Information credibility: confirmed by other sources, consistent with what is known, recent. Information from a reliable source but two years old in an area of massive displacement is weak; information from an average source confirmed by three others is solid.

The SDR ends with a gap analysis: for each decision question, what is known, with what level of confidence, and what remains to be collected? It is this table, not a list of topics, that justifies the questionnaire.

• Source triangulation: the same information confirmed by independent sources.
• Method triangulation: a household survey and focus groups that converge.
• Investigator triangulation: different teams obtaining the same results.

3Primary data: choosing methods and controlling bias

Four methods complement each other. The key informant interview (KII) questions people who have an overview: village chief, nurse, teacher, trader, head of a women's association. Quick and cheap, it gives trends but not proportions. The focus group discussion (FGD), 6 to 10 homogeneous people (same sex and age group), explores perceptions, norms and preferences. Direct observation (latrines, markets, queues) verifies what is reported. The household survey on a representative sample is the only method that produces generalisable proportions, provided that sampling is rigorous (week 6).

The most frequent biases:

• Selection bias: interviewing only villages accessible by road, or households present during the day (missing daily workers).
• Courtesy and aid-expectation bias: people exaggerate or minimise according to what they think the enumerator wants to hear or the aid they hope for. It is reduced by clearly explaining that no aid is promised and by favouring observable indicators.
• Enumerator bias: leading wording, a man questioning women about violence. Train, pre-test the questionnaire and form mixed teams.
• Survivorship bias: in an area of high mortality or departure, the most affected households are no longer there to answer.

Minimum ethics: informed consent, the possibility to refuse without consequence, no sensitive question without a referral protocol (gender-based violence, child protection), and personal data reduced to the strict minimum.

4Measuring severity: explicit rules and people in need

Comparing areas requires a common severity scale. The Joint Intersectoral Analysis Framework (JIAF) and the Integrated Food Security Phase Classification (IPC) both use five phases, from 1 (minimal) to 5 (catastrophic). The IPC classifies an area according to a 20% rule: the area is placed in the most severe phase that, together with the higher phases, concerns at least 20% of its population.

Example. Population of 100,000 people: phase 1: 35%; phase 2: 30%; phase 3: 22%; phase 4: 10%; phase 5: 3%. Phases 4 and 5 combined: 13%, below the threshold. Phases 3, 4 and 5 combined: 35%, above 20%. The area is therefore classified in phase 3. The number of people in need of urgent action (phase 3 or above) is 35,000.

The number of people in need (PIN) of a sector is not the total population of the area: it is the sum of people whose situation reaches the defined severity threshold (often phase 3 or above). Adding up the PIN of several sectors counts the same person several times. For the intersectoral figure, the maximum per area and per group is therefore used, or an analysis at household level.

Severity must be disaggregated by sex, age and disability. For disability, the Washington Group Short Set asks six questions (seeing, hearing, walking, remembering, self-care, communicating); a person is counted as living with a disability if they answer "a lot of difficulty" or "cannot do at all" to at least one question.

5Protection, gender and market analysis

An assessment that looks only at material gaps misses the risks. Protection analysis identifies threats (who does what to whom), people's vulnerabilities to these threats, and their capacities: risk = threat × vulnerability / capacity. It identifies groups excluded from aid (people without documents, minorities, isolated older people) and the risks that aid itself can create: distributions that expose people to theft, dangerous journeys, tensions between displaced people and the host population.

Gender analysis describes who decides, who controls resources, who has access to services, and how the crisis changes these roles. For example, it avoids paying cash assistance to a household member who does not control its use, or setting distribution times incompatible with women's tasks.

Market analysis determines whether cash transfers are feasible: are markets functioning, can traders respond to an increase in demand without inflation, is physical and financial access safe? The reference tools (rapid market assessment, emergency market mapping and analysis) answer three questions: availability of goods, prices and their trends, restocking capacity.

Do no harm: for each response option, write at least one risk it creates and the measure that reduces it.

6From analysis to decision: response analysis

Response analysis links what is known about needs to what the organisation will do. It compares options (in-kind, cash, vouchers, services, advocacy) on explicit criteria: relevance to priority needs, people's preferences, feasibility (access, markets, security, capacities), cost-effectiveness, risks, and what other actors are already doing (3W or 4W matrix: who does what, where and when).

A simple method is the weighted matrix: each criterion receives a weight set before scoring the options, to avoid choosing first and justifying afterwards. For example, relevance 30%, preferences 20%, feasibility 20%, cost-effectiveness 20%, risks 10%. An option scoring 4/5 for relevance, 5 for preferences, 3 for feasibility, 4 for cost-effectiveness and 3 for risks totals 0.3 × 4 + 0.2 × 5 + 0.2 × 3 + 0.2 × 4 + 0.1 × 3 = 3.9.

The final decision must also take into account the organisation's added value: presence, acceptance, expertise, local partners. If an actor already adequately covers a need in an area, the best decision is often to go elsewhere or to complement, not to duplicate.

The final product is an analysis note of a few pages: priority needs by area and by group, severity, coverage gaps, compared options, recommendation and level of confidence. This is what management will read, not the database.