Strategic MEAL: accountability, evidence and decisions

1MEAL as a decision system

MEAL (monitoring, evaluation, accountability, learning) is not a reporting function. For a senior manager, it answers three questions: are we doing what we promised? (monitoring), is it really changing people's situation? (evaluation), do people have a say and are they safe? (accountability). Learning links all three to decisions.

A proportionate MEAL system generally devotes 3 to 10% of a project's budget to monitoring and evaluation, depending on its size and complexity. It is formalised in a MEAL plan: an indicator tracking table (definition, method, frequency, person responsible, disaggregation), a survey calendar, feedback mechanisms, an evaluation plan and data protection rules.

Relevance test: for every piece of data collected, you must be able to name the decision it informs and the person who will make it. Data that informs no decision costs the teams and the people interviewed time, and increases data protection risks.

2Indicators, sampling and data quality

A good indicator is valid (it measures what it claims to measure), reliable (two enumerators get the same result), sensitive (it changes when the situation changes), disaggregated (sex, age, disability using the Washington Group questions) and realistic to collect. For food security, standardised indicators (food consumption score, reduced coping strategies index) are preferred so that results can be compared with other actors.

For a simple random sample survey, the sample size is calculated as n = z² × p × (1 − p) / e². With a 95% confidence level (z = 1.96), an unknown prevalence (p = 0.5) and a 5% margin of error (e = 0.05): n = 3.8416 × 0.25 / 0.0025 ≈ 384. With cluster sampling, multiply by the design effect (often 1.5 to 2); then add a margin for non-response (for example 10%). With a design effect of 2 and 10% non-response: 384 × 2 / 0.9 ≈ 854 households.

Data quality is controlled at every step: training and testing enumerators, a questionnaire piloted in the local language, automatic checks in the tool (KoboToolbox, ODK), re-checking a sample of interviews (back-checks), daily analysis of outliers and of abnormally short interview durations.

3Evaluating: criteria, attribution and contribution

Evaluations most often use the six revised OECD DAC criteria (2019): relevance, coherence, effectiveness, efficiency, impact and sustainability. For humanitarian action, ALNAP proposes complementary criteria such as coverage, connectedness and coordination.

The hardest question is causality: is the observed change due to the project? Three families of approaches:

• Experimental approaches (randomised trials): they establish attribution with the greatest rigour, but are expensive, raise ethical issues in emergencies (you do not deny aid to a control group in need) and do not always explain why something works.
• Quasi-experimental approaches (difference-in-differences, propensity score matching, regression discontinuity): they use natural comparisons, for example households just above and just below an eligibility threshold.
• Theory-based approaches (contribution analysis, process tracing, outcome harvesting): they check whether the planned causal chain actually happened and rule out alternative explanations.

Worked difference-in-differences example. The average food consumption score rises from 28 to 41 among beneficiary households, and from 30 to 36 in a comparison group. The estimated effect is (41 − 28) − (36 − 30) = 13 − 6 = 7 points, not 13: part of the improvement is explained by the harvest season.

4Accountability to people and the Core Humanitarian Standard

The Core Humanitarian Standard (CHS), revised in 2024, sets nine commitments to affected people and communities: assistance that is appropriate, effective and timely, that strengthens local capacities and avoids negative effects, based on communication, participation and feedback, with complaints handled, coordination, learning, competent and well-managed staff, and resources managed responsibly.

An effective complaints and feedback mechanism offers several channels (hotline, boxes, help desks, committees, digital channels), adapted to people who cannot read, older people and people with disabilities; it separates sensitive complaints (sexual exploitation and abuse, fraud, corruption), which are handled confidentially through a dedicated channel; and it closes the loop by responding to people within an announced timeframe.

Useful management indicators for a senior leader: rate of responses within the deadline, share of complaints that led to a programme change, share of feedback coming from women and people with disabilities, and time taken to handle sensitive complaints.

Warning sign: a mechanism that receives almost no complaints is not proof of a perfect programme. Most often it is proof that people do not know about it, cannot access it or do not trust it.

5From data to organisational learning

Learning happens when evidence changes decisions. The most effective tools are simple: quarterly progress reviews centred on three questions (what works, what does not work, what are we changing?), after-action reviews at the end of each emergency phase, a management response to the recommendations of every evaluation (accepted, partially accepted, rejected, with an action plan and a person responsible), and follow-up of those plans.

Adaptive management formalises this cycle: explicit assumptions, leading indicators, scheduled decision points, and room for manoeuvre negotiated with the donor to change activities without a heavy amendment. Several donors explicitly encourage it in unstable contexts.

Finally, data protection is a leadership obligation: minimising the data collected, informed consent, encryption, restricted access, retention periods, and an impact assessment before sharing beneficiary lists with third parties (authorities, payment providers). The ICRC Handbook on Data Protection in Humanitarian Action (2nd edition, 2020) is the sector's reference.