GCRI · Disaster Resilience and Recovery · Foundations
How anticipatory action is measured and modelled: triggers, evaluations and data
Anticipatory action depends on two kinds of method: models that decide when to act on a forecast, and evaluations that test whether acting early helped. This article explains how triggers are designed and verified, how effects are measured with quasi-experiments and randomised trials, how costs and benefits are estimated, how global counts are compiled, and where the uncertainty lies.
- As of
- 28 October 2026
- Ukufunda
- 13 minutes
- Ubungqina
- 13 sources, cited in order
- Indlela yokusebenza
- A review of the methods literature on forecast-based financing and action-based forecasting, two household-level evaluations (a quasi-experiment in Bangladesh and randomised trials in Bangladesh, Nepal and Niger), agency learning frameworks, WMO guidance on impact-based forecasting, the data methodology of the Anticipation Hub global overview and published benefit-cost estimates.
- Idatha
- The trial and study results come from their published papers and, for the Niger trial, a World Bank reproducibility package whose data are due to be released through the World Bank Microdata Library; global counts come from the Anticipation Hub overview for 2025, which describes its collection method; cost comparisons from the Global Assessment Report 2025 and FAO.
Methods and evidence
- Umbuzo
- How are anticipatory action triggers designed and verified, how are the effects of acting early measured, and how reliable are the global figures and cost estimates used to describe the practice?
- Approach
- A review of the methods literature on forecast-based financing and action-based forecasting, two household-level evaluations (a quasi-experiment in Bangladesh and randomised trials in Bangladesh, Nepal and Niger), agency learning frameworks, WMO guidance on impact-based forecasting, the data methodology of the Anticipation Hub global overview and published benefit-cost estimates.
- Ubungqina
- 13 sources, cited in order and listed under References
- Data availability
- The trial and study results come from their published papers and, for the Niger trial, a World Bank reproducibility package whose data are due to be released through the World Bank Microdata Library; global counts come from the Anticipation Hub overview for 2025, which describes its collection method; cost comparisons from the Global Assessment Report 2025 and FAO.
- Imida
- Rigorous evaluations are few and concentrate on cash transfers ahead of floods and droughts in South Asia and the Sahel; effects measured over weeks or months may not persist; global counts are voluntary, self-reported minimum totals; benefit-cost ratios depend on assumptions. The article is not guidance for any specific trigger or programme.
01Ubungakanani bomsebenzi
What measuring anticipatory action involves
Two families of method: deciding when to act, and testing whether acting helped.
Anticipatory action methods are the techniques used to decide when to act on a forecast and to measure whether acting early reduced harm. Anticipatory action itself is acting ahead of a predicted hazardous event to prevent or reduce its impacts before they fully unfold, using activities, triggers and funding agreed in advance 7.
The first family of methods is ex ante: designing and verifying triggers. A 2015 paper setting out forecast-based financing proposed matching threshold forecast probabilities with appropriate actions and releasing funds when those thresholds are reached 1. The second family is ex post: evaluating effects on households and comparing costs and benefits. The UN Office for the Coordination of Humanitarian Affairs (OCHA) builds both into its frameworks through a learning plan with a quantitative evaluation of household welfare, a qualitative evaluation of people's experience and an evaluation of trigger performance 2.
This article complements the platform's foundational article on anticipatory action and its anticipatory action topic on the Disaster Resilience and Recovery platform. Forecasting science is covered on the Climate and Hazards platform's climate services and warnings topic.
02Stakes
Why the methods matter
Money released on a forecast needs rules that can be defended, and evidence that it works.
Anticipatory action moves money before the hazard is certain. In 2025 there were 146 activations in 54 countries, releasing US$119.9 million, and 262 active frameworks held about US$292 million in pre-agreed financing 7. The Secretary-General reports that prearranged financing through OCHA-managed pooled funds has passed US$160 million 8.
Decisions on that scale need defensible rules and tested evidence. The forecast-based financing literature identifies uncertainty about which forecast probabilities are worth acting on, and the discomfort of donors with funding events that are likely but not certain, as central barriers 1. Evaluations answer the other question donors and governments ask: whether help delivered early does more good than the same help delivered later 4.
Laws now encode these methods. The Philippine imminent disaster law requires a pre-disaster risk assessment, a lead time of three days extendable to five, and disbursements graduated to the forecast severity 9. The platform's methods and uncertainty page records how such evidence is used.
03Triggers
Designing a trigger
Thresholds, lead times and the willingness to act in vain.
A trigger states the forecast variable, the threshold, the lead time and the area. The Anticipation Hub summarises the requirements: a viable trigger must be verifiable, data-driven, transparent, timely and operationally feasible, built on historical and forecast data, governed by clear protocols and designed to balance lead time against forecast accuracy 7. It notes that trigger development is also influenced by donor requirements 7.
The action-based forecasting method makes the trade-offs explicit. In a 2016 study applied to a Red Cross forecast-based financing pilot in Uganda using the Global Flood Awareness System, the authors verified forecasts against each action's parameters, such as how long the action remains useful, and linked forecasts to action through the magnitude of past floods and the willingness to act in vain for each specific action 3. They defined a danger level of flooding and selected probabilistic triggers suited to each action, and argue that the method can be applied across hazards 3.
As a teaching example only: an action that stays useful for a month, such as distributing water purification tablets, can tolerate a lower probability threshold than one that is costly and irreversible, such as evacuation. The verification statistics that matter are the share of past events the trigger would have caught and the share of activations that would have been in vain.
04Impact-based
From hazard forecasts to impact-based triggers
Why triggers increasingly use forecast impacts, not hazard values alone.
A hazard forecast says what the weather will be; an impact-based forecast says what it will do. The World Meteorological Organization (WMO) guidelines on multi-hazard impact-based forecast and warning services argue that the focus should evolve from what the weather will be to what the weather will do, by including the vulnerability and exposure of people and assets 10. They note that an accurate and timely warning does not by itself guarantee safety 10.
Triggers follow the same path. The Philippine law bases declarations on a pre-disaster risk assessment that classifies forecast impacts as severe, considering local risk profiles, institutional capacity and the potential for cascading hazards 9. Combining a hazard threshold with exposure and vulnerability data narrows action to the places where the forecast hazard is likely to cause harm.
Where action happens is part of the design. The Anticipation Hub reports that 103 activations in 2025 took place in areas predetermined in a framework, and that 190 active frameworks identified predetermined areas while 4 planned to select areas only after a trigger is reached 7. Predetermined areas make prepositioning easier but can miss communities outside them 7.
05Harmonization
Harmonizing triggers and new forecast tools
Why several agencies acting on different triggers is a problem, and how it is being solved.
When several organisations run their own triggers for the same hazard and place, they can act at different times on different information. The Anticipation Hub identifies the harmonisation of triggers as a recurring theme of 2025, with triggers identified jointly and transparently and common arrangements for monitoring and activation 7.
Two models are emerging. In Sri Lanka, partners including the Food and Agriculture Organization of the United Nations (FAO), the national Red Cross Society and the World Food Programme (WFP) are aligning the triggers in flood frameworks under development, with experts from several government institutions 7. In Mozambique, several organisations acted ahead of floods on a government advisory that uses harmonised triggers, with the National Institute for Disaster Risk Management and Reduction (INGD) declaring activation once thresholds were reached 7.
Forecast tools are changing too. The Hub reports that triggers continued to use artificial intelligence and machine learning in 2025; in Nigeria, an OCHA-led flood framework combines forecasts from the Global Flood Awareness System with a machine-learning flood forecast, setting thresholds for streamflow from both models to reduce the chance of missing events 7.
06Uvavanyo
Measuring effects: from quasi-experiments to trials
How evaluations have moved from comparisons to randomised designs.
The first household evidence came from comparison groups. A quasi-experimental study of the 2017 Bangladesh floods compared 1,039 households that received an unconditional grant of about US$60 before the early flood peak, released on a forecast and pre-defined trigger, with similarly vulnerable flood-affected communities that did not, using a post-disaster household survey and qualitative data 6. It found better access to food, less high-interest debt and less psychosocial stress, but no sustained benefit after a second flood peak 6.
Quasi-experiments are vulnerable to differences between the groups that the design cannot remove. Randomized trials address this by assigning timing at random. The key design choice is the comparison: early help against no help measures the value of the transfer, while early help against the same help delivered later measures the value of timing, which is the question anticipatory action has to answer 4.
Agency programme results add breadth but not rigour. WFP reports, for example, that in Ethiopia cash combined with early warnings was more effective than cash alone and reduced harmful coping strategies by 28 per cent 11.
07Trials
Randomized trials: what timing does
Two recent trials that compare early and later help.
A randomized trial in Bangladesh and Nepal by researchers from the World Bank, WFP and the University of Chicago assigned 300 villages in Bangladesh and 138 in Nepal to receive a one-time transfer of the same value either within days of the flood trigger or about one to one and a half months after the peak 4. In Bangladesh, the early group received transfers by mobile money two days before the flood peak 4.
The trial measured outcomes in three survey rounds. In the short run, early recipients had a food consumption score 1.1 points (2.4 per cent) higher, were 4.2 percentage points more likely to reach acceptable food security, had a reduced coping strategies index 6.8 per cent lower, depression scores 0.101 standard deviations lower and life satisfaction 0.179 standard deviations higher 4. Outcomes converged in later rounds, so the authors conclude that total benefits were higher for the early group; the paper is a working paper, not yet peer-reviewed 4.
A World Bank trial in Niger used satellite-based drought triggers to compare large transfers paid before the lean season, a conventional humanitarian response during the lean season and smaller regular transfers through the year 5. Large early transfers produced greater net gains in economic welfare and psychological well-being before and during the lean season, but none of the differences persisted after the lean season or nine months later, and early timing changed borrowing without a discernible effect on livelihoods 5.
08Ezoqoqosho
Costs, benefits and their assumptions
How benefit-cost ratios are built and why they vary.
Benefit-cost estimates compare the value of avoided losses with the cost of acting, including activations in vain. The UN Office for Disaster Risk Reduction (UNDRR), in the Global Assessment Report 2025, cites projects under Mozambique's national drought framework with a benefit-cost ratio of about 2.25, monetised benefits of US$99 per household against a cost of US$44.19 12. FAO states that every dollar invested in anticipatory action can generate seven dollars in benefits for rural families 13.
Such figures depend on what is counted. Benefits may include avoided asset sales, livestock deaths, debt and lost income, but may omit effects on mental health, which the randomised trial measured directly 4. Costs depend on how often a trigger fires in vain over a period, which follows from the threshold choices described above 3.
Trial results can feed better estimates. Measured short-run gains with later convergence, as in Bangladesh, Nepal and Niger, suggest that the value of anticipatory action lies largely in reducing suffering and harmful coping during the shock rather than in lasting income gains 4,5. Estimates that assume permanent effects would overstate the benefits.
09Idatha
Where the global data come from
How activations, frameworks and funding are counted, and what the caveats mean.
The main global dataset is the Anticipation Hub's annual overview. For 2025 it collected data from members of the Anticipatory Action Task Force, an online survey promoted to other organisations and members of the Grand Bargain secretariat; entries were checked for duplication, returned to submitters for verification and compared with regional data, and unverifiable entries were excluded 7.
The Hub states the limits plainly. Reporting is voluntary, so some frameworks and activations are likely missed, especially locally led and informal initiatives, and the figures should be cited as minimum totals 7. The method has been used since the 2023 report but differs from the first, 2022 report, and more organisations submit data each year, so part of the growth in activations and frameworks likely reflects reporting rather than practice 7. Whether people covered only by early warnings are counted as reached is left to each organisation 7.
Other data are softer still. The Hub's data on institutionalisation come from a survey of country and regional experts and are described as indicative, subjective and non-exhaustive 7. Donor data on build and fuel funding are self-classified by donors 7. UN reports give separate counts, such as 29 UN-supported frameworks in 23 countries in May 2026 8.
| Umthombo | Design | Main limitation |
|---|---|---|
| Action-based flood forecasting, Uganda 3 | Trigger verification against past floods | Single pilot; depends on flood records |
| Bangladesh 2017 floods study 6 | Quasi-experiment, 1,039 treated households 6 | Comparison groups not randomised |
| Bangladesh and Nepal trial 4 | Randomized timing, 438 villages 4 | Working paper; floods only |
| Niger drought trial 5 | Randomized timing and size of transfers | Effects did not persist |
| Anticipation Hub global overview 7 | Voluntary self-reported survey | Minimum totals; method changes |
10Izixhobo
Guidance and instruments for anticipatory action methods
Learning frameworks, forecasting guidance, legal rules and reproducibility.
There is no international standard for trigger design or evaluation. Practice draws on agency learning frameworks, such as OCHA's learning framework of independent evaluation, process learning, and monitoring and evaluation 2, on WMO guidance for impact-based forecasting 10 and on the methods literature 1,3.
National law sets binding rules where it exists: the Philippine act defines lead time, the pre-disaster risk assessment, graduated disbursement and the handling of funds when the hazard does not occur 9.
Reproducibility is becoming part of the evidence. The Niger trial's reproducibility package was verified by a World Bank analytics team, and its data are due to be released through the World Bank Microdata Library 5. The Bangladesh and Nepal trial is registered in a public trial registry 4.
11Frontiers
What is changing in the methods
Four areas where methods are moving.
Harmonized and government-led triggers. Joint thresholds and national meteorological leadership reduce conflicting activations 7.
Machine-learning forecasts in triggers. Combined thresholds from conventional and machine-learning models aim to catch more events 7.
Timing trials beyond floods. Randomized comparisons now cover drought as well as floods 5,4.
Better accounting of reach. The Hub aims to report people covered by early warnings more consistently 7.
12Umanyano
Anticipatory action methods across the Consortium
Evidence, governance and finance for acting on forecasts.
Within the Consortium, The Global Centre for Risk and Innovation (GCRI) leads on science, technology and evidence: trigger verification, impact evaluation and data quality, through the Disaster Resilience and Recovery platform and its pages on forecasting for early action, decision analysis for disaster risk and the social science of disasters. Nexus Foundry is operated at national and regional level by Nexus Consortiums through their host and lead organisations, within a multilateral programme led by these three pillar not-for-profit institutions on a public-private-planet partnership.
The Global Risks Forum (GRF) leads on governance and cooperation, through its Diplomacy and Climate and Nature Governance platforms. The Global Risks Alliance (GRA) translates the evidence into finance questions, without lending, underwriting or rating, through its anticipatory finance topic. Work moves between the three in a cycle from question to evidence, analysis, publication in Reports and decision by competent authorities.
What this article does not claim: it describes methods and public evidence on anticipatory action. It does not design or endorse any trigger, model or programme, and it is not a forecast or warning. Working papers cited are not yet peer-reviewed. GCRI does not regulate or certify, GRF does not regulate or finance, and GRA does not lend, underwrite or rate. Naming an organisation, study or tool is not a claim of partnership, endorsement or deployment.
Imibuzo
Imibuzo neempendulo
How is an anticipatory action trigger set?
By choosing a forecast variable, threshold, lead time and area, and verifying them against past events with each action's cost and the acceptable rate of acting in vain in mind 3,7.
Is there randomised evidence that acting early helps?
Yes, for cash transfers. Trials in Bangladesh and Nepal and in Niger found larger short-run gains from early transfers than from later ones, though the differences faded over time 4,5.
How reliable are the global counts of activations?
They are voluntary, self-reported minimum totals, and part of the growth since 2022 likely reflects changes in data collection 7.
Where do benefit-cost ratios for anticipatory action come from?
From project evaluations and models that compare avoided losses with costs, such as a ratio of about 2.25 for drought projects in Mozambique 12; results depend on what is counted.
Imiqathango
Glossary
- Trigger verification
- Testing a trigger against past forecasts and events to see how often it would have acted correctly, missed an event or acted in vain.
- Action lifetime
- The period over which an anticipatory action keeps its value, used to choose the forecast lead time and threshold.
- Randomized controlled trial
- An evaluation in which units are assigned to groups at random so that differences in outcomes can be attributed to the intervention.
- Food consumption score
- A household food security indicator combining how often different food groups were eaten, weighted by nutritional value.
- Minimum total
- A count known to be incomplete, cited as a floor rather than as the full number.
Reuse and cite
Text CC BY 4.0. Cite the page with its date; data tables are linked under each figure.
Imithombo
References
13 sources, numbered in the order they are first cited
- Coughlan de Perez et al., Natural Hazards and Earth System Sciences (European Geosciences Union) (2015-04-23). Forecast-based financing: an approach for catalyzing humanitarian action based on extreme weather and climate forecasts. Natural Hazards and Earth System Sciences 15, 895–904. doi.org. Accessed 2026-10-28.
- UN Office for the Coordination of Humanitarian Affairs (Web page, current as of 2026-10-28). Anticipatory action. www.unocha.org. Accessed 2026-10-28.
- Coughlan de Perez et al., Hydrology and Earth System Sciences (European Geosciences Union) (2016-09-05). Action-based flood forecasting for triggering humanitarian action. Hydrology and Earth System Sciences 20, 3549–3560. doi.org. Accessed 2026-10-28.
- Christian et al., Becker Friedman Institute, University of Chicago (authors from the World Bank, WFP and the University of Chicago) (2025-11-01). Fast Action for Floods: RCT Evidence on Forecast-based Cash Transfers from Bangladesh and Nepal (Working Paper No. 2025-140). bfi.uchicago.edu. Accessed 2026-10-28.
- World Bank (Pople, Premand, Dercon, Vinez and Brunelin) (2025-05). The Earlier the Better? Cash Transfers for Drought Response in Niger (Policy Research Working Paper 11138), reproducibility package. reproducibility.worldbank.org. Accessed 2026-10-28.
- Gros et al., International Journal of Disaster Risk Reduction (2019-12-01). Household-level effects of providing forecast-based cash in anticipation of extreme weather events: quasi-experimental evidence from humanitarian interventions in the 2017 floods in Bangladesh. International Journal of Disaster Risk Reduction 41, 101275. research.vu.nl. Accessed 2026-10-28.
- Anticipation Hub (2026-05-28). Anticipatory action in 2025: a global overview. www.anticipation-hub.org. Accessed 2026-10-28.
- United Nations Secretary-General (2026-07-31). Implementation of the Sendai Framework for Disaster Risk Reduction 2015-2030: Report of the Secretary-General (A/81/307). documents.un.org. Accessed 2026-10-28.
- Congress of the Philippines (2025-09-12). Republic Act No. 12287: Declaration of State of Imminent Disaster Act. lawphil.net. Accessed 2026-10-28.
- World Meteorological Organization (2015-01-01). WMO Guidelines on Multi-hazard Impact-based Forecast and Warning Services (WMO-No. 1150). etrp.wmo.int. Accessed 2026-10-28.
- World Food Programme (Programme page, current as of 2026-10-28). Anticipatory Action for climate shocks. www.wfp.org. Accessed 2026-10-28.
- UN Office for Disaster Risk Reduction (2025-05-27). Global Assessment Report on Disaster Risk Reduction 2025: Resilience Pays. Summary for Policymakers. reliefweb.int. Accessed 2026-10-28.
- Food and Agriculture Organization of the United Nations (2025-11-14). In Brief to The Impact of Disasters on Agriculture and Food Security 2025: Digital solutions for reducing risks and impacts. doi.org. Accessed 2026-10-28.
Irekhodi
Editorial record
Facts and policies are stated as of 28 October 2026. This article is reviewed at least every 180 days and whenever a cited source changes, including peer-reviewed publication of the cited working papers, the next Anticipation Hub global overview and new randomised evaluations of anticipatory action.
Published by the GCRI editorial desk for the Disaster Resilience and Recovery platform. To report an error or request a correction, use the platform's corrections and concerns route; corrections are dated and listed with the article.
Text is published under the Creative Commons Attribution 4.0 licence (CC BY 4.0): reuse it with credit to GCRI and a link to this page. Third-party data keep their own terms.
Cite this article
GCRI editorial desk (2026, October 28). Anticipatory action methods: triggers, trials and evidence. Disaster Resilience and Recovery. https://therisk.global/disaster-resilience/anticipatory-action-methods-and-evidence/
@online{gcri2026anticipatoryactionmeth,
author = {{GCRI editorial desk}},
title = {{Anticipatory action methods: triggers, trials and evidence}},
publisher = {Disaster Resilience and Recovery},
year = {2026},
month = {10},
url = {https://therisk.global/disaster-resilience/anticipatory-action-methods-and-evidence/},
language = {en}
} TY - ELEC AU - GCRI editorial desk TI - Anticipatory action methods: triggers, trials and evidence PB - Disaster Resilience and Recovery PY - 2026 DA - 2026/10/28 UR - https://therisk.global/disaster-resilience/anticipatory-action-methods-and-evidence/ LA - en ER -
Republish
Republish or translate this article free of charge, in print or online, under CC BY 4.0: keep the byline and Disaster Resilience and Recovery, link to the original and note any change. Figures, maps and twins made by the institutions carry the same licence; third-party data keeps its own credit line.
Attribution to paste
<p>How anticipatory action is measured and modelled: triggers, evaluations and data — GCRI editorial desk, Disaster Resilience and Recovery. First published 28 October 2026 at <a href="https://therisk.global/disaster-resilience/anticipatory-action-methods-and-evidence/">https://therisk.global/disaster-resilience/anticipatory-action-methods-and-evidence/</a>. Republished under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/).</p>