Skip to main content

EvalXchange 2025: Powering Decisions through Evidence, Efficiency & AI

Posted on 26/06/2025 by Alexandra Priebe, Wilson Olarasha Kaikai
 WFP/Arlette Bashizi.
WFP/Arlette Bashizi.

Authors: Alexandra Priebe, Wilson Olarasha Kaikai and Federica Zelada

In April, the World Food Programme (WFP) Office of Evaluation hosted EvalXchange, a series of virtual learning events.  EvalXchange is now in its fifth year, uniting knowledge from WFP evaluations and those of partners, creating a space for participants to learn, share experiences, and improve the practice of evaluation. This year’s theme centered around efficiency. 

Session 1: EvalX Live-The Evaluation Talk Show: Beyond the Report

Styled as a talk show, this engaging session brought evaluation users and practitioners from around the globe. Together they discussed how evaluation functions can better leverage evidence to more efficiently meet organizational needs, and to adapt evidence to specific demands across topics and countries. 

Keeping users at the center of discussion, the first segment looked at evidence in action.  This segment spotlighted users from across WFP who engage with innovative use products and targeted learning sessions to inform not only policy and strategy development, but also the design of implementation models.  While evaluation reports remain important resources for learning, it takes time to read and digest them. The Office of Evaluation is responding innovatively by facilitating the generation of new evidence offerings, Summaries of Evaluation Evidence (SEEs), which maximizes the use of existing evaluation reports to support learnings that meet users’ immediate evidence needs for forward-looking decision-making. Targeted learning sessions are also a critical resource for summarizing evidence across evaluation reports. They allow for interactive engagement with technical leads to expedite learnings and help drive evidence-based decisions.

Evaluators from UNICEF, Konterra Group, and WFP unpacked approaches to evidence use during the second segment. Moving from instrumental to conceptual use requires identification of key users as evaluation brokers, who can leverage their international position and lead discussions on evaluation findings with the larger group of stakeholders. Practitioners highlighted several learnings from developing rigorous and compelling SEEs: 1) Being clear with commissioners about what SEEs can and cannot do. SEEs are not new evaluations but are limited by the evaluations that they draw from; 2) Look for magic patterns, not magic bullets. SEEs must look at patterns emerging across multiple evaluations; 3) Broader is better – looking across regions and across the system can create more interesting patterns; 4) Get the questions right – more conceptual questions tend to be more interesting to users; and 5) Engage the SEE lead/team, as they have the most nuanced understanding of the topic and leverage interaction events to energize the conversation. The best SEEs are the ones that are explicitly linked to an ongoing learning event and are thematic or broader in scope.

Session 2: Unpacking Efficiency: Options, Challenges and Lessons for Evaluation

A total of 127 evaluation practitioners from UN agencies, NGOs, international financial Institutions, governments, donors, and academia reflected on the different perspectives and experiences on how to evaluate efficiency in humanitarian and development contexts.

Evaluation practitioners recognized that in the current context, organizations are operating in an environment where the growing need is growing while funding for humanitarian response and development programs are decreasing. Therefore, there is increasing pressure, demand and call for organizations to be more efficient, effective and impactful. Consequently, there is an increased demand by evaluands to evaluate the efficiency of programs to help organizations establish the extent to which they are making the best use of available resources for achieving desired objectives and outcomes. 

The OECD/DAC and ALNAP provide useful guidance on how to evaluate the efficiency of programs, but evaluating efficiency is not always straightforward. There are several different approaches, methods and tools for evaluating efficiency, which presents opportunities and challenges for evaluators. The different dimensions (including economy and cost-effectiveness) of efficiency are not always independent from each other. Therefore, they need to be analyzed in conjunction with the other evaluation criteria (such as relevance and effectiveness). Evaluators should be aware of tradeoffs between efficiency and other considerations, such as humanitarian principles (impartiality, humanity), protection, and environmental footprint, among others. In addition, evaluators are confronted with the challenge of assessing the cost-effectiveness of an intervention, as not all costs and benefits are easily measurable. 

This session highlighted that, in order to design and choose appropriate approaches and methods of evaluating program efficiency in different contexts, evaluators should complement the use of OECD/DAC guidance, with practical examples and lessons of how to address challenges of evaluating efficiency.

Session 3: Exploring the AI Frontier: The Adventure

The session highlighted how AI tools can enhance evaluation by improving credibility and transparency. Participants agreed that AI offers genuine opportunities to accelerate and streamline the use and reuse of evidence, while emphasizing that success depends on ensuring these solutions align closely with evaluation needs. Harnessing AI in evaluation is a complex but valuable process that requires continuous learning, collaboration, and adaptation. Evaluators invested significant effort in understanding AI and bridging the gap with data science, though further work is ongoing to develop tools that efficiently deliver evidence for more impactful decision-making.

Collaboration between evaluators and data scientists was seen as essential to building shared understanding and trust, which helps ensure tools are both technically sound and practically relevant in real-world evaluation contexts. Lessons from developing and using AI in real situations show that success depends on carefully designing tools, ensuring accurate information, and sharing results in ways that are easy to use and understand. These elements are essential for effectively bringing AI into evaluation work.

Involving diverse teams and partners supports integration with existing systems and future initiatives, preventing siloed solutions. A modular design offers the flexibility to update or replace components as technology evolves, which reduces risks such as vendor lock-in. Maintaining operational efficiency by balancing affordability with predictable costs and planning for specialized expertise (especially in machine learning operations) is critical for ensuring long-term sustainability.

Finally, sharing experiences, lessons learned, and challenges is essential for navigating this evolving landscape and to better harness AI-driven solutions.