Resilient institutions do more than respond to disruption. They build capacity to learn, question assumptions and change course while decisions can still make a difference.
Dr Emmanuel Y Jimenez, Director General of the Independent Evaluation Department at the Asian Development Bank (ADB), reflects on what decades of evaluation have taught multilateral development banks about turning evidence into action.
Resilience has become a central concern for institutions in development, business, government and beyond. But the concept is often better stated than practised. From your experience assessing whether large institutions are delivering on what they set out to do, what does genuine resilience demand of an organisation?
Genuine resilience depends on capacities built before a crisis occurs. For multilateral development banks (MDBs), this means supporting developing member countries to prevent and mitigate shocks, not only responding once a crisis has begun. It also requires MDBs themselves to work across internal silos. This is especially important for regional institutions such as ADB, which operates across countries exposed to economic, natural and health-related disruptions.
Independent evaluations have found that MDBs have generally succeeded in providing timely financial support in response to shocks.1 Much of this has taken the form of quick-disbursing budget support to central governments, easing balance-of-payment pressures or financing urgently needed public expenditure, including COVID-19 vaccines. MDBs have also developed instruments such as catastrophe bonds, to allow countries to access funds quickly during a shock. Moreover, they have helped countries develop institutions that enable people, especially the poor and vulnerable, to withstand shocks and recover. For example, I was privileged to work with the World Bank team that provided support to Asian countries developing social protection programmes after the 1997-98 Asian financial crisis; these countries were able to reach families more quickly during subsequent natural disasters and the COVID-19 pandemic.
These evaluations offer lessons about what resilience demands. One lesson is that assistance has been more effective in countries with more capable governments, underlining the importance of strengthening institutional capacity, especially during non-crisis periods. This is not always easy: immediate needs can crowd out longer-term but potentially more consequential measures, such as institutional reform.
A second lesson is the importance of coherence, both within an institution and across its partnerships. During crises, working across bureaucratic silos and allowing flexibility in institutional procedures have been essential to providing timely and effective support. Alignment with other development partners also matters. In Myanmar and Afghanistan, for example, when ADB could no longer work directly with the central governments, it partnered with United Nations humanitarian agencies to support education, health and food security. Resilience, in other words, rests not only on resources, but also on an institution’s ability to adapt how it works. Genuine resilience depends on capacities built before a crisis occurs. For multilateral development banks (MDBs), this means supporting developing member countries to prevent and mitigate shocks, not only responding once a crisis has begun. It also requires MDBs themselves to work across internal silos. This is especially important for regional institutions such as ADB, which operates across countries exposed to economic, natural and health-related disruptions.
Independent evaluations have found that MDBs’ have generally succeeded in providing timely financial support in response to shocks.1 Much of this has taken the form of quick-disbursing budget support to central governments, easing balance-of-payment pressures or financing urgently needed public expenditure, including COVID-19 vaccines. MDBs have also developed instruments such as catastrophe bonds, to allow countries to access funds quickly during a shock. Moreover, they have helped countries develop institutions that enable people, especially the poor and vulnerable, to withstand shocks and recover. Asian countries with functioning social protection programmes, many established after the 1997-98 Asian financial crisis, were able to reach families more quickly during subsequent natural disasters and the COVID-19 pandemic.
These evaluations offer lessons about what resilience demands. Assistance has been more effective in countries with more capable governments, underlining the importance of strengthening institutional capacity, especially during non-crisis periods. This is not always easy: immediate needs can crowd out longer-term but potentially more consequential measures, such as institutional reform.
A second lesson is the importance of coherence, both within an institution and across its partnerships. During crises, working across bureaucratic silos and allowing flexibility in institutional procedures have been essential to providing timely and effective support. Alignment with other development partners also matters. In Myanmar and Afghanistan, for example, when ADB could no longer work directly with the central governments, it partnered with United Nations humanitarian agencies to support education, health and food security. Resilience, in other words, rests not only on resources, but also on an institution’s ability to adapt how it works.
Evaluation is often seen as something that happens after the fact. How can institutions bring evaluative thinking earlier into strategy and decision-making, before choices become too entrenched to change?
Bringing evaluative thinking earlier into the programme or project cycle requires institutions to treat learning as part of accountability, not as an alternative to it. Lessons from ex-post evaluations—what worked, what did not and why—can be immensely useful in designing future projects. MDBs such as ADB and the World Bank assess completed operations and document the lessons.
With thousands of completed projects, one challenge is making such evidence easily accessible to the staff designing new operations. Artificial Intelligence is increasingly helping staff review a curated evidence base and retrieve relevant lessons when they are needed. Access alone, however, is not enough. Staff also need incentives to use those lessons. Their deliverable should not simply be a loan; it should be the development impact that the loan helps to produce.
Evaluation should therefore be built into project design from the outset, rather than left solely to independent evaluation units, assessing results after project completion for accountability. Evaluation during project implementation can provide real-time checks on projects, assess whether a project is on track and allow timely adjustments. There is evidence that such monitoring and mid-course correction can significantly improve implementation.2
The same principle applies at the institutional level. MDBs evaluate their engagement in sectors and themes as well as major corporate initiatives over seven-to-twelve-year periods. The thematic choice and timing of these evaluations determine whether they can inform operational decisions. ADB, for example, independently assesses the Asian Development Fund, which finances grants to its most vulnerable member countries, before donors consider its replenishment. Evaluative thinking has influence when it is connected to a decision that has yet to be made.
Much of the conversation on resilience focuses on sudden shocks. How can evaluation help institutions recognise slower pressures, such as underperforming programmes, outdated assumptions or strategies that have drifted from their original intent?
Institutions such as MDBs, which promote long-term economic development, are accustomed to slow-burning pressures. Their projects are derived from 5-10 country partnership strategies, which can take 1-3 years for design and approval, followed by another 5-7 years for implementation. By the time it closes and is evaluated, the context and some of the assumptions on which it was based may have changed considerably.
Building evaluation into a project can help an institution recognise these changes before they become entrenched. One research study examined 100 impact evaluations and 1,135 projects approved by the World Bank between 2005 and 2011. It found that projects with impact evaluations avoided common delays and halved the gap between planned and actual disbursements. According to the study, "Evidence-based mid-course corrections, a clearer implementation road map, strengthened capacity on the ground, and observer effects are possible channels to explain the results.”3
Evaluation must also distinguish between a failure to deliver and a justified change in course. When conditions alter substantially, as they did during the COVID-19 pandemic, outcome targets or implementation schedules may need to change. An ex-post evaluation should examine the validity of the reasons for those changes rather than merely assess a project mechanically against assumptions that no longer hold.
Some interventions also take much longer to produce observable results. It took years of evaluative research, following cohorts through different stages of their lives, to demonstrate the effects of early childhood education on adult outcomes, including further education and career trajectories. That was possible only where monitoring and evaluation systems had been built directly into the programmes. Recognising slower pressures therefore requires both timely feedback and the patience to follow outcomes over the appropriate period.
Many organisations are under pressure—from funders, boards and stakeholders—to demonstrate impact. What are the risks when impact becomes something to be reported rather than something to be learned from?
The volume of lending has often been used as an indicator of MDB performance because it is easy to measure. Volume is important but it is no longer sufficient as the only measure of success. Like many organisations, MDBs are under pressure to demonstrate the development impact of what that lending enables countries to buy. It is no longer enough to get money out the door; they must show that a project has contributed to a changed outcome.
In June 2026, the ADB approved new procedures to strengthen the focus on development impact in the design and implementation of its projects. One risk is that this becomes a “box-ticking exercise”, the bane of many a bureaucratic procedure. To guard against that, teams must provide a robust theory of change that identifies and justifies the assumptions linking what is financed to the outcomes promised in their project proposals. This should make it easier to evaluate not only whether a project achieved the desired result, but why it did and did not.
A second risk concerns the relevance of the indicator being reported. In any large organisation, stakeholders may differ over which results matter most. At the corporate level, organisations understandably want a limited set of indicators that demonstrate strategic coherence. ADB now tracks 35 main results in its corporate scorecard, compared to 60 three years ago; the World Bank tracks 22 indicators in its corporate scorecard, down from about 150. These measures may be important in aggregate but less relevant in a particular country. A recent ADB evaluation of its engagement with an upper-middle-income country found that teams felt pressure to prioritise indicators that could be aggregated to the corporate level, even though they did not reflect local priorities. When that happens, reporting can displace learning rather than support it.
ADB works across highly varied development contexts in Asia and the Pacific. What have these different contexts revealed about how evaluation needs to be designed, communicated and used?
ADB supports countries where the highest per capita income is about 30 times the lowest, and where the most populous country has around 800,000 times more people than that of the smallest.4 This diverse context has a significant bearing on how we evaluate corporate, country and subregional strategies. The challenge is to preserve overall coherence while allowing enough flexibility to respond to markedly different needs.
Our evaluation teams have therefore had to develop criteria for effective country differentiation. A recent evaluation of ADB’s engagement with its upper-middle-income countries concluded that the institution should move away from offering a common menu of support as countries transition to higher-income status. Its operations should demonstrate clear value in relation to the increasingly complex needs of each country. Alongside its traditional role of lending to governments—which upper-middle-income countries require less of—ADB can provide knowledge services and mobilise private capital.
Context also affects how evaluation is communicated and used. Our department has shifted its approach to evaluation capacity building in the region. Until a few years ago, we supported individuals in learning how to design and conduct evaluations. Many alternatives, including online courses, are now available. We instead offer a learning platform that helps countries build evaluation systems suited to their own institutions and needs. The aim is not to transfer a single model, but to help countries develop systems that can use and sustain.
Independent evaluation needs to be credible, but it also needs to be close enough to decision-making to be useful. How do you balance independence and influence?
MDBs have institutional guardrails to ensure independence, which is required for accountability. At ADB, the Independent Evaluation Department (IED) reports not to the President or management, but to the institution’s shareholders through the Board of Directors. Its head serves a single, non-renewable term and is contractually prohibited from subsequently taking up a staff or consulting position with ADB.
To ensure influence, the current mantra of many evaluation units of MDBs like the ADB is “independence but not isolation.” It begins with what gets evaluated. The most influential independent evaluations are topical and relevant to decisions facing management. We thus consult extensively with both the Board and management (the evaluee) in developing a three-year rolling work programme.
In this sense, influence is “hardwired” into the process. When management agrees with a recommendation from a major thematic or corporate evaluation, it must prepare and implement an action plan that is monitored and tracked for up to five years. In the past year, more than two-thirds of actions due for completion had largely or completely implemented.
Formal accountability processes are not enough. Independent evaluation should also be seen as useful. One way to do so is to make the evaluations more accessible. At IED, we have worked to make evaluations shorter, more readable and more clearly focussed on their central messages, without losing rigour. We have also developed an AI tool that helps staff retrieve and distil lessons easily when they need them.
Evaluations are also more influential when staff feel they have been developed and presented constructively. A ‘gotcha’ approach often elicits defensiveness, and learning diminishes. Findings that are clearly explained and well evidenced are more likely to encourage reflection and openness to feedback. Early engagement also helps: I've found that hard messages are easier to absorb if the evaluee is not taken by surprise. Independence protects the credibility of the findings; engagement increases the likelihood that they will be used.
Across your work with the World Bank, 3ie and ADB, do you think large institutions have become better at learning from evaluation? What remains hardest to change?
MDBs have become much better at generating evidence. All the major ones have independent evaluation departments to ensure coherence in standards and practices. MDBs also conduct increasing numbers of rigorous impact evaluations through their research and analytical work. But while they generate more evidence, it is less clear whether they have become better at learning from it.
Research shows that using evidence to inform project design can lead to better outcomes, while evaluation during implementation can improve delivery. Yet evaluations of the learning ecosystems at ADB and the World Bank show that embedding the systematic use of such evidence in institutional practice remains challenging.
Systemic change requires a stronger culture of evidence-seeking. Institutions need the capacity not only to generate evidence, but also to interpret and use it. More difficult still is creating incentives for staff to seek it out. Generating evidence and incorporating it into business processes takes time and resources. If the findings are taken seriously, they may also raise the standard for what an institution is willing to support. That is difficult when the easiest measure of success (and hardest to forego given the business model of a bank) is still the volume of lending.
For leaders outside the development sector, what does it take to build a culture of evaluation rather than simply an evaluation function? Where should they begin?
One lesson from international development is that a culture of evaluation starts at the top. In MDBs, independent evaluation began because shareholders demanded accountability. But formal requirements alone are not enough. Executive leaders must send a clear message that honest assessment, including evidence of failure or underperformance, is valued and will be acted upon.
Leaders should begin with a diagnosis of both capacity and the incentives. Can the organisation generate and use credible evidence? Do staff feel empowered to speak truth to power? Are they supported or punished when they provide an honest assessment of what works and what does not?
Finally, what gets measured gets done. Leaders should ask whether evaluative evidence is easy to access and whether systems are in place to track how it is used. An evaluation function can produce reports. A culture of evaluation is present only when people across the organisation seek evidence, discuss it openly and use it to make different choices.
For a list of references to this article, please click the link here.