This evaluation assesses the World Bank’s approach to knowledge and learning in its lending operations. The evaluation is guided by the high-level question, How can the World Bank create optimal conditions for learning in and from its financing operations? The evaluation’s analysis covers all financing instruments; country, sectoral, and operational knowledge; and tacit and explicit knowledge (box 1.1 defines these and other key terms). The evaluation applies a mixed methods, phased approach that combines case studies of 34 lending projects with desk reviews, interviews, a questionnaire, and quantitative analyses. It aims to support World Bank management’s ongoing efforts to improve the institution’s performance as a “Knowledge Bank” as outlined in The Knowledge Compact for Action: Transforming Ideas into Development Impact—For a World Free of Poverty on a Livable Planet and Realizing the World Bank Group’s Knowledge Potential for Effective Development Solutions: A Strategic Framework (World Bank Group 2021, 2024).

Overall, the evaluation finds that the World Bank’s knowledge ecosystem has a formal, linear model for embedding knowledge in its lending processes that creates learning moments at specific points of the project cycle, especially during the preparation stage. The World Bank also has a strong informal learning culture that often helps project teams identify actionable sector and country knowledge. However, this ecosystem has imbalances, weak lesson learning, and fragmented institutional support.

Knowledge and Learning in the Linear Model

The World Bank uses a linear model to embed knowledge in its financing through key entry points in a project’s preparation, approval, implementation, and completion phases. The linear model excels at producing explicit technical knowledge sourced from both inside and outside the World Bank to inform project designs. Consequently, sampled project designs used diverse knowledge sources that predictably included much World Bank–generated knowledge, including project diagnostics and various types of advisory services and analytics, but also frequently turned to knowledge from clients, multilateral institutions, and other external sources. These projects rarely used the World Bank’s “core” advisory services and analytics, which are a specific set of World Bank reports prioritized by the Knowledge Compact that serve to inform decision-making and policy making at strategic levels for the World Bank, its clients, and other partners and stakeholders in a way that transcends specific operations. Indeed, only 2 percent of all references in Project Appraisal Documents are to any of the core or extended core advisory services and analytics report series. Development policy financing operations more often cited these reports.

The linear model is not ideal for fostering learning during the project implementation phase and underemphasizes country knowledge. The project preparation phase has budgets and formal requirements to produce specific knowledge inputs and technical assessments, which the implementation phase lacks. Learning diminishes once a project is launched. Moreover, the linear model focuses its explicit knowledge work on generating sectoral knowledge rather than country knowledge, which tends to be tacit and shared unsystematically.

Lending instruments vary in what type of knowledge input they emphasize. Investment project financing operations use studies to identify or justify new projects, to design a project’s implementation arrangements, or to identify the specific pieces of infrastructure it will finance. Development policy financing operations drew on policy studies, and Program-for-Results (PforR) financing operations used integrated risk assessments to bolster projects’ technical soundness and support for client systems.

PforR and Multiphase Programmatic Approach operations have knowledge mandates that better suit them to learning. The investment project financing and development policy financing instruments have formal policies, procedures, and guidance to acquire or use knowledge in the operations’ design phase but not during the implementation phase. PforR operations, by contrast, have explicit incentives and mechanisms for learning during implementation because of their annual validation of disbursement-linked indicators (DLIs). The DLI process creates a formal, mandated space for knowledge sharing and learning among World Bank teams and clients. This space is supported by DLI data on PforR’s progress toward achieving project outcomes. The Multiphase Programmatic Approaches similarly focus on knowledge because they require learning agendas that span the Multiphase Programmatic Approaches’ multiple phases and operations. However, these learning agendas are not always of the same quality, with some being comprehensive and fully implemented and others not. World Bank operations have a formal lesson learning mandate during Implementation Completion and Results Reports, but these are underused by teams who frequently view them as compliance tools rather than learning tools.

Learning from tacit knowledge and informal exchanges is deeply embedded in the World Bank’s organizational culture. Yet the World Bank has not optimized the transfer of its tacit knowledge. Indeed, some of the most valuable and widespread learning moments are the least formal ones, often on the margins of the linear model. For example, the World Bank has a structured and valuable peer review process for improving project design quality. Quality enhancement reviews are also valuable because they occur earlier and are less formal than other review meetings. These processes create the most value for project teams when they result in early and frank advice and dialogue in an informal safe space.

Certain forms of knowledge are better preserved by the World Bank for later use than others. The evaluation found that review and decision meetings, peer reviews, and studies underpinning project designs were well documented, although that does not ensure use. In contrast, trust-funded studies, project-commissioned innovations, and informal client learning during implementation were not often documented, making such knowledge hard to reuse or build on. Further, the World Bank does not document the lessons from dropped and canceled operations, hindering learning from these experiences and from mistakes and failures more broadly. Additionally, existing tools to capture and learn from country knowledge have shortcomings, making country office staff the primary repository for tacit country knowledge (World Bank 2020d).

Ecosystems for Knowledge and Learning

The World Bank relies on a broader decentralized knowledge ecosystem that goes beyond the formal linear learning model for lending. This ecosystem includes the resources, opportunities, capacities, and client engagements that create effective learning conditions for staff outside of mandated project-related requirements. It is within this broader knowledge ecosystem that successful teams and units create the most valuable learning opportunities using tacit knowledge, regular exchanges within and beyond the team, safe spaces, and sustained client engagements. Trust funds often support them in this process. In sum, the linear model for embedding knowledge in financing relies on formal knowledge mandates in the project design and approval stages, whereas the broader knowledge ecosystem relies on tacit knowledge, informal approaches, and staff’s and managers’ motivation to drive learning.

Global units’ knowledge production often focuses on corporate and global unit priorities and not on operational needs or country knowledge. The absence of a centralized knowledge system means business units must develop their own knowledge systems, with some thriving and some falling behind. The World Bank has no minimum standards and has few incentives and mechanisms to ensure excellence in knowledge sharing and use.

Learning is often driven by task team leaders’ motivation and external trust funds. The evaluation finds that in the absence of formal learning mandates, it is the staff’s motivation to pursue knowledge and learning that often determines the level of knowledge and learning the World Bank generates. Moreover, without internal administrative budgets for knowledge and learning, teams must turn to trust funds to finance these activities. These become incredibly valuable for new projects that do not have prior operations from which to learn. However, this also means that much of this important work depends on the availability of trust fund resources.

Tacit learning is preferred by staff and embraced by some managers because it is a trusted and easy-to-access source of actionable knowledge. This informal learning stems from candid conversations in informal safe spaces. These spaces are valuable because of their confidentiality, the access they provide to trusted experts, and their ability to generate actionable knowledge that teams can directly apply in their projects’ contexts. Approximately half of the sampled World Bank teams collaborated with development partners on learning initiatives, and nearly all task teams shared knowledge with clients and fostered joint learning. Yet informal sharing of tacit knowledge also has risks and limitations, such as narrow diffusion of knowledge, bias toward perceived successes, the tendency to ignore failures, and inaccuracies. Despite its prevalence, the transfer of tacit knowledge is not systematized within the institution. For example, the World Bank’s few knowledge and learning staff primarily focus on explicit knowledge sharing but do not regularly support informal knowledge exchanges nor have the remit to support World Bank–wide knowledge management.

Conclusions

The World Bank’s approach to learning (which typically encompasses a linear, formal project-driven model and multidimensional, informal systems) has three major imbalances. First, the linear model creates a focus on generating sectoral knowledge and producing reports but often neglects applying this knowledge to operations and country-level processes. It also steers teams’ and management’s attention to the design phase, leading to less attention to knowledge and learning during the implementation phase. Second, the World Bank’s lesson and country learning often relies on unpacking tacit experiences in prior or parallel operations. However, this approach is not always reliable because it depends on chance discoveries and shortchanges smaller countries and new engagement areas that typically lack prior or parallel operations to learn from. This is part of a larger issue of a culture that greatly values informal knowledge exchanges but organizes them with little methods and support. Third, the World Bank’s knowledge ecosystem is generally fragmented, disconnected, and underresourced. This lack of consistency and minimum standards leads to inefficiencies in the knowledge flows.

Recommendations

Recommendation 1. Make better use of the learning opportunities that are already embedded in the lending processes. Operations Policy and Country Services should revise the procedures and guidance for lending to incentivize more consistent learning throughout the lending cycle. Specifically, Operations Policy and Country Services should set clear expectations to the type of knowledge and learning, the moments and processes for learning activities, and the level of client engagement required from project teams (examples of what such process tweaks could entail are offered in chapter 4). At the same time, managers at all levels should create spaces for knowledge and learning, including via quality enhancement reviews and informal and early meetings with peer reviewers. Managers should also role model attention to knowledge and learning and openness to discussing failures.

Recommendation 2. The Knowledge and Learning Department should ensure core knowledge management capacity and set World Bank–wide standards and processes for knowledge capture, storage, sharing, and access. The World Bank should rationalize the core capacity to organize key learning events with instructional designers and knowledge management professionals. The essential knowledge management capacities, standards, and processes should have oversight by senior management. Implementing this recommendation could entail:

• Setting standards for how knowledge is tagged, classified, stored, and shared across units.
• Supporting the Global Practices in adopting these standards, for example, by insisting that knowledge from key events such as Knowledge Weeks, client workshops, and long-term learning engagements is tagged, classified, stored, and shared across units.
• Professionalizing and enhancing the capacities of knowledge management staff. Supporting career management for knowledge management staff.
• Working with the Global Practices and Regions to better leverage professional knowledge management staff and thereby relieve task team leaders.
• Working with the Global Practices and Regions to bring enhanced methods and support to knowledge sharing activities, including enhanced attention to instructional design.
• Bringing methods and support to enhance informal knowledge exchanges. Investing also in communities of practice and technical help desks.
• Periodically surveying staff’s need for, and satisfaction with, knowledge management.