The Graveyard of Promising Pilots: Why Scale Fails and What to Do About It

by

Edward Addai

Edwardaddai64@gmail.com

Introduction

There is a graveyard of promising pilots in the social sector. You know their stories because you have lived them. A beautifully designed education programme in fifty Malawian schools doubles learning outcomes. A community health initiative in three Ugandan districts cuts maternal mortality by a third. A child protection protocol in Indonesia keeps vulnerable children safer. The evidence is robust. The funders are impressed. Then when it comes to scaling up the voltage drops.

The pattern is so familiar it has become an expectation: somewhere between 50 and 90 percent of social sector initiatives that show promise at small scale fail to deliver equivalent results when expanded (List, 2025). This is not merely an implementation problem. It is a conceptual problem. We have been thinking about scale incorrectly, as an exercise in replication rather than transformation, as a technical challenge rather than an adaptive one.

This thought piece argues that scaling for impact requires a fundamental shift in mindset. Drawing on complexity science, institutional economics, and ecosystem literature, with evidence from health, education and child protection in developing countries, I will explore what scaling truly means, why it matters urgently, why it proves so difficult, including its hidden resource intensity, and how to approach it differently.

What Scaling Really Means

Let us begin by clarifying what we mean by scale. The science of scaling has been defined as "the systematic pursuit and application of knowledge to optimise sustained equitable outcomes of innovation and scaling" (Redman et al., 2025). This definition moves us beyond the common conflation of scale with size alone.

Scale encompasses three distinct dimensions. The first is population coverage, i.e. reaching more people. The second is package expansion, i.e. diversifying the intervention. The third is systemic integration involving embedding the approach into structures, financing and norms that sustain it over time. A school feeding programme that reaches 10,000 additional children has achieved population coverage. One that adds deworming tablets has achieved package expansion. But one that shifts national education policy to mandate school meals, with dedicated budget lines and accountability mechanisms, has achieved systemic integration.

 

But systemic integration is not enough. Drawing on ecosystem literature, interventions exist within broader ecosystems of actors, institutions, resources, and relationships (Frenken & Neffke, 2025). No single organisation controls all elements needed for population-level change. This is why scaling is also fundamentally about managing for collective impact i.e. the commitment of actors from different sectors to a common agenda for solving complex social problems (Kania & Kramer, 2011).

The means, true scale requires altering the interconnections across this ecosystem. And this requires organisations to develop dynamic capabilities: the capacity to integrate, build, and reconfigure internal and external competencies to address rapidly changing environments (Teece, Pisano & Shuen, 1997 to sense ecosystem shifts, seize opportunities, and transform routines in response.

Cambodia's long-term area programmes illustrate this. After a decade across three districts, programmes achieved open defecation-free status, increased exclusive breastfeeding from 13% to 94%, reduced child stunting by more than 50 percent, and increased skilled birth attendance from 15% to full coverage (World Vision International, 2025). These outcomes were sustained because multiple actors shifted together: service providers improved capabilities, communities became aware of their rights, and local government allocated budget for sustainability.

Why Scale Matters

The rationale for scale is stark. In Guinea-Bissau, fewer than 30% of students complete primary school (Eisen, 2025). In Ethiopia, 48.6 million people are under 18, with 57% prevalence of childhood maltreatment (Mackenzie et al., 2025). In Indonesia, over 19,000 incidents of child violence were recorded in 2024 alone (Mackenzie et al., 2025).

These are not problems that can be solved by discrete programmes operating at the margins. They require population-level change. They require scale. And they require collective action across multiple actors including governments, civil society, communities, each contributing different capabilities.

Why Scaling Is Difficult

If the case for scale is compelling, the practice remains elusive. John List's "voltage drop" captures the phenomenon: when an enterprising idea falls apart at scale (List, 2025). Causes include false positives from underpowered trials, unrepresentative samples, and the "chef versus ingredients" problem, where success is attributable to exceptional practitioners rather than replicable mechanisms.

The ecosystem perspective adds another layer: interventions are embedded in ecosystems of actors with their own interests and capabilities (Frenken & Neffke, 2025). Scaling disrupts these ecosystems. New actors enter. Relationships shift. The intervention that worked in a stable pilot ecosystem may fail in a different one.

If scaling requires navigating ecosystem dynamics, organisations need dynamic capabilities to do so. Yet many social sector organisations are configured for stable environments. Their capabilities are operational, delivering services efficiently, rather than dynamic, sensing and responding to change (Teece, 2007). The PERFORM2Scale study in Ghana, Malawi and Uganda found that repeated cycles of a management strengthening intervention fostered autonomy, innovation, and teamwork among district health teams (Raven et al., 2024). These emergent outcomes required teams to develop new capabilities. Where such capabilities were absent, scale faltered.

Institutions are "the rules of the game" (North, 1990). Interventions encounter existing incentive structures, power relations, formal rules and informal norms that shape what is possible. The PERFORM2Scale scalability assessment found that while the intervention was credible, its scalability was constrained by limited government buy-in and lack of sustained financial investment (Bulthuis et al., 2022).

There is also an uncomfortable truth we rarely name: doing scale well requires significant resources, often more than the pilot itself. The coordination mechanisms for collective impact including shared measurement systems, backbone support, continuous communication, require dedicated funding (Kania & Kramer, 2011). Building dynamic capabilities demands investment in learning systems, staff development, and adaptive management. Ecosystem mapping and stakeholder engagement take time and money. Yet funders often assume scale means efficiency and that per-unit costs should fall. The transition to scale involves transition costs that are front-loaded and easily underestimated.

How to Scale for Impact: Building Dynamic Capabilities for Collective Impact

If scaling requires navigating ecosystem dynamics through collective action, and if organisations need dynamic capabilities to do so, and if all this requires significant resources, how should social sector leaders proceed? Here are six concrete practices.

Map Your Ecosystem Before You Design

Most pilots are designed in isolation, then struggle when they encounter the real world. Start by understanding who else matters. Identify all actors who influence your problem including government agencies, other NGOs, community groups, private sector actors. Map their interests, capabilities, and relationships. Ask who needs to be part of this for change to stick. What coordination mechanisms already exist? The Cambodia programmes succeeded because they understood that shifting local government budgets was as important as shifting community behaviour. This mapping takes time and money, so budget for it from the beginning.

Build Sensing Systems That Tell You What's Really Happening

You cannot adapt to what you cannot see, yet most monitoring systems track only whether you delivered activities, not what is changing in the ecosystem. Build simple, rapid feedback loops that capture what different actors are experiencing. The BEFIT tablet programme in Malawi complemented its randomised trials with annual surveys of teachers, parents, and children (Global Partnership for Education, 2026). Community-defined competencies, such as attentiveness, independence, motivation, respect, were incorporated into monitoring. Findings informed adjustments to implementation, enabling continuous improvement at-scale. This learning infrastructure cost money and was essential to success.

Invest in Relationships That Bridge Levels

Problems at the policy level cannot be solved by action only at the community level. But no single organisation controls all levels. Identify the actors best positioned at micro (community), meso (organisational), and macro (policy) levels. Then invest in relationships that connect them. The PERFORM2Scale project succeeded across 27 districts because "resource teams" and "champions" bridged district, regional and national action (Raven et al., 2024). These actors operated at multiple levels, creating the connective tissue that collective impact requires.

Embrace Complexity

Child protection scholars distinguish between "complexity reduction" i.e. trying to simplify systems to make them manageable, and "complexity absorption" i.e. building the capacity to work within complexity (Gillen et al., 2025). The latter means developing people who can synthesise multiple perspectives, tolerate uncertainty, and make decisions in dynamic conditions. Create spaces for reflection and adaptation. Value learning alongside accountability. Measure not only outcomes but also the system conditions that produce them. Cultivate complexity literacy throughout your team.

Assess Scalability and Resource Needs from the Start

Use frameworks like WHO/ExpandNet's CORRECT criteria - Credibility, Observability, Relevance, Relative advantage, Easy to implement, Compatibility, Testability (Bulthuis et al., 2022). But also ask deeper questions: What collective action is required? What dynamic capabilities will we need to develop? What will this actually cost, including transition costs, coordination infrastructure, and capability building? The PERFORM2Scale scalability assessment found that while the intervention was scalable, certain aspects required adaptation, including greater involvement of regional and national actors and improved documentation of results (Bulthuis et al., 2022). Continuous assessment enabled critical reflections on next steps.

Engage with Political Economy and Build Institutional Capabilities

This means understanding whose interests are served or threatened by change, building coalitions that can sustain momentum through political cycles, and working to shift the institutional rules that shape what is possible. Map the political economy of scale-up. Identify interests, incentives, and power relations. Build coalitions across sectors and levels. Work patiently on institutional change comprising the rules of the game that determine what endures. And invest in the collective infrastructure including shared measurement, backbone support, continuous communication that enables multiple actors to develop dynamic capabilities together.

The Ghana cocoa project illustrates this approach. Addressing child labour required linking climate action, sustainable livelihoods, and child protection (ILO, 2026). Farmers were trained to produce biochar from cocoa pod husks, restoring soil fertility and improving productivity. Community child protection committees were reinforced. The project aligned with Ghana's National Plan of Action for the Elimination of Child Labour. This was not rapid scale-up. It was patient institutional and ecosystem building involving building dynamic capabilities across multiple actors, and it produced transformation that rapid replication could never have achieved.

Conclusion

Scaling for impact requires moving beyond the replication mindset. It requires engaging with complexity, understanding institutions as the rules of the game, thinking in terms of ecosystems, managing for collective impact, and building dynamic capabilities at organisational and institutional levels. And it requires honesty about a difficult truth: doing this well costs money. More than the pilot cost. More than we usually budget for.

The transition to scale involves transition costs. Coordination infrastructure costs. Capability building costs. These are not inefficiencies. They are investments in the ecosystem conditions that make population-level change possible.

The children in learning poverty in Zambia, where "Teaching at the Right Level" is being tested (de Barros, 2025). The families excluded from social protection in Guinea-Bissau (Eisen, 2025). The children at risk of violence in Indonesia and Ethiopia (Mackenzie et al., 2025). They do not need more promising pilots. They need population-level change. And population-level change requires collective action, dynamic capabilities, and adequate resources to build the ecosystems within which solutions can flourish.

You can only change the world at scale. The question is whether we are willing to resource what that actually requires.

References

Bulthuis, S. et al. (2022) 'Assessing the scalability of a health management-strengthening intervention', Health Research Policy and Systems, 20(1), e85.

de Barros, A. (2025) 'Scaling Differentiated Instruction', National Academy of Education.

Eisen, K. (2025) 'Human Capital Investment Project', The Borgen Project, 12 June.

Frenken, K. and Neffke, F. (2025) 'Economic Geography and Complexity Theory', SFI Press.

Gillen, A. et al. (eds) (2025) Systems complexity in child protection and welfare. Routledge.

Global Partnership for Education (2026) 'Malawi: Strengthening EdTech evidence', 4 February.

International Labour Organization (2026) 'Climate-smart cocoa project in Ghana', 9 March.

Kania, J. and Kramer, M. (2011) 'Collective Impact', Stanford Social Innovation Review, 9(1), pp. 36-41.

List, J. (2025) cited in TASO, 'Making an impact at scale', 17 April.

Mackenzie, C. et al. (2025) 'Advancing child protection in Indonesia and Ethiopia', Child Abuse & Neglect, 170, 107762.

North, D.C. (1990) Institutions, Institutional Change and Economic Performance. Cambridge University Press.

Raven, J. et al. (2024) 'From PERFORM to PERFORM2Scale', Health Policy and Planning, 39(8), pp. 841-853.

Redman, S. et al. (2025) 'The science of scaling in health', ASPA Conference 2025.

Teece, D.J. (2007) 'Explicating dynamic capabilities', Strategic Management Journal, 28(13), pp. 1319-1350.

Teece, D.J., Pisano, G. and Shuen, A. (1997) 'Dynamic capabilities and strategic management', Strategic Management Journal, 18(7), pp. 509-533.

World Vision International (2025) 'More than a decade of impact in Cambodia', 5 September.