Why Systems Change Fails When We Fund the Wrong Unit of Change

May 06, 2026

9 min read

3 views

74%

The Hidden Mistake in Most Change Efforts

What if the reason so many social programs, public health initiatives, and reform efforts stall is not that they lack evidence, but that they are built to fund the wrong thing?

We usually finance projects, organizations, and outcomes. But many of the problems we care about, from chronic disease to inequality to climate vulnerability, do not move through single organizations at all. They move through relationships, trust, coordination, shared learning, and adaptation. In other words, they move through networks.

That creates a deep tension. Traditional funding wants a clear recipient, a budget line, a timeline, and measurable outputs. But systems change rarely behaves like a machine. It behaves more like a living ecosystem. You can irrigate one plant, but if the soil is poor, the roots are entangled, and the weather keeps shifting, the plant will not thrive unless the whole environment changes.

The real question is not, “Which organization deserves funding?” The deeper question is, what is the actual unit of change?

The Difference Between Buying Results and Building Conditions

A project is easy to understand because it has boundaries. A hospital runs a campaign. A nonprofit trains community health workers. A city launches a pilot. You can count the people reached and the dollars spent. That clarity is seductive, because it makes control feel possible.

But many high-stakes problems do not respond primarily to isolated interventions. They respond to the conditions that allow interventions to spread, adapt, and stick. That is where networks matter. A network is not just a collection of actors. It is a structure that determines how quickly ideas travel, how much trust exists between people, and whether learning can move from one setting to another.

This is why a single well-designed intervention can fail in one place and take off in another. The difference is often not the intervention itself, but the network around it. If local leaders trust one another, if practitioners share stories, if there are bridging relationships between sectors, then small successes can become contagious. If not, even excellent ideas stay trapped inside pilot programs.

Systems change is not mainly about creating better answers. It is about creating better pathways for answers to spread.

That shift in emphasis changes everything. It means the work of change is not only technical. It is relational, infrastructural, and temporal. It unfolds through conversations, reputation, coordination, and repeated experiments that slowly accumulate legitimacy.

Why Evidence Alone Rarely Moves a System

Public health has long lived with a frustrating contradiction: some interventions are supported by strong evidence, yet adoption remains uneven. We often assume the problem is ignorance. If people knew more, they would do the right thing. But systems rarely resist because they lack facts. They resist because facts have to compete with habits, incentives, institutional routines, and local context.

This is where a realist way of thinking becomes especially useful. Instead of asking only, “Does it work?”, a more revealing question is, what works, for whom, in what circumstances, and through what mechanisms? That question recognizes that context is not noise. Context is the mechanism that determines whether an intervention lands or bounces off.

Consider a vaccination campaign. In one neighborhood, uptake may rise because trusted community leaders endorse it, clinics are accessible, and communication is culturally resonant. In another neighborhood, the same campaign may underperform because transportation is unreliable, rumors spread faster than official messages, and historical distrust is high. The evidence did not change. The surrounding system did.

This is why the fantasy of one-size-fits-all scaling so often disappoints. Scaling is not just replication. It is translation. It requires understanding which parts of a model are essential, which are adaptable, and which depend on local relationships. A networked approach to change respects this reality. It treats variation not as a problem to eliminate, but as something to learn from.

The consequence is profound: the most valuable thing a funder can often support is not delivery, but adaptation capacity. The ability to interpret local conditions, test changes, exchange knowledge, and coordinate action may matter more than any single intervention package.

Funding the Network, Not Just the Node

This is where many change efforts get stuck. They invest in charismatic hubs, flagship nonprofits, or isolated pilots, then wonder why momentum dissipates when the grant ends. Funding one node in a network is not the same as strengthening the network itself.

Think of a road system. If you repair only one intersection, traffic may improve locally, but the system is still fragile. If you fund the roads, signage, connections, and traffic rules, the whole city becomes more navigable. Social change works similarly. The crucial asset is not simply the actor at the center. It is the connective tissue between actors.

That connective tissue includes several often-overlooked ingredients:

Shared language, so different groups can coordinate without translating every time.
Trust, so people share failures as well as successes.
Boundary spanners, who move between communities and carry learning across silos.
Feedback loops, so local experience informs strategy quickly.
Adaptive governance, so the network can change course without collapsing.

A funder who understands this does not ask only, “How many services were delivered?” They also ask, “How much learning moved across the system?” and “Did relationships become more capable of carrying future change?”

This is a harder kind of investment to justify in the short term, because it does not always produce immediate, visible outputs. But that is precisely why it is strategic. Networks create the conditions under which outcomes become repeatable instead of accidental.

There is a useful analogy here to gardening. You can pour water directly on a plant and see instant effects. But if the soil is compacted, the roots are shallow, and the surrounding ecosystem is degraded, the plant remains vulnerable. Funding isolated projects is like watering leaves. Funding networks is like rebuilding the soil.

A Better Mental Model: Change as a Learning System

The deepest synthesis here is that systems change should be treated less like program delivery and more like distributed learning.

This model solves a problem that neither pure funding logic nor pure evidence logic can solve on its own. Funding logic is good at accountability, but too often makes the world look simpler than it is. Evidence logic is good at rigor, but too often assumes that knowledge will travel by itself. A learning system combines both. It funds experimentation, but also funds the relationships and routines that let experiments inform each other.

In this frame, a network is valuable because it reduces the cost of learning. Instead of each organization discovering the same lesson alone, the network lets one site’s insight become another site’s shortcut. That means the system becomes smarter over time. Failures are not wasted, because they are shared. Successes are not random, because they are analyzed, translated, and adapted.

This is especially important in public health, where context changes constantly. New pathogens emerge. Demographics shift. Misinformation spreads. Policy landscapes move. Static solutions age quickly. A learning network is therefore not a luxury. It is the basic operating system for resilience.

The goal is not to create a perfect intervention. The goal is to create a system that can keep getting better under real-world conditions.

That requires a different kind of funder, one who sees themselves not as a buyer of outputs but as a designer of ecosystems. It also requires a different kind of measurement. Instead of only counting endpoints, we should also track the health of the network itself: who is connected to whom, how often knowledge travels, whether trust deepens, whether adaptation happens faster over time.

If that sounds abstract, imagine two emergency response systems. One is a collection of excellent but isolated units. The other is a web of responders who share protocols, communicate constantly, and can reroute resources quickly. In a crisis, the second system will outperform the first, even if the individual parts are less impressive on paper. That is because coordination is a capability, not an afterthought.

What Funders and Practitioners Need to Do Differently

If the real unit of change is the network, then both philanthropy and practice need to evolve. This does not mean abandoning outcomes or rigor. It means broadening what counts as strategic investment.

First, fund relationship infrastructure. Convenings, peer learning groups, community of practice meetings, and cross-sector partnerships can look soft from the outside. In reality, they are the channels through which adaptation becomes possible. Without them, every organization is left to reinvent the wheel in isolation.

Second, support boundary-spanning roles. Many systems fail not because no one cares, but because no one is paid to connect the dots. The person who translates between government and grassroots, or between clinicians and community leaders, often carries disproportionate value. These roles deserve stable support, not ad hoc gratitude.

Third, measure mechanisms, not just outputs. Did trust increase? Did knowledge move faster? Did local actors gain more ability to modify interventions? Did the network become less dependent on a single leader? These are not vanity metrics. They are leading indicators of whether change can survive contact with reality.

Fourth, accept that adaptation is not drift. In complex environments, a program that changes in response to context is often stronger than one that stays rigid. The challenge is to distinguish thoughtful adaptation from incoherence. That means defining the core principles of an intervention while leaving room for local design.

Finally, fund over a longer horizon. Networks do not mature on the timetable of a quarterly report. Trust grows slowly. Shared norms take repeated exposure. Learning compounds. If we want durable system change, we must tolerate the slower, less linear pace at which real capacity emerges.

Key Takeaways

Stop asking only which program works. Ask what relationships, incentives, and communication pathways make work possible in the first place.
Treat context as part of the intervention. A good idea can fail if the surrounding system is not prepared to carry it.
Fund connective tissue. Convening, coordination, translation, and trust building are not extras, they are infrastructure.
Measure learning, not just delivery. Track how quickly a network can absorb, adapt, and spread knowledge.
Design for adaptation. In complex systems, durable change comes from models that can evolve without losing their core purpose.

The Real Shift: From Heroic Fixes to Capable Systems

The most tempting story in social change is the heroic one: identify the right organization, the right intervention, the right evidence, and the problem bends. But the world is not usually changed by isolated heroes, however competent they are. It is changed by capable systems that can sense, learn, coordinate, and respond.

That is the deeper connection between networked funding and realist thinking about public health. Both reject the fantasy that transformation comes from applying a universal solution to a passive field. Both insist that outcomes are produced through interaction, not just intention. And both point toward a more mature idea of strategy: not controlling the world from above, but strengthening the conditions under which good action can travel.

If you want to change a system, do not only fund the answer. Fund the network that can carry the answer, test it, adapt it, and make it resilient enough to survive the next shock.

Because in the end, the most important question is not whether a solution exists. It is whether the system around it is capable of learning it.