When Change Refuses to Stay Linear: Why Good Theories of Change Need Feedback, Not Faith

May 28, 2026

10 min read

5 views

86%

The uncomfortable truth about change

What if the main reason our plans fail is not that they are too ambitious, but that they are too clean?

We like change to look like a bridge: input on one side, outcome on the other, with a neat sequence of steps in between. Funders ask for logic models. Managers ask for milestones. Teams ask for roadmaps. Yet the real world rarely behaves like a spreadsheet. People adapt, institutions push back, incentives mutate, and small interventions often produce effects that are either delayed, distorted, or completely different from what was intended.

That is the deeper tension at the heart of every serious attempt to change a system: we need causal direction, but we live inside complexity. If we oversimplify, we mistake a story for a strategy. If we overcomplicate, we lose the ability to act at all. The challenge is not choosing between certainty and chaos. It is learning how to think and work in a world where cause and effect are real, but rarely linear.

This is why the most useful theories of change are not blueprints. They are working hypotheses. They help us decide what to try, what to watch, and what to revise. And once you see them that way, change stops being a matter of proving that a plan is right and becomes a discipline of learning whether reality is cooperating.

Why linear plans break in living systems

A bridge works because steel and concrete do not revise themselves. Human systems do. The moment you introduce a new policy, program, or incentive, the system begins to respond. Teachers reinterpret the policy. Patients skip appointments for reasons no model predicted. Employees find workarounds. Communities absorb the intervention in ways that are shaped by history, trust, and power.

This is where many theories of change become fragile. They assume that if A leads to B, and B leads to C, then the chain will hold. But in complex environments, the chain is usually not a chain at all. It is a web of feedback loops, delays, thresholds, and side effects.

Consider a simple public health example. A campaign to increase vaccination might succeed in boosting appointments, but if clinics become overbooked and wait times lengthen, trust may erode. A well intended education reform might improve test scores in one district while narrowing the curriculum in another. A workplace productivity tool might increase output for a month, then quietly burn people out. In each case, the intervention works and fails at the same time, depending on which part of the system you measure and when.

In complex systems, success is not just about causing the desired effect. It is about not triggering the wrong counter effects.

This is the first major insight: causal thinking is still necessary, but it must become adaptive. A theory of change is not a prophecy. It is a map of assumptions that needs to be tested against terrain that keeps changing.

The hidden danger of certainty

One of the most seductive habits in planning is to confuse clarity with truth. A neat diagram can make an uncertain world feel governable. But clean diagrams have a cost. They can make us overconfident in what we know, and blind to what we do not.

The danger is not just technical. It is political and psychological. Once a theory of change hardens into doctrine, people stop using it as a learning tool and start using it as a justification machine. Evidence that fits is celebrated. Evidence that conflicts is explained away. The model survives, but only because reality is forced to bow to it.

That is why complexity matters. It restores a basic humility: the world is not obligated to match our preferred storyline. A good change strategy must therefore include not only a theory of how things should work, but a way of noticing when they are working differently.

Think of a ship captain crossing unfamiliar waters. A map matters, but so do the instruments, the weather, and the willingness to adjust course. The captain who treats the map as sacred is dangerous. The captain who throws it away is reckless. The skill lies in keeping the map open to correction.

This suggests a better standard for change work: not “Is the theory elegant?” but “How well does the theory help us detect surprise?” Elegance can inspire. Surprise detection keeps us honest.

A better mental model: theories of change as sensors, not scripts

The most powerful way to reconcile causal pathways with complexity is to stop treating theories of change as scripts for action and start treating them as sensing systems.

A script says: do step one, then step two, and the outcome should follow. A sensing system says: if we do step one, here are the signals we should watch for, here are the assumptions most likely to break, and here are the conditions under which we should adapt.

This shift changes everything.

Inputs become bets, not guarantees. When you fund a community initiative, launch a new service, or redesign a process, you are not buying a result. You are placing an informed bet that a chain of mechanisms may hold.
Outputs become evidence, not proof. A completed training session or a published policy tells you something, but not enough. The real question is whether behavior, relationships, and incentives changed in the way your model assumed.
Outcomes become moving targets. In a complex system, the outcome you want may shift as people respond. The goal is not fixed victory, but continuous alignment between intention and reality.
Failure becomes data, not disgrace. If the system surprises you, that is not automatically a sign that the effort was wrong. It may mean the model was incomplete, the context shifted, or a hidden variable mattered more than expected.

This is the essential upgrade from simplistic planning: design for learning, not just delivery.

Imagine planting seeds in different soils. A script says, water each one the same amount and expect the same result. A sensing approach says, observe moisture, sun, drainage, and growth patterns, then adjust. The second approach does not eliminate uncertainty. It gives you a way to work intelligently inside it.

Contribution analysis in a complex world: proving less, learning more

When change is messy, the question is often not “Can we prove causation beyond all doubt?” but “What can we plausibly attribute, and how confident are we?” That distinction matters.

Contribution thinking is valuable because many real interventions are not isolated experiments. They happen in environments where multiple actors are influencing the same outcomes. A nonprofit is not the only force affecting school attendance. A public agency is not the only force shaping housing stability. A company is not the only force affecting employee engagement.

In such settings, the smartest question is not whether we caused everything, but whether we contributed to a causal pathway that can be defended with evidence and reasoning. That requires tracing mechanisms, checking alternative explanations, and looking for patterns that fit the full context.

This is a more mature form of accountability. It rejects magical thinking on one side and impossible standards on the other. It says: we should be able to explain why this change likely happened, what else may have mattered, and how our assumptions held up.

There is a practical virtue in this modesty. If you insist on total proof, you often end up measuring too late. If you accept disciplined contribution, you can learn in time to improve the next iteration.

The goal is not to claim perfect authorship of change. The goal is to understand your share of the causal story well enough to act responsibly.

That is a profound difference. It turns evaluation from a courtroom into a laboratory.

The real question: what kind of truth does a theory of change need to tell?

A theory of change is often judged as if it were a prediction engine. But its better role is to act as a sensemaking device. It organizes what we think matters, makes assumptions visible, and gives us a basis for discussion when the system behaves unexpectedly.

This means the truth it tells is not binary. It is not simply right or wrong. It can be useful, incomplete, outdated, or dangerously confident. A theory can be directionally correct while missing a crucial delay. It can identify the right leverage point while misreading the mechanism. It can be elegant on paper and weak in practice.

The deeper question is not whether the theory matches reality perfectly. It is whether it helps us make better decisions under uncertainty.

Here is a useful test:

Does the theory identify the mechanism by which change is supposed to happen?
Does it make explicit the assumptions that could fail?
Does it specify what evidence would count as a sign of progress, drift, or unintended consequence?
Does it allow for iteration, or does it lock the team into defending the original design?

If the answer to these questions is yes, the theory is doing real work. If not, it is probably ceremonial.

A ceremonial theory of change is like a compass glued to a wall. It looks serious, but it cannot guide movement. A useful theory of change is more like a cockpit display: imperfect, partial, but designed to help pilots navigate while the weather changes.

From plans to adaptive loops

Once you see change this way, the unit of strategy is no longer the plan. It is the adaptive loop.

An adaptive loop has four parts:

Hypothesis: If we do X, mechanism Y should produce outcome Z.
Signal: What early signs should appear if the hypothesis is working?
Disconfirmation: What would tell us the hypothesis is wrong or incomplete?
Revision: How will we adjust based on what we learn?

This structure is powerful because it blends direction and flexibility. You still need a theory, otherwise learning is random. But you also need revision, otherwise theory becomes ideology.

Here is a concrete example. Suppose a city wants to reduce youth unemployment by offering training programs. A simplistic plan says: train more young people, and jobs will follow. An adaptive loop asks deeper questions. Do employers trust the credential? Are transportation barriers limiting attendance? Are young people dropping out because the schedule conflicts with care responsibilities? Is the issue truly skills, or is it network access, discrimination, or labor market mismatch?

Each of those questions suggests a different mechanism. Each mechanism suggests a different intervention. And each intervention should be monitored for both intended and unintended effects.

This is the practical heart of complexity aware change work: move from program certainty to mechanism curiosity.

Key Takeaways

Treat theories of change as hypotheses, not verdicts. Their job is to guide learning, not to protect ego.
Look for mechanisms, delays, and feedback loops. The most important effects in complex systems are often indirect.
Measure contribution, not total control. In real systems, you usually shape outcomes alongside many other forces.
Build in disconfirmation early. Decide in advance what evidence would make you revise the plan.
Use change strategies as sensing tools. The best frameworks do not just explain action, they improve your ability to notice when action is going off course.

The final shift: from certainty seeking to reality listening

The deepest lesson here is not that linear thinking is wrong. It is that linear thinking is only one instrument in the orchestra. It works best when the system is stable, the environment is predictable, and the causal chain is short. But many of the problems that matter most, from education and health to climate, governance, and organizational change, live in systems that learn back.

In those settings, the right posture is neither passivity nor arrogance. It is disciplined attentiveness. You still need direction. You still need a theory. You still need to act. But you must hold your theory lightly enough to notice when the world is telling a different story.

That is the real sophistication in change work: not having a perfect theory of the future, but having a better practice for meeting the unexpected.

A theory of change worth using does not promise control. It promises orientation. And in a complex world, orientation is often the difference between motion and progress.