Why Good Theories of Change Fail Before They Begin

The hidden problem is not measurement, it is belief

Most plans do not fail because people forgot to count something. They fail because the plan quietly depended on a set of beliefs nobody had fully examined. A strategy can be beautifully written, logically sequenced, and full of indicators, yet still rest on a fantasy: that people will behave as expected, that institutions will cooperate, that incentives are aligned, that a small intervention will travel cleanly through a messy world.

That is the real tension at the heart of any serious attempt to explain change. We want causal clarity, but the world gives us uncertainty. We want a crisp chain of logic, but real systems respond through human judgment, politics, chance, and adaptation. The challenge is not simply to prove that an intervention worked. It is to understand the fragile web of assumptions that made success possible in the first place.

A theory of change is often treated as a diagram. In practice, it is a wager.

The quality of a plan is not determined by how elegant its logic looks on paper, but by how honestly it confronts the assumptions that keep the logic alive.

That shift matters because it changes the basic question. Instead of asking, “What activities lead to what outcomes?” we should ask, “What must be true for this chain of change to hold, and how likely is it that those truths survive contact with reality?”

Why causal logic is really a map of assumptions

Any causal pathway is only as strong as the assumptions beneath it. If a training program is supposed to improve service delivery, the pathway may assume that participants attend, that the content is relevant, that they can apply it, that their managers support them, and that the surrounding system does not punish experimentation. Each step sounds obvious until it breaks. Then the whole elegant theory starts to look less like a mechanism and more like a stack of maybes.

This is why many projects are overconfident at the start and confused at the end. They confuse a sequence of intended effects with a guarantee of actual effects. But causal chains in social settings are never self executing. They depend on trust, timing, coordination, interpretation, and context. In other words, they depend on assumptions that are often invisible precisely because they seem too mundane to question.

A useful way to think about this is to distinguish between causal logic and causal confidence.

Causal logic asks whether the pathway makes sense in theory. Causal confidence asks how much uncertainty remains because one or more assumptions may fail. The first is about coherence. The second is about risk.

A good theory of change does not pretend to eliminate uncertainty. It makes uncertainty legible.

Consider a public health campaign designed to increase vaccine uptake. The causal logic may look simple: messages increase awareness, awareness increases intent, intent increases uptake. But the assumptions are doing most of the work. People must trust the messenger. They must have access to the vaccine. The message must not trigger backlash. Competing rumors must not dominate the attention economy. A single hidden assumption, such as “the community sees official messages as credible,” can determine the fate of the whole program.

The point is not that theory of change is wrong. The point is that its real power appears only when it is used as a tool for surfacing assumptions, not just plotting outcomes.

The real work happens in the gaps between steps

When a strategy fails, the failure is rarely dramatic at the exact point the plan predicts. More often, it slips through a gap between steps. That gap is where assumptions live.

Imagine a workforce development initiative. Step one is recruitment, step two is training, step three is job placement. On paper, this seems orderly. But the real question is not whether the steps were executed. It is whether the transitions between steps were plausible. Could the participants attend regularly while juggling caregiving and transportation barriers? Did training lead to credentials employers valued? Did employers trust the pipeline? Did participants have enough support after placement to avoid early dropout?

Each transition is a test of an assumption. Each assumption is a hinge. When the hinge breaks, the door does not merely move slowly. It stops opening.

This is why high quality causal work should look less like proving a straight line and more like stress testing a bridge. Bridges do not fail because engineers drew a line from one bank to another. They fail when load, weather, material fatigue, and unexpected forces expose weak points. A theory of change should be treated the same way. The critical task is not to admire the path, but to identify where the path is most vulnerable.

That leads to a more useful mental model: the assumption stack.

An intervention rests on several layers of assumptions, often in this order:

1. Behavioral assumptions: Will people do what the theory expects?
2. Capability assumptions: Do they have the skills, time, authority, or resources?
3. Institutional assumptions: Will organizations support the desired behavior?
4. Contextual assumptions: Will the external environment remain sufficiently stable?
5. Interpretive assumptions: Will stakeholders understand the intervention in the intended way?

The higher you go in the stack, the more invisible, political, and fragile the assumptions become. Many plans focus on the lower layers because they are easier to observe. Yet the upper layers often determine whether the effort succeeds at all.

Contribution analysis is really an argument about plausible influence

This is where contribution thinking becomes so valuable. In complex settings, we usually cannot prove direct causation the way we might in a lab. Social change does not arrive with a receipt. Instead, we ask whether a plausible, evidence backed story connects the intervention to the observed outcomes.

That sounds modest, but it is actually rigorous. It forces us to move from certainty theater to disciplined judgment. Did the intervention plausibly contribute to change, and what else might explain it? Were the expected intermediate effects observed? Did the context behave as anticipated? Which assumptions held, which broke, and which were never tested?

This approach changes evaluation from a verdict into an investigation.

That matters because many organizations make a category error. They treat evaluation as an end point, a final scorecard. But in a complex system, the most valuable evaluation is not the one that pronounces success or failure. It is the one that reveals which assumptions were carrying the weight of the design. In that sense, contribution analysis is not merely about accountability. It is about learning where reality negotiated with intention.

Think of a nonprofit trying to reduce school absenteeism. If attendance improves after the intervention, the temptation is to celebrate. But a contribution lens asks a deeper question: was the improvement driven by the program, or by a new transportation policy, a change in weather, or a shift in family routines? More importantly, which parts of the theory actually mattered? Perhaps the text reminders worked, but only because the school also improved breakfast access. Then the intervention was not a self contained cause. It was part of a causal ensemble.

That is the kind of insight that makes future design smarter. It tells you not just whether something worked, but under what conditions it worked, and what hidden supports it needed.

The best causal explanation is not the one that eliminates complexity. It is the one that reveals which complexities are doing the real work.

A better way to work with assumptions

Once you see assumptions as the skeleton of a theory of change, the next step is not to eliminate them. That would be impossible. The goal is to work with them intelligently.

The first move is to name assumptions explicitly. Vague confidence is the enemy of learning. If a team cannot state what must be true for its plan to work, then it cannot know what to monitor. Making assumptions visible turns hidden fragility into an operational issue.

The second move is to rank assumptions by risk. Not every assumption deserves equal attention. Some are foundational, some are peripheral, and some are merely convenient. Ask two questions: How likely is this assumption to fail? And how catastrophic would failure be? The intersection of those two answers tells you where to focus.

The third move is to design tests, not just activities. Too many plans add activities without adding learning. If a strategy depends on community trust, create a way to observe trust early. If success depends on partner coordination, monitor whether coordination is actually happening. If uptake depends on usability, test whether people can use the thing without assistance. A good plan includes probes for its own weak points.

The fourth move is to treat surprises as diagnostic evidence. When an assumption fails, do not immediately interpret that as a failure of execution. Sometimes it is evidence that the original theory was too thin. The surprise is not noise. It is information about the shape of the system.

Here is a practical way to use this approach:

• Write your core pathway in one sentence.
• Underline every unstated belief required for that sentence to be true.
• Mark each assumption as strong, uncertain, or fragile.
• For the fragile ones, decide what evidence would show they are holding or breaking.
• Revisit the list after implementation, not just before launch.

This process changes the role of leadership too. Leaders are often rewarded for decisiveness, but in complex change work, the better skill is calibrated doubt. Not paralysis. Not cynicism. Calibrated doubt is the ability to say, “Here is the path we think will work, here are the assumptions that make it possible, and here is how we will know if reality disagrees.”

The deeper lesson: change is a negotiation with reality

The most powerful insight that emerges from combining causal pathways with assumption work is this: change is not something you impose on the world. It is something you negotiate with the world.

That reframes everything. A strategy is no longer a command. It is a conversation with context. An intervention is not a magic lever. It is an offer to a system that may accept, modify, delay, or reject it. A theory of change is not a promise. It is a disciplined articulation of how you think the negotiation might unfold.

This is why the most mature organizations are not the ones with the most elaborate diagrams. They are the ones that can say, with some humility, “Here is where our theory is strong, and here is where it is most exposed.” They know that clarity is not the same as certainty. They know that causality in the social world is often conditional, indirect, and co produced by actors who are responding to the intervention while simultaneously reshaping it.

That is a harder story than linear success. But it is a truer one, and truer stories lead to better action.

Key Takeaways

1. Treat every theory of change as a set of assumptions, not just a chain of outcomes. If you cannot name the assumptions, you cannot manage the risk.
2. Focus on the transitions between steps. Most failures happen in the gaps, where behavior, institutions, and context must align.
3. Separate causal logic from causal confidence. A pathway can make sense and still be fragile.
4. Use evaluation to learn which assumptions held. Good contribution work clarifies plausible influence and exposes hidden dependencies.
5. Design tests for your weakest assumptions early. Do not wait for the final report to discover what reality was trying to tell you.

Conclusion: stop asking whether the plan worked in theory

The deepest mistake in change work is not optimism. It is unexamined optimism. We often ask whether our plan is logically sound, as if logic alone could protect us from the world. But the world does not reward elegance. It rewards designs that can survive the collapse of at least a few assumptions.

So the better question is not, “Is this theory of change correct?” It is, “What must be true for this theory to work, and how will we know if those truths are breaking down?” That question turns planning into inquiry and evaluation into learning.

In the end, the value of a theory of change is not that it predicts the future. It is that it makes the future discussable. And once assumptions become discussable, change becomes more honest, more adaptive, and far more likely to happen for real.