Most theories of change we inherit at the start of an evaluation engagement are flowcharts. They have boxes for activities, outputs, outcomes and impact, arrows running from left to right, and a colour scheme that suggests coherence. The diagram sits on slide nine of the proposal deck, and a board has approved it.
In most cases it is useless as a guide to evaluation, because it commits the programme to nothing that evidence could disconfirm.
A decorative TOC is easy to recognise. Every arrow points up and to the right. It has no branches, no failure modes and no assumptions stated specifically enough to be wrong. The boxes name categories such as "improved livelihoods" and "strengthened systems" and leave out the mechanisms. Taken at face value, the diagram claims only that doing things will produce other things, which is true of most activity.
A useful theory of change names the mechanism through which the activities are supposed to produce the outcome, identifies the assumptions that mechanism depends on, and says what evidence would show an assumption to be false. One test is whether the team can write down, before fieldwork, a finding that would force a revision of the theory. If it cannot, what it has is a programme description.
Most TOCs fail at the mechanism. A livelihoods programme's flowchart that reads "training → enterprise formation → income gains" hides three or four assumptions that do the causal work: that participants have access to start-up credit, that local markets can absorb additional supply at non-trivial margins, that household labour allocation permits sustained enterprise time, and that the trained skill, and no other factor, is the binding constraint on enterprise formation. Each of those assumptions is testable. None of them appears in the diagram.
This matters because the TOC should be the advance statement of what evidence would update the programme's beliefs. Evaluators who arrive after baseline are either testing the theory or producing post-hoc narratives that fit whatever the data showed. Which of the two they are doing depends on whether the theory was specific enough beforehand to conflict with at least some possible findings.
Contribution analysis, done properly, imposes this discipline. The method requires writing down the alternative explanations for any observed outcome (secular trends, rival programmes, selection effects, measurement artefacts) and specifying what evidence would distinguish each of them from the programme's claimed contribution. Programme teams find this uncomfortable because it makes them admit that other explanations are plausible. That admission is what makes the theory testable.
Decorative TOCs also have a political use. They signal coherence to funders without committing the programme to anything specific. A board approves the diagram more easily than it would approve a paragraph that begins, "We are betting that the binding constraint on women's enterprise formation in this district is a credit gap, and that local demand is sufficient, and if we are wrong about that, the programme will fail."
We advise programme teams to write the theory of change as prose paragraphs first and draw the diagram afterwards. Prose makes the team commit to mechanisms that a diagram lets it pass over. The paragraphs should name the binding constraint the programme believes it is addressing, the assumptions under which addressing that constraint produces the outcome, and at least two findings from baseline or early implementation that would force a revision.
This takes longer than drawing a flowchart, and the extra time is where the value comes from. A theory written down in fifteen minutes reflects fifteen minutes of thinking. One that went through three rounds of argument within the programme team reflects considerably more.
We should add a caveat. Theories of change written as falsifiable claims are easier to evaluate and harder to fund. Funders who say they want rigorous evaluation often respond to specific, falsifiable theories with discomfort, because such theories make programme failure visible. We do not have a clean answer to this tension. Still, choosing between a TOC that can be wrong and a TOC that cannot be evaluated amounts to choosing whether the programme intends to learn from evidence at all.
Useful background reading: BetterEvaluation on contribution analysis, John Mayne's original CGIAR paper on contribution analysis, and the ActKnowledge ToC Basics primer.