What Is a Theory of Change?

A theory of change explains how an intervention is expected to produce a result and under what conditions. The Medical Research Council framework presented by Skivington and colleagues (2021) calls for spelling out that explanation, testing it against evidence, and refining it as the intervention progresses. If a first trial reveals a step the team missed, the team can revise both the intervention and its explanation of how the intervention works.

A theory of change often takes the form of a chain of outcomes. Each link states an assumption the team can examine. Mayne (2017) proposes asking what has to be true for one intermediate outcome to lead to the next and identifying any missing conditions. The team needs enough detail to know where to look and what would change its decision.

How Do You Turn a Business Goal into a Testable Hypothesis?

Start with the business outcome and an action that could contribute to it. “Improve quality” points in a direction. A drop in the number of applications returned for missing documents gives the team a result it can track. To identify an observable practice, the team asks who checks the documents, when, and what information they need. It can then investigate barriers to that check with the COM-B model.

Once the team chooses an intervention, it explains what should change at the point of action. Making information available in the tool people already use can cut the time they spend looking for it. That time saving is an intermediate effect the team can check. It then looks at whether people take the expected action more often and whether that shift comes soon enough to affect the business outcome. In the 4T model (in French), working through that explanation helps the team design the test before it intervenes.

What Does a Theory of Change Look Like in Billing?

Back to the fictional billing example. For standard contracts, the portal pulls together recorded hours and billable work into a summary the project manager checks. The hypothesis is that reviewing the summary will take less time than assembling it from several tools. The project manager will then have more time to send it to the client by the first business day after the month-end close. The company expects more invoices to go out on time if clients sign off early enough for accounting to process the approvals before its deadline.

The team has several points to check. A summary might be available sooner and still need extensive corrections. Earlier delivery might also leave approval dates unchanged if clients review the summaries on a fixed schedule. At each step, the team connects the expected effect to a condition and to something it can measure.

What Should Change What It Depends On What the Team Measures
Implementation
The portal makes a prefilled summary available to the project manager.
The data reach the portal at the close. The project manager can access the summary. Share of standard contracts with a summary available on time.
Intermediate effect
Preparing the summary takes less time.
The prefilled data are accurate enough and exceptions are easy to spot. Time spent preparing and correcting summaries before sending them, measured across comparable cases.
Target behavior
The project manager sends the summary to the client by the first business day after the month-end close.
The time savings occur before the deadline, and the project manager has time to check and send the summary. Share of all expected summaries actually sent by the deadline.
Approval
The client approves the summary before accounting’s processing cutoff.
The client receives the summary in time and can approve it earlier under its review schedule. Share of all expected summaries approved before the processing cutoff.
Business outcome
More of the corresponding invoices go out on schedule.
Accounting can process the approved summaries in time. Other required documents are available. Share of all expected invoices issued by the scheduled deadline.

Which Assumptions Should You Test First?

Start with assumptions whose answers could change what happens next. Some concern basic conditions for delivering the intervention. If the people expected to use a feature lack access, the immediate question is how it was rolled out. Other assumptions concern behavior. If people have access and the feature saves time, but the action is still rare, the team can go back to the work and find out what is taking priority.

The pilot also tests the link to the business outcome. People might change a practice while another step still determines the overall turnaround time. HM Treasury’s Magenta Book (2026) calls for examining competing explanations. Better results in pilot teams might coincide with additional staffing or lower demand. Tracking those changes helps the team interpret the results. Where feasible, following comparable teams that temporarily keep their usual process can help assess how much of the change is attributable to the intervention.

How Should You Read the Pilot Results?

An implementation measure shows whether the intervention reached people under the conditions planned. A behavioral indicator (in French) shows whether they took the expected action in the situations the team chose to observe. The business measure shows whether the outcome is moving in the desired direction. Taken together, these measures help the team tell the difference between a rollout problem and a design that needs work even when delivery goes as planned.

If the measures move as expected at each step, the team has evidence consistent with its explanation. How much it can conclude about the intervention’s contribution also depends on the quality of the data, the comparisons available, and what else changed during the pilot. The contribution analysis approach described in the Magenta Book (2026) weighs those factors. The team revises the links that did not hold and records the conditions in which it saw a result.

Where Does the Theory of Change Fit in MEET©?

In the MEET© protocol, the team starts connecting behaviors to the business outcome in Model. It refines that explanation in Engineer, after observing the work and choosing the intervention it wants to try. That is when the team decides which assumptions to test and what evidence to gather.

In Experiment, the team checks the theory against what it observes and adjusts the intervention as needed. Transfer raises a further question. Will the conditions that supported the change during the pilot be there in day-to-day work? The person responsible revisits those conditions before expanding the practice to other teams or reducing pilot support.

Frequently Asked Questions

What is the difference between a theory of change and a logic model?

A logic model typically maps a program’s resources, activities, and outcomes. A theory of change explains the proposed causal links and the conditions they rely on. One diagram can do both if it shows why each effect is expected to happen.

Who should help build a theory of change?

Include the person accountable for the business outcome, the people whose work is expected to change, and those who will put the intervention in place. People on the receiving end of the action can also explain what has to happen before it affects the next outcome. They are often the ones who can spot a dependency that is hard to see in the project schedule.

How much detail is enough?

Describe the steps where new evidence might change a decision. If teams use different tools or work under different rules, spell out those conditions when they could change the expected effect. A short first version is enough to organize a pilot and assign responsibility for collecting observations.

References

HM Treasury and Evaluation Task Force (2026). Magenta Book: Central Government guidance on evaluation. Guide and Annex A, analytical methods

Mayne, J. (2017). Theory of change analysis: Building robust theories of change. Canadian Journal of Program Evaluation, 32(2), 155-173. 10.3138/cjpe.31122

Skivington, K., Matthews, L., Simpson, S. A., et al. (2021). A new framework for developing and evaluating complex interventions: Update of Medical Research Council guidance. BMJ, 374, n2061. 10.1136/bmj.n2061

About the Author

Maxime Rabéchault is the founder and president of UNREST, a change enablement consultancy. He applies behavioral science to organizational transformation and leadership development.