Realist Evaluation
Examine for whom, under what circumstances and why change occurs through hypotheses linking context, mechanisms and outcomes. Explore why the same program produces different results. [1]
- Difficulty
- Advanced
- Cost
- High
- Duration
- Illustrative estimate (adjust for scale and staffing): 2–4 weeks for literature review and CMO hypotheses; several months to 1 year or more for empirical work
How it works at a glance
Realist evaluation examines people’s responses to resources and opportunities provided by a program. Develop hypotheses linking context (C), mechanism (M) and outcome (O), and examine them using cases and data from different conditions. [1]
A mechanism is not an activity label; it combines opportunities provided with participants’ interpretations and responses. Use analytical findings to revise initial hypotheses and specify conditions needed for implementation elsewhere. [1]
When to use it and when it is unsuitable
When to use it
- Explaining why a program worked in one setting but not another
- Assessing transferability when considering implementation or scaling in different contexts
- Systematically learning about effective mechanisms from existing evaluations and studies (realist synthesis)
- Refining program theory and improving intervention design
When it is unsuitable
- When proving average intervention effectiveness is the purpose rather than explaining mechanisms; RCT or DiD is appropriate
- When prior knowledge and literature are insufficient for CMO hypotheses and exploratory research is needed first
Participants, roles, and deliverables
Participants and roles
Evaluation researchers (consider expertise in both quantitative and qualitative approaches), program theory specialists and frontline implementers. An illustrative plan assigns literature review to 1–2 researchers.
Main deliverables
- CMO hypothesis map
- Refined CMO configuration list
- Analysis of effects by context
- Transferability and adaptation conditions report
Implementation steps
- Reframe research questions as CMO questionsReframe “Does this program work?” as “In what contexts, through which mechanisms, and with what outcomes?” This determines the overall evaluation direction.
- Develop initial CMO hypothesesUse literature, previous evaluations and implementer interviews to propose several CMO configurations. State them as “If this mechanism operates in this context, this outcome is expected.”
- Design data collectionDesign collection to examine and revise each CMO element. Include contextual variables, proxies for mechanisms such as cognitive or emotional change, and diverse outcomes.
- Test and refine CMO hypothesesCompare data with hypotheses to analyze whether C → M → O holds and in which contexts it does not. Repeatedly revise, reject and refine hypotheses.
- Refine configurations and assess transferabilitySummarize tested and refined configurations and identify contexts where mechanisms operate. Specify transferability and necessary adaptations.
- Recommend policy and practice changesUse CMO analysis to recommend contexts for implementation and what is needed for mechanisms to operate.
Strengths and limitations
Strengths
- Explaining mechanisms provides deeper understanding of why something works beyond whether it works
- Analyzing context specificity can reduce failure risks during scaling or transfer
- Applies rigorous causal thinking to complex social interventions where RCTs are difficult
Limitations
- CMO hypothesis development requires advanced conceptual and theoretical skills
- Empirically measuring mechanisms is difficult and relies heavily on proxies
- Unsuitable for explaining average-effect estimates to donors seeking numbers
Common pitfalls
- Being satisfied with a CMO configuration and neglecting data collection to test it
- Confusing mechanisms with program activities such as delivering training. Mechanisms involve reasoning and resources in relation to the program (participants’ cognition, emotions and choices)
- Using only broad categories such as urban / rural rather than identifying specific contextual factors affecting mechanisms
Conditions of use and evidence
- Consider mechanisms as combinations of program resources or opportunities and people’s responses to them. [1]
Checking conditions for applying training at work
The following fictional example compares organizations that did and did not use plans after evaluation planning training.
Scroll horizontally to view the table
| Hypothesis number | Context (contextual conditions) | Mechanism (changes in cognition, emotions or choices) | Outcome (expected result) | Rationale and sources |
|---|---|---|---|---|
| H1 | Improvements can be proposed at regular meetings | Seeing the training draft creates confidence that it can be explained | Staff bring the plan to a meeting and determine items | Untested hypothesis for this example |
Common errors in judgment
Writing “2 training sessions delivered” under M
How to revise
Describe how exercises affected participants’ judgments
ReasonThis distinguishes implementation activities from mechanisms.
Applying one successful organization’s CMO to all organizations
How to revise
Examine different conditions and non-use cases and revise hypotheses
ReasonThe method examines how responses vary by context.
References and checked sections
- Supplementary Guide: Realist evaluationHM Treasury / Evaluation Task Force · 2026
- Section consulted
- 3. Developing realist theory: Context, mechanism and outcome; 5. Applying the realist frame to evaluation
- How it relates to the text
- Consider mechanisms as combinations of program resources or opportunities and people’s responses to them.
Editorial and checking information
How to cite this page
Change history
- — Updated explanations to correspond to the sources and added method-specific fictional completed examples, judgment examples and a change log.


