> For the complete documentation index, see [llms.txt](https://eval.playbook.org.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://eval.playbook.org.ai/social-impact/how-to-evaluate/process-evaluation-why-arent-outcomes-changing.md).

# Process Evaluation: Why Aren’t Outcomes Changing?

Level 4 impact evaluations reveal the intervention’s effect but may not explain the mechanism behind it. A PE ([see primer here](/level-linkages/linkage-across-levels/process-evaluations.md)) could offer an explanation. When impact is limited, it could identify if implementation was weak. When impact is strong, it can identify conditions that made it possible which should be replicated when scaling.

<table><thead><tr><th width="137.19921875">ToC Domain / Assumption</th><th width="391.46484375">Example PE questions</th><th>Methods</th></tr></thead><tbody><tr><td>User-to-beneficiary pathway</td><td>When the user is a frontline worker, is the intended beneficiary (e.g. patient, student, farmer) actually receiving and able to act on the information? What are the barriers and facilitators of a strong user-to-beneficiary pathway?</td><td>Interviews with beneficiaries; observation of frontline worker-beneficiary interactions</td></tr><tr><td>System readiness and sustainability</td><td>What institutional enablers support or constrain the integration and use of AI-enabled products within a broader delivery system? Are there bottlenecks outside the product (in supply chains, health systems, or institutional capacity) that mediate impact?</td><td>Document review; stakeholder interviews; review of administrative and information system data</td></tr></tbody></table>

In a level 4 evaluation, process data collected across study sites can also be used directly in the analysis process to generate richer insights on impact via:

* **Mediation and subgroup analysis**. If you have sufficient variation in implementation across study units and sample sizes, evaluators can test whether impact concentrates among subgroups defined by implementation quality or contextual factors.
* **Treatment-on-the-Treated analysis** with varying implementation intensity. In practice, not all treatment units might receive the intervention as designed. Process data capturing implementation fidelity or intensity allow evaluators to move beyond the intention-to-treat estimate and conduct ToT analysis that uses randomization as an instrument for actual exposure or treatment intensity.<br>

***

<details>

<summary>💬 Want to suggest edits or provide feedback?</summary>

{% embed url="<https://tally.so/r/A788l0?originPage=level-4-impact-evaluation%2Fhow-is-level-4-evaluation-performed%2Fprocess-evaluation-why-arent-outcomes-changing>" %}

</details>