Participatory Monitoring and Evaluation (PM&E)

• PM&E is a collaborative practice that actively involves local stakeholders in data generation, interpretation, and evidence-driven decision-making.
• It employs diverse methodologies—from digital apps to communal artifacts—to facilitate real-time monitoring, feedback loops, and learning cycles.
• Empirical studies highlight that while PM&E enhances local agency and accountability, stakeholder participation varies across project phases.

Participatory Monitoring and Evaluation (PM&E) denotes monitoring and evaluation processes that are collaboratively developed and implemented with key stakeholders, especially local communities and beneficiaries, so that knowledge production, interpretation, and adaptation are embedded in practice rather than separated as an external audit. In the contemporary cases documented across health information systems, citizen sensing, urban environmental governance, flood-risk management, donor-funded projects, and therapeutic making, PM&E is characterized by stakeholder involvement in data generation, feedback loops that connect evidence to action, and the use of monitoring artifacts, platforms, or meetings as sites of collective sense-making. The cases also show that participation is unevenly distributed across the project cycle: many systems are strongly participatory in monitoring, but weaker in indicator definition, formal evaluation design, or shared control over institutional response (Amuni et al., 4 Sep 2025, Russpatrick et al., 2021, Ahmed et al., 2017, Wolff et al., 2021, Parikh et al., 2024).

1. Conceptual foundations

PM&E is presented in the literature as a family of practices rather than a single standardized protocol. In donor-funded projects in Kisumu Central Sub-County, it is understood as “monitoring and evaluation processes that are collaboratively developed and implemented with key stakeholders,” where local communities and beneficiaries actively participate in needs assessments and baseline studies, ongoing monitoring, project reviews, meetings and workshops, and evaluations and outreach activities (Amuni et al., 4 Sep 2025). In Rwanda, the closely related formulation “evaluation for improvement” frames monitoring and evaluation as a continuous, formative process that systematically assesses current data-use situations, identifies shortcomings in practice and software, designs interventions, and intends re-evaluation after changes (Russpatrick et al., 2021).

Several cases do not explicitly use the PM&E label, yet align closely with its core logic. The Bangladesh participatory sensing framework defines participatory sensing as “the concept where individuals, groups and communities engages in the data collection actively to construct knowledge,” linking citizen-generated reports to expert decision making in environmental governance (Ahmed et al., 2017). The Communal Loom is explicitly designed to merge artistic co-creation with “recording data about patients’ well-being and perspectives,” thereby resisting the separation between intervention and evaluation in therapeutic settings (Parikh et al., 2024). The flood-monitoring work in Fiji and Indonesia is stronger on participatory monitoring than on fully co-owned evaluation, but its action-research cycle, feedback practices, and community use of data for advocacy and local decision-making place it on a clear PM&E continuum (Wolff et al., 2021).

A recurrent conceptual distinction concerns monitoring versus evaluation. Some systems provide robust participatory monitoring infrastructure but only embryonic evaluation procedures. The Bangladesh platform supports continuous citizen reporting, real-time mapping, and descriptive statistics, yet does not specify participatory interpretation workshops, agency performance reviews, or formal impact evaluation methods (Ahmed et al., 2017). The RISE flood case similarly shows strong community involvement in sustained data collection, while initial indicator definition and centralized data interpretation remain more researcher-led (Wolff et al., 2021). This suggests that PM&E is best treated as a graded configuration of participation across design, monitoring, analysis, interpretation, and response, not as a binary property.

2. Participation across the project cycle

The empirical cases show participation occurring at multiple stages, but with markedly different depth. In Rwanda, facility and district staff are involved in diagnosis, analysis, redesign, and intended reassessment through field visits, focus groups, observation of monthly coordination meetings, questionnaires, WhatsApp feedback, online workshops, and mock-up co-design with HISP Rwanda, HISP Tanzania, the Ministry of Health, and the global DHIS2 core team at the University of Oslo (Russpatrick et al., 2021). In the Pittsburgh air-quality case, residents shape the problem definition, identify needed indicators, host sensors and the camera, submit smell reports, generate animated smoke images, and use evidence in complaints, meetings, and regulatory confrontations (Hsu et al., 2018).

Participation is frequently strongest in implementation and weaker in early design. In Kisumu, stakeholders are most visibly involved through meetings and workshops, outreach activities, project monitoring, baseline assessment, needs assessment, and project evaluation, while project design has the lowest involvement at 3.9% (Amuni et al., 4 Sep 2025). In the RISE flood-monitoring project, researchers defined gauges and core monitoring protocols, but community members influenced safe and accessible locations, maintained the daily reporting practice, supported interpretation, and repurposed the messaging group for warning, mutual support, and advocacy (Wolff et al., 2021). The Communal Loom occupies an intermediate position: participants produce the primary data and physically encode it into the woven artifact, but the initial implementation does not state that participants directly co-designed the questionnaire, making it “partial” PM&E in the sense that data production and use are participatory while indicator definition is less so (Parikh et al., 2024).

These patterns expose a central PM&E misconception: frequent engagement does not automatically imply meaningful influence. The Kisumu study reports regular opportunities for stakeholder interaction and frequent forums, yet also concludes that stakeholders’ perspectives and opinions were “not diligently incorporated into programming as anticipated” (Amuni et al., 4 Sep 2025). The Bangladesh framework similarly envisages accountability feedback from authorities, but in the prototype the effective loop is mostly horizontal among citizens rather than vertical from authorities back to citizens (Ahmed et al., 2017). A plausible implication is that the analytical distinction between consultation, co-production, and co-governance is indispensable when evaluating participation claims.

3. Methods, artifacts, and data infrastructures

PM&E systems in these studies rely on heterogeneous monitoring media: tangible artifacts, mobile apps, dashboards, sensors, photographs, message groups, and hybrid evidence platforms. The Communal Loom couples a printed questionnaire, a “bubble sheet-like sheet that translates questionnaire responses into a weaving pattern,” rolls of yarn corresponding to possible responses, and a custom fixed heddle loom. Each participant contributes a row of eight woven segments corresponding to eight survey questions, formalized as

$$W(p) = \left(f_1(r_{p,1}), f_2(r_{p,2}), \dots, f_8(r_{p,8}) \right),$$

with the group artifact represented as an ordered list of participant rows separated by neutral separators (Parikh et al., 2024). The resulting woven scroll functions simultaneously as communal art and as a “machine-readable form of data record.”

Digital PM&E architectures are equally varied. The Bangladesh pollution framework uses a three-tier design: an Android application, a server for storage and computation, and a web application for city authorities. Citizens report incidents using built-in phone resources—camera, GPS, microphone, and storage—and authorities access real-time pollution maps and reports through the third tier (Ahmed et al., 2017). In Rwanda, the DHIS2 case centers not on data capture novelty but on the redesign of analytical use: monthly coordination meetings, Excel tables with color coding and comments, dashboard printing, and a proposed Data Form App that can create report formats matching Ministry of Health templates, share them, consolidate data, and export as XLS/PDF (Russpatrick et al., 2021).

Environmental PM&E cases often combine objective and subjective streams. The AirProbe International Challenge involved Air Ambassadors carrying a low-cost Sensor Box and using an Android app, while Air Guardians and Ambassadors jointly placed AirPins on city maps to estimate black carbon. Perception error was operationalized as

$$\text{APD} = \text{AirPin value} - \text{AirSquare value},$$

and temporal coverage for monitored spatial tiles was quantified via Shannon entropy,

$S = -\sum_{i=1}^{8} f_i \log_2 f_i.$

These metrics enabled evaluation of learning, participation patterns, and coverage quality (Sîrbu et al., 2015). The Pittsburgh system integrated animated smoke images, government and community PM2.5 sensors, crowdsourced smell reports on a 1–5 scale, wind visualization, and automatic smoke detection counts, thereby allowing residents to triangulate source visibility, exposure, and atmospheric transport (Hsu et al., 2018). The RISE flood system used fixed gauges and crest indicators photographed by residents and sent through a messaging app, yielding time-indexed water-level observations $h_i(t)$ for settlements and wet seasons (Wolff et al., 2021).

Taken together, these cases suggest that PM&E is not methodologically tied to surveys or workshops. It can be embodied, sensor-mediated, spreadsheet-based, map-centric, image-centric, or woven into everyday communal practice. What remains constant is the coupling of stakeholder action with interpretable data structures.

4. Evaluation logics and learning cycles

Evaluation in PM&E is often formative, recursive, and embedded in ordinary workflows. The Rwanda project explicitly adopts an action research cycle: identify and assess routine data-use situations, identify shortcomings and suggest improvements, implement and evaluate changes, then repeat the cycle (Russpatrick et al., 2021). The Communal Loom case uses observations during the session, an informal interview with the art therapist, and artifact analysis; the evaluation is primarily qualitative and exploratory, consistent with the pictorial format (Parikh et al., 2024). The RISE flood case is likewise reflexive and inductive, using fieldworker notes, structured discussions, open-ended questionnaires, and quantitative indicators such as photo counts and data completeness (Wolff et al., 2021).

Several studies formalize evaluation through mixed methods. The Kisumu study uses a Convergent Parallel design, collecting quantitative and qualitative data concurrently and integrating them to assess stakeholder involvement in M&E and project performance (Amuni et al., 4 Sep 2025). The Pittsburgh air-quality study combines server-log analysis with a survey assessing before/after changes in awareness, self-efficacy, and sense of community, testing differences with a right-tailed Wilcoxon signed-rank test (Hsu et al., 2018). The AirProbe study uses longitudinal phase comparisons and Kolmogorov–Smirnov tests to compare perception distributions and validate a stochastic transformation model of trust and partial opinion updating (Sîrbu et al., 2015).

A major evaluative theme is the feedback loop. In PM&E-aligned systems, data are not merely archived; they are returned to participants as prompts for action, redesign, or reflection. In Rwanda, WhatsApp groups and monthly coordination meetings function as venues where users discuss underperformance, request features, and shape redesign priorities (Russpatrick et al., 2021). In the Communal Loom, the final woven piece becomes a visible session record that can support reflection and conversation (Parikh et al., 2024). In Makassar, monthly flood reports with contributor recognition helped sustain motivation and made visible how community-generated data informed infrastructure design and local disaster-risk reduction (Wolff et al., 2021). Where those loops remain weak—as in the Bangladesh prototype, where authority-to-citizen response is not yet operationalized—the system remains stronger on participatory monitoring than on participatory evaluation (Ahmed et al., 2017).

5. Empirical domains and documented outcomes

The documented applications of PM&E span therapy, health systems, pollution governance, donor-funded development, and climate-risk management. In the Communal Loom deployment, a scroll woven by 28 Coler participants encoded responses to questionnaire items about the local public library and its use; the authors report that mapping survey questions to weaving patterns provided “an easy onramp” for people not used to designing their own weaving patterns and “made it easier for residents to share their feelings and to start new conversations around them” (Parikh et al., 2024). In Dhaka, the participatory sensing prototype was deployed for two weeks in December 2016, generated 53 data points, and found that 34% of incidents were related to garbage management; among 21 feedback respondents, 66.7% “Strongly Agree” and 28.6% “Agree” that the app would be able to bring effective changes in decision making (Ahmed et al., 2017).

In Rwanda, the central empirical finding is not lack of data but misalignment between formal platform analytics and situated work practices: DHIS2 dashboards and analytical tools were in limited use because district and facility users preferred Microsoft Excel for analysis, annotation, denominator management, and printing. Participatory redesign led to dashboard printing being prioritized in the core roadmap and to co-design work on a Data Form App (Russpatrick et al., 2021). In Kisumu, the mean score for “There are frequent forums where stakeholders are engaged in our work” is $\bar{x} = 4.05$ with SD $1.07$, while “Stakeholders contribute to the growth of this organization/project” has $\bar{x} = 3.79$ and SD $0.94$; yet only 3.9% of respondents identify stakeholder involvement in project design, and the study concludes that stakeholder involvement in M&E influenced project performance while also showing that incorporation of stakeholder views remained incomplete (Amuni et al., 4 Sep 2025).

Environmental citizen-science cases provide high-resolution evidence of both monitoring output and social learning. The AirProbe initiative collected 6,615,409 valid geo-localized sensor points and 70,758 AirPins, and reported that direct involvement in measurement activities rather than indirect information exposure can enhance learning and environmental awareness (Sîrbu et al., 2015). The Pittsburgh monitoring system logged 542 unique users and 1480 sessions from August 2015 to July 2016; the survey study found significant increases in self-efficacy and sense of community, and participants described the system as providing “uncomfortable” information that supported agonistic discussion with regulators (Hsu et al., 2018). The RISE flood project collected 5,301 photos from 26 community members in 13 informal settlements between 2018 and 2020, generating local flood records used to inform wetland and infrastructure design, health-risk analysis, and community advocacy (Wolff et al., 2021).

These outcomes indicate that PM&E can produce more than compliance data. It can generate communal artifacts, local evidence bases, design requirements, advocacy resources, and changes in psychological and organizational capacities. This suggests that PM&E outcomes are often distributed across epistemic, social, and political dimensions, not only program performance metrics.

6. Tensions, limitations, and contested issues

A persistent tension in PM&E concerns the balance between standardization and agency. The Communal Loom required fixed answer-to-yarn mappings to maintain analyzable structure, yet some participants found the process “overly mechanical and repetitive” and felt that it did not allow them to exercise their own creative independence (Parikh et al., 2024). The Rwanda case reveals an analogous tension in digital health systems: standardized core-platform development cannot satisfy all locally situated requirements, and the core team explicitly rejected the possibility of turning DHIS2 into “Excel inside the browser,” forcing trade-offs between global genericity and local app development (Russpatrick et al., 2021). In donor-funded projects, the same issue appears as the gap between regular consultation and actual influence over programming decisions (Amuni et al., 4 Sep 2025).

Data credibility, inclusiveness, and ethics are recurrent challenges. Open environmental reporting platforms risk false or misleading reports and rely on basic social verification mechanisms such as registration, support ratings, and future reputation systems (Ahmed et al., 2017). Citizen-science platforms can exclude those without smartphones, internet access, time, or digital fluency, as seen in Fiji and in the bandwidth concerns raised by users in Dhaka (Wolff et al., 2021, Ahmed et al., 2017). The flood-monitoring study explicitly warns that citizen science must not shift responsibility for flood management from government to vulnerable communities (Wolff et al., 2021). The Pittsburgh case similarly frames PM&E as adversarial and agonistic rather than consensual, emphasizing that participatory evidence is often generated under unequal power relations and used to contest institutional inaction (Hsu et al., 2018).

Another controversy concerns whether participatory monitoring should be equated with PM&E as a whole. Multiple cases caution against that simplification. The Bangladesh framework is a participatory monitoring platform with underdeveloped evaluative and accountability procedures (Ahmed et al., 2017). The RISE flood project evolves from a “citizens as sensors” model toward more collaborative use, but remains weaker on joint indicator definition and fully participatory evaluation (Wolff et al., 2021). The empirical record therefore supports a more discriminating interpretation: PM&E reaches its strongest form when stakeholders are not only data providers but also co-analysts, co-designers, and recognized claimants on institutional response.

Across these cases, the most durable lesson is that PM&E depends less on any single tool than on the quality of reciprocal relationships among monitoring, interpretation, redesign, and accountability. When data collection is embedded in meaningful activity, when findings are returned in usable form, and when stakeholders can influence what happens next, PM&E becomes a practical mechanism for learning, voice, and negotiated change. When participation is limited to reporting upward without visible response, it remains participatory in appearance but only partially participatory in effect.