Every development programme in Nepal follows a broadly similar programme cycle: problem identification, design, implementation and adaptation, with a monitoring and evaluation (M&E) framework as the dominant management and assessment tool. In M&E, a baseline survey establishes conditions before a project begins. Monitoring then tracks output against various indicators. At the end, an end-line survey assesses whether those targets were met, informing decisions about the next programme cycle. 

This system serves a clear organisational purpose by making development legible to donors. This article looks closely at how M&E, as a tool, allows the machinery to keep the programme cycle turning, regardless of durable change or unintended consequences. 

Indicators rely on the magic of numbers and the appearance of certainty and objectivity. Counting reshapes institutional attention towards what can be counted, without addressing what it cannot capture. This dynamic has driven Nepal’s M&E practice. However, indicators are never neutral instruments of measurement. What gets counted depends on which groups and organisations can afford to count; in Nepal, that has consistently been the donors and their implementing partners, not the communities being measured. 

Hence, with the deployment of statistical measures, depoliticisation operates through the technology of indicators. Moreover, this evaluation method systematically obscures the most important things. A monitoring framework can count the number of toilets built, but cannot tell you whether people are using them, whether they were built in the right location, or whether they were designed in ways that work for women. M&E systems rarely consider unintended consequences because of their structural features.

Moreover, the logical framework approach (the logframe), which structures most donor-funded project monitoring in Nepal, tracks outputs rather than outcomes because outputs are measurable within project time frames, whereas outcomes require longer-term, more complex assessment. A programme can meet every logframe target yet have minimal impact on the actual conditions it was designed to address. 

M&E systems have documented programme success, thereby legitimising its failure, rather than serving as tools meant to prevent it. Anthropologist David Mosse argues that development’s representations of success are socially produced. This framing often forces actors in the aid chain, including communities themselves, to present things in ways the system rewards.

Nepal’s post-earthquake housing reconstruction shows clearly what M&E counts and what it misses. It was a large owner-driven reconstruction programme, covering nearly 700,000 houses, and the reports cited the number of grant instalments paid out. That number rose steadily, suggesting recovery. However, the obvious measure of a finished house would have been the completion certificate, but so few were issued that The Asia Foundation’s independent monitors decided the figure would misrepresent housing reconstruction progress and used instalments instead. This shows how indicators were chosen based on availability rather than on purpose; and regardless of the problem being known, it was worked around rather than fixed. 

Meanwhile, Vivek Rawal and colleagues found that 18,505 households had been identified as vulnerable and promised an extra $500. However, the promise existed only on paper, and no one received the payment. As a result, families took on debt when the grant fell short. The M&E architecture could not register the unintended consequence of debt created by the programme’s design failure. 

Similarly, researchers have identified what is called the ‘accountability–learning dichotomy’—the structural tension between performance measurement as a tool for holding organisations accountable to donors and performance measurement as a tool for organisational learning. This tension has led to a power struggle between donors and organisations, which often results in organisations manufacturing a ‘culture of consensus’ to cover up the power imbalance. Donors also get trapped in the system, as they are sometimes reluctant to withdraw support from failing programmes for fear of being accused of abusing power. Hence, programmes continue not because the evidence supports them, but because withdrawing support has its own institutional and political costs.

Over the years, in response to criticism, the development sector has expanded its monitoring vocabulary from M&E to monitoring, evaluation and learning (MEL) to monitoring, evaluation, accountability and learning (MEAL) to monitoring, evaluation, research, learning and adapting (MERLA). The intent behind adding these terms was clear. ‘Learning’ was added to ensure that the findings would be useful for programme improvement; ‘accountability’ to ensure accountability towards the communities, not just donors; and ‘research’ to ensure that systematic evidence generation would be integrated with monitoring. Even as the acronyms changed, the structural conditions that produced M&E’s initial failures did not. The projects continued on short cycles, accountability continued towards donors, and quantitative outputs continued to be relied upon even in these new systems.

It is difficult to integrate ‘learning’ when the same organisations commission the evaluation of their programmes. It is difficult to integrate ‘accountability’ towards communities when there is no formal power to challenge the evaluation findings that might misrepresent their experiences. Likewise, integrating ‘research’ into evaluation becomes only a lip service when it lacks methodological rigour. Therefore, despite adopting new vocabularies, without institutional and methodological transformation, these changes are simply hollow. 

A re-orientation for development practitioners requires building monitoring, evaluation and learning systems that can understand and work with failure, and practitioners have been working around its limits. Some programmes have stopped treating evaluation as a fixed system of baselines and endlines. M&E consultants Florencia Guerzovich and Tom Aston call it ‘learning cadences’—pausing to reflect when the work actually calls for it, not when reporting demands it. Similarly, some have replaced concrete indicators with more fluid descriptions—programmes working slowly or working for some people and not others; while others have started counting what really matters—shifting relationships, observations of what people are actually doing, informal discussions, etc. 

However, there are limits to these adaptations as well, as they happen within existing systems. Perhaps adaptations need to take place against the very structure of programme cycles, and funders need to accept the limited impact of their money, given how fluid, messy and complex our social realities are. Otherwise, M&E will remain a perfectly engineered machine, endlessly turning and indifferent to change and consequences.