Qualitative MEL data analysis means systematically coding beneficiary interviews, focus group discussions, and Most Significant Change stories against the outcomes in a programme's Theory of Change, then aggregating themes across countries and languages for donor reporting. Most MEL guidance and software focuses on quantitative indicators and logframes. Tools like Skimle close the qualitative gap, with native multi-language analysis and full traceability from theme back to source quote.
Monitoring, Evaluation, and Learning (MEL, sometimes MERL when "research" is added) is the framework donors expect grantees to report against. A logframe sets out indicators. A Theory of Change explains the causal pathway from activities to outcomes. And underneath both sits a pile of qualitative data: beneficiary interviews, key informant interviews, focus group discussions (FGDs) in programme communities, and Most Significant Change (MSC) stories collected from participants in their own words.
The quantitative side of MEL has decades of tooling behind it: logframe templates, indicator tracking tables, DHIS2 dashboards, Kobo Toolbox forms. The qualitative side has had comparatively little. A MEL officer with 200 FGD transcripts from three countries, recorded in four languages, has to read, code, and synthesise that material largely by hand, then write it up against indicators that were designed with numbers in mind. This guide covers how that synthesis work actually happens, where it breaks down at scale, and how AI-assisted analysis tools fit into a MEL workflow without compromising the audit trail donors require.
What is MEL and how does it differ from MERL?
MEL stands for Monitoring, Evaluation, and Learning. Monitoring is the ongoing tracking of programme activities and outputs against a plan. Evaluation is the periodic, often more independent assessment of whether the programme achieved its intended outcomes and why. Learning is the deliberate process of feeding findings back into programme design and adaptation, rather than letting evaluation reports sit unread after submission.
MERL adds "Research" to the acronym, used by organisations that run a dedicated research function alongside routine monitoring, often producing case studies, impact stories, or academic-style outputs in addition to donor-mandated reporting. The terms overlap heavily in practice and many practitioners use them interchangeably. The MERL Tech community, active since 2014, has built an entire conference and practitioner network around the specific challenge of applying technology, including natural language processing, to monitoring, evaluation, research, and learning work in the humanitarian and development sectors.
A MEL system is built around a Theory of Change: an explicit statement of how a programme's activities are expected to produce outputs, which lead to outcomes, which contribute to longer-term impact. INTRAC's guidance on Theory of Change describes it as the framework that links short-, medium-, and long-term results into a single causal chain, which then determines what a MEL system needs to track at each level. Activities and outputs are usually counted: number of training sessions held, number of farmers reached, number of clinic visits. Outcomes and impact (changed behaviour, improved wellbeing, shifted attitudes) are much harder to count and are where qualitative data carries most of the analytical weight.
Why qualitative data is the harder half of MEL
A logframe indicator like "number of women trained in financial literacy" is straightforward to monitor. An outcome indicator like "women report increased confidence in household financial decision-making" requires asking people, in their own words, what changed for them and why. That is qualitative data, and it arrives in forms that resist the spreadsheet logic that quantitative MEL is built around:
- Beneficiary interviews: one-on-one conversations with programme participants, often conducted by field staff in local languages
- Key informant interviews (KIIs): conversations with community leaders, government officials, or service providers who can speak to programme effects beyond their own direct experience
- Focus group discussions (FGDs): structured group conversations, typically 6-10 participants, used to surface shared experiences and community-level dynamics
- Most Significant Change (MSC) stories: first-person accounts of the change participants consider most significant, collected and then selected through a panel review process
The Most Significant Change technique was developed by Rick Davies during his PhD fieldwork with the Christian Commission for Development in Bangladesh in 1994, and has since spread across development programmes in Africa, Asia, Latin America, and Australasia. It is explicitly participatory: stories are collected from beneficiaries, then reviewed and selected by panels (often including beneficiaries themselves) to surface the changes a community values most, rather than the changes a donor's logframe anticipated. That participatory strength is also what makes MSC data hard to process. Hundreds of stories, often translated and transcribed inconsistently, need to be read and synthesised before they can support a single line in a donor report.
tools4dev, a long-running resource for NGO M&E practitioners, makes a similar point in its guide to qualitative measurement: not every programme result can be counted, tracked, and fit neatly into an M&E framework, and methods like interviews, FGDs, and MSC stories exist precisely to capture the results that resist quantification. The guide is candid that these methods take real analytical effort. Reading and synthesising a single batch of MSC stories or FGD transcripts by hand, across a programme spanning multiple districts or countries, can consume weeks of an already stretched MEL team's time.
How much qualitative data does a typical MEL programme generate?
The scale varies by programme size, but it is rarely small. A multi-country programme running quarterly FGDs across five implementation sites, two rounds of beneficiary interviews per year, and an annual MSC story collection round can easily generate 300-600 qualitative source documents a year once transcribed. A single end-of-programme evaluation might require synthesising several years of accumulated qualitative material in a matter of weeks.
| Data type | Typical volume per round | Typical source format | Common languages involved |
|---|---|---|---|
| Beneficiary interviews | 20-100 per site | Audio recording, then transcript | Local/national language, sometimes translated |
| Key informant interviews | 5-20 per site | Notes or transcript | National language, sometimes English/French |
| Focus group discussions | 6-30 groups per round | Notes, audio, transcript | Local language, often multiple within one programme |
| MSC stories | 20-200 per collection round | Written or transcribed story | Mix of local languages and donor-reporting language |
This is the volume problem that most MEL guidance does not fully address. Standard MEL training covers how to design a Theory of Change, build a logframe, and select indicators. It says comparatively little about what happens when the qualitative half of that indicator set produces hundreds of documents that someone has to actually read, code, and turn into a defensible finding before the donor report is due.
How donors expect qualitative MEL findings to be reported
Donors including USAID and FCDO require grantees to report against the indicators in their approved MEL plan and Theory of Change, and increasingly expect qualitative findings to be presented with the same rigour as quantitative ones: a clear method, a defined sample, and a traceable link from finding to underlying data. A report that says "beneficiaries reported increased confidence" without being able to show how many beneficiaries, across which sites, using what method, is a weaker evidentiary basis than one that can say "34 of 52 women interviewed across four districts described a specific instance of independent financial decision-making, most commonly relating to school fees."
This is the same defensibility requirement that runs through public sector and regulatory work more broadly. Our guide on analysing stakeholder consultation responses covers the equivalent problem in policy consultations: a summary without traceability is not analysis, it is an assertion that cannot be checked. MEL reporting for donor audits faces the same test, often years after the original data was collected, when an auditor or evaluator wants to verify a claim made in an annual report.
Minimum practice for qualitative MEL reporting:
- Document the coding framework and how it maps to Theory of Change outcomes before coding begins
- Code consistently across sites, languages, and collection rounds, not just within a single round
- Keep every coded theme traceable to its source interview, FGD, or MSC story
- Report frequencies and patterns, not just illustrative quotes ("12 of 40 respondents" rather than "several respondents")
- Note where qualitative findings diverge by site, gender, age group, or other relevant programme metadata
Why translation has been the bottleneck
Most international development programmes collect qualitative data in the language participants speak, not the language the donor report will be written in. A beneficiary interview conducted in Swahili, Wolof, or Bangla in the field traditionally needed to be translated into English or French before a MEL officer based at headquarters could read and code it. That translation step is expensive, slow, and lossy. Academic research into multilingual qualitative methods has documented the conceptual equivalence problems that arise even with careful forward-and-back translation, where nuance specific to the original language gets flattened or altered in the process of rendering it into the reporting language.
For MEL teams working across several countries in a single regional programme, this multiplies. A programme spanning Kenya, Tanzania, and Uganda might need parallel translation pipelines for Swahili, local vernaculars, and English, each adding cost and turnaround time before analysis can even start. Skimle takes a different approach. Large language models process meaning across languages directly, so a project containing interviews in Swahili, French, and Amharic can be coded into a single, consistent set of themes without a separate translation step for every document. You can read more about how Skimle handles multi-language analysis, including how it keeps the analysis language-agnostic while still producing reports in whichever language the donor expects.
How Skimle supports qualitative MEL analysis
Skimle was built around the same traceability requirement that MEL reporting depends on: every theme it surfaces links back to the specific excerpt, and therefore the specific interview, FGD, or MSC story, that supports it. For MEL teams, the relevant features are:
Predefined categories mapped to your Theory of Change. The predefined categories analysis mode lets a MEL team define their outcome categories in advance, matching the structure of their logframe or Theory of Change, then apply that framework consistently across every document in the project. This produces a deductive analysis that reports directly against the indicators a donor expects, rather than a generic thematic summary that needs to be manually mapped afterwards.
Metadata by site, language, and respondent type. Programme documents can be tagged with country, district, gender, age band, or respondent type (beneficiary, key informant, community leader). The metadata analysis feature then lets a team compare how themes differ by site or demographic group, which is usually what a donor mid-term review actually wants to see: not "beneficiaries reported X" but "beneficiaries in District A reported X more often than District B, and here is the pattern by gender."
Native multi-language processing. Documents in different languages can sit in the same project and be analysed together, without a separate translation step before coding starts. Output reports are generated in whichever language the MEL team needs for the donor.
Traceable insights for audit. Every theme links back to the source document and the exact excerpt that supports it, which is the standard a MEL team needs to meet when a donor, auditor, or external evaluator asks to see the evidence behind a reported finding.
Export to office formats. The export to Word and Excel feature produces a structured output that slots into the standard donor report template, with quotes and source attribution intact, rather than requiring a separate manual write-up step.
For MEL teams running an inductive first pass on MSC stories or open-ended interview data before they have a fixed coding framework, the inductive analysis mode surfaces themes from the data itself, which is often the more appropriate starting point for participatory methods like MSC where the goal is to find out what beneficiaries value, not confirm what the programme expected.
What this costs, realistically
Budget is a real constraint in this sector in a way it often is not for corporate research teams. Programme budgets are typically restricted to the activities named in the grant agreement, and "qualitative analysis software" rarely has its own budget line. This is a similar constraint to the one academic researchers face, which our guide to qualitative research on a limited budget covers in more detail, including why a EUR 1,000-plus annual licence for a tool like NVivo is hard to justify when most of its feature set goes unused.
Skimle's free tier includes AI-assisted thematic analysis, which covers the basic need for many smaller NGOs and single-country programmes: load your interviews and FGD transcripts, run an analysis, and trace findings back to source. Larger or multi-country MEL teams that need predefined categories mapped to a Theory of Change, metadata cross-tabulation by site, or higher document volumes can move to a paid plan, but the entry point does not require an institutional procurement process or a multi-year licence commitment. For a sector where many organisations operate on tight, donor-restricted budgets, that matters more than feature breadth.
Frequently asked questions
What is the difference between MEL and M&E?
M&E (Monitoring and Evaluation) is the older, narrower term covering just the tracking and assessment of programme performance. MEL adds "Learning" explicitly: the deliberate process of using monitoring and evaluation findings to adapt programme design, rather than treating evaluation as a one-off compliance exercise at the end of a grant cycle. Most major donors, including USAID, now use MEL or MEL plan as the standard term in grant agreements.
How does qualitative data fit into a Theory of Change?
A Theory of Change sets out a causal chain from activities to outputs to outcomes to impact. Outputs (trainings held, materials distributed) are usually tracked quantitatively. Outcomes (changed behaviour, attitudes, or capacity) are harder to count directly and are typically assessed through qualitative methods: interviews, FGDs, and techniques like Most Significant Change that ask beneficiaries to describe what changed for them and why. A well-designed MEL plan specifies which outcome indicators rely on qualitative evidence and how that evidence will be coded and reported.
Can AI analysis be used for donor-facing MEL reports?
Yes, provided it preserves traceability. AI-assisted tools that link every reported theme back to its specific source interview or story are appropriate for donor reporting, because the underlying claim can be checked. Tools that produce a fluent summary without showing which documents support which finding create a documentation gap that becomes a problem if a donor, auditor, or evaluator asks to see the evidence behind a claim.
Do beneficiary interviews need to be translated before analysis?
Not necessarily, if the analysis tool can process the source language directly. Traditional workflows translated interviews into the donor-reporting language before coding, which is slow, costly, and introduces its own accuracy risks through conceptual loss in translation. Tools that use large language models can analyse text in the original language and still produce reports in the language a donor expects, removing the translation step from the critical path.
How many MSC stories or FGD transcripts should a MEL team expect to analyse per reporting cycle?
It depends heavily on programme scale, but a multi-site programme collecting quarterly FGDs and an annual MSC round can generate several hundred qualitative documents per year. There is no fixed rule for how many are "enough" to support a finding. What matters is being able to report the actual sample (how many interviews, from which sites, covering which respondent types) rather than an unspecified "beneficiaries said."
Ready to handle the qualitative side of MEL at the scale your programme actually generates?
Try Skimle for free and see how AI-assisted analysis handles beneficiary interviews, FGD transcripts, and MSC stories across languages with full traceability back to source. For higher-consideration deployments across multi-country programmes, you can also book a demo to talk through your specific MEL workflow.
Related reading:
- If your work spans public sector or regulatory consultation alongside MEL reporting, see how to analyse stakeholder consultation responses
- For more on Skimle's approach to mixed-language datasets, read analysing interviews and other documents in multiple languages
- If budget is the main constraint, our guide to qualitative research on a limited budget covers free and low-cost options in more depth
- For the broader public sector and NGO audience, see how Skimle fits public sector and policy work
About the authors
Henri Schildt is a Professor of Strategy at Aalto University School of Business and co-founder of Skimle. He has published over a dozen peer-reviewed articles using qualitative methods, including work in Academy of Management Journal, Organisation Science, and Strategic Management Journal. His research focuses on organisational strategy, innovation, and qualitative methodology. Google Scholar profile
Olli Salo is a former Partner at McKinsey & Company where he spent 18 years helping clients understand the markets and themselves, develop winning strategies and improve their operating models. He has done over 1000 client interviews and published over 10 articles on McKinsey.com and beyond. LinkedIn profile
Sources
- Most significant change technique - Wikipedia
- 5 ways to measure qualitative results - tools4dev
- Monitoring, evaluation and learning: a toolkit for small NGOs - INTRAC
- Theory of Change - Monitoring and Evaluation Planning Series 16 - INTRAC
- MERL Tech community
- Official Development Assistance falls in 2024 - Focus 2030