NGO case studies prove their impact, and get referenced in Google’s AI Overviews or a ChatGPT answer, when they name a person, give a dated starting point, and link to a source a reader can check. Full stop. So, why do so many not do this? It’s a shame and a loss these days when every bit of valuable humanitarian work needs its results maximized and publicized.
A fundraiser needs those details to recount the story and a journalist needs them to verify it and make it newsworthy to as big an audience as possible. We reviewed 50 online English-language case studies from 10 mid-to-large international organizations – including the World Bank, UNDP, and CARE.
We found that most of these studies reported a number without showing what the work actually changed. That gap costs a case study its second life: in a donor report, media pitch,partner brief, or AI-generated answer.
| What we measured across the 50 case studies | Share |
| Used numbers to describe a result | 78% |
| Gave evidence the work caused that result | 24% |
| Included a named person the work aimed to help | 28% |
| Stated or signaled a limit | 44% |
| Linked to a report, study, or evaluation | 66% |
| Gave a useful next step beyond “subscribe” or “donate” | 64% |
Contents
Why do most NGO case studies fail to prove their results?
Most fail because they report scale, not change. Only 12 of the 50 case studies (24%) gave enough evidence to show the featured work caused or helped produce its result, and the other 38 counted activity, such as people reached or training delivered, without a starting point to measure it against. The strongest entries named a person or a community, gave a dated starting point, separated the work from the change, and linked to a source.
The European Commission’s own results-based management framework for international partnerships treats a dated baseline tied to a specific result as standard practice for tracking foreign-aid programs. A case study that skips it gives your donor’s own monitoring team more work, not less.
The full findings, including the 25-point scoring guide behind every score, are in our white paper on the 50-case-study review. It’s available by signup.
What are the four case-study formats, and why does it matter?
The 50 case studies split into four formats: 20 stories or features, 16 results explainers, 9 report or research landing pages, and 5 project or program profiles. A report landing page usually introduces a longer document, while a story or results explainer is more likely to account for the evidence and the narrative itself.
The four formats do different publishing jobs, so a short landing page and a long story shouldn’t be scored as if they were the same thing. IIED’s case study on gender-focused economic models of climate resilience in Madagascar has concise landing copy, but most of its evidence sits in the downloadable report behind it. Check which of the four jobs a case study is doing before judging its evidence.
What’s the difference between reporting activity and proving change?
Activity is what a program delivered. Change is what happened because of it, and a case study needs to show both, with a baseline to compare against. A program that trained 2,000 farmers reports activity; a program that shows the share of farmers using a new practice, before and after training, reports change.
The World Bank’s case study on disaster risk and climate change in Colombia gives dated results but no beneficiary voice and no comparison that could test cause and effect, so a fundraiser reusing the numbers would still need to track down the evidence behind them. Practical Action’s case study on off-grid solar energy in Burkina Faso has a similar gap: it reports what was delivered, not what changed for the people who received it.
We call this the difference between a contribution claim and a causal claim. A contribution claim says a program helped produce a result without claiming it was the only cause. A causal claim says the program caused the result, and it should appear only when a study or comparison can rule out other explanations. Only one case study in our sample used a comparison strong enough to support a causal claim: a Mercy Corps evaluation on drought resilience in Ethiopia.
Why do readable case studies still make claims the evidence doesn’t support?
A case study can read well and still overstate its result. In our sample, 30 case studies (60%) earned the top score for search readability, and 31 (62%) earned the top score on our AI Source-Clarity check, which looks only at whether the visible text names the people, place, dates, figures, result, and source. Neither score checks whether the claim itself holds up.
The World Bank’s case study on building a resilient future in the Pacific is clearly organized and full of figures, but several of those figures are portfolio outputs or expected benefits rather than measured results. Good structure doesn’t turn a forecast into proof. The easiest sentence for a reader, or a search system, to quote is usually the least promotional one: it names who acted, where, when, what changed, and how that change was measured.
Why do evidence links and calls to action do different jobs?
An evidence link answers one question: why should I believe this? A call to action answers a different one: what should I do next? A single generic subscribe or donate button answers neither.
Of the 50 case studies, 33 (66%) linked to a report, study, or evaluation, and 32 (64%) gave a useful next step, such as reading the underlying research or viewing a related case. IIED’s case study on climate risk and resilience in Zambia explains its method and then sends readers to the report. That’s a useful next step, but it leaves the actual findings inside the download instead of on the page readers and search engines can see.
What are the eight parts of a case study that hold up under scrutiny?
A case study holds up when it separates delivery from change, limits its claim to the evidence, and gives readers a way to check it. Our review of the 50 case studies points to eight parts that do this work, in the order you can use them.
- Person, place, and problem. Open with a named person, community, or decision-maker in a specific place, and describe the problem the way they experience it.
- Starting point. State the condition before the work began, with a date and a unit. If no number exists, describe the starting condition and name the source.
- The work and the partners. Explain what happened, who did it, and which partners or public bodies took part.
- What was delivered and what changed. Separate the activities and products from the changes in behavior, income, skills, resilience, or the environment.
- A claim the evidence supports. Say the program caused the result only when a study or comparison can test other explanations; otherwise, say it helped or contributed, and name other influences.
- A useful human voice. Use named quotations that explain a decision, a problem, or a change, not quotations that only praise the organization.
- A way to check the claim. Link to the report, study, data, or project record behind the result.
- A limit and a next step. Say what’s uncertain or unfinished, then give the reader something useful to do next.
Why does a consistent brand voice matter for readers and search systems alike?
Brand voice is how you repeatedly name your programs, describe the people you work with, explain your evidence, and state uncertainty. Stable names and consistent claims help a reader, a journalist, or a search system trace one case study back to your reports, your partner pages, and your other media coverage.
CGIAR’s case study on climate services in Guatemala uses the same institution names and dates across multiple supporting reports, so the trail from claim to evidence is easy to follow no matter where a reader enters it. A program that goes by three slightly different names across your website makes that trail harder to find, for a human reader and an AI system alike.
What makes a case study citable in Google AI Overviews and ChatGPT?
A case study becomes citable when the visible text, not a linked report, names the people, place, dates, figures, result, and source. Search systems can only repeat what’s on the page, and 60% of US adults now say they read AI summaries at the top of search results, Pew Research Center found in its 2026 survey on AI adoption.
Five checks sit outside our 25-point scoring guide and cover the wider technical publishing task.
| Check | What it means |
| Identity | Use the same full name for the organization and program every time, and repeat the same author, location, and result names across the case study, the About page, and partner materials. |
| Access | Confirm that robots.txt, hosting settings, and CDN rules let search crawlers reach and index the page. Google’s own guidance says a page must be indexed and eligible for a standard search snippet to appear as a supporting link in AI Overviews or AI Mode, with no special markup required beyond that. |
| Case-study details | Put the main result in visible text, and use Article structured data for the headline, author, and publication dates, matching what’s on the page. |
| Evidence | Link each main result to the report, method, dataset, or outside source that supports it. |
| Measurement | Track discovery, clicks, and referrals in Search Console and analytics, including referral traffic from ChatGPT search. OpenAI’s crawler, OAI-SearchBot, has to reach a page for it to appear there, and blocking the crawler through robots.txt keeps a case study out of ChatGPT’s summaries, though a bare link can still surface from other signals. |
Which case studies got it right?
Two case studies in our sample scored 23 out of 25, the highest in the review: CARE’s story on farmers in Tigray, Ethiopia, and IFAD’s story on irrigation in Masvingo Province, Zimbabwe. Both combined a named human voice with a dated starting point, real figures, a source link, and a stated limit.
CARE’s case study lets a farmer named Kalayu lead the story, uses several of his quotations, compares harvest frequency before and after the work, and states plainly that water remains insufficient. IFAD’s case study combines several farmers’ quotations with before-and-after harvest figures, hectares reached, and the same kind of limit. Other strong entries used different methods: Mercy Corps linked named farmers in Guatemala to a mobile service and change measured over time, WRI drew on research across Bhutan, Ethiopia, and Costa Rica without claiming more than the evidence showed, and CGIAR tied groundwater recharge in India to a dated project and reported results. None of these claimed more than its evidence could support.
How can your team fix weak case studies without a research department?
You don’t need an evaluation department to close the evidence gap. You need one brief, used before the interview even begins, that requires a starting point, a result, a date, a location, a source, a named viewpoint, and a limit.
Practical Action’s case study on climate-resilient farming for Rwandan refugees centers a named farmer, but its broader claims rely only on organization and funder testimony. A brief that asks for the evidence source before drafting would close that gap without adding staff.
Learn more about our case study services for NGOs.
Frequently asked questions
What’s the difference between a case-study format like a story and a results explainer?
A story or results explainer usually carries the evidence and the narrative itself, while a report or research landing page usually introduces a longer document and leaves most of the evidence there. Our sample had 20 stories or features, 16 results explainers, 9 report landing pages, and 5 project profiles, and each format has a different bar for what counts as enough evidence on the page itself.
How long should an NGO case study be?
Length doesn’t decide the score. In our sample, the median case study ran 892 words, the shortest 259, and the longest 4,253, with no clear link between length and how well a case study proved its result. A short case study that names its evidence beats a long one that buries it.
What’s the difference between a contribution claim and a causal claim?
A contribution claim says a program helped produce a result without claiming to be the only cause. A causal claim says the program caused the result, and it needs a study or comparison that can test other explanations. Most case studies in our review made a contribution claim; only one used a comparison strong enough for causal attribution.
Does structured data help a case study get cited by AI search tools?
It helps search systems match the visible text to the right headline, author, and date, but Google’s own guidance says there’s no special markup required beyond standard Article structured data that matches the page. The bigger gap in our sample was missing evidence in the visible text, not missing markup.
Seeking professionals to write your case studies? Please get in touch with us at MacroLingo.
