Evidence you can actually trace

Why this exists

Development practice is full of confident claims and thin evidence. Programmes get copied across borders on the strength of a compelling story rather than a careful reading of what actually worked, for whom, and under what conditions. This library exists to make the evidence legible: to sit next to the theory and show what interventions look like in the field — including the failures, the trade-offs, and the cases where the honest answer is "we don't really know yet."

It is built for students, researchers, and practitioners who need cited evidence and real-world lessons rather than marketing.

How studies are selected

Cases are chosen to span regions, decades, and sectors — and, deliberately, to include interventions that failed or produced mixed results alongside the celebrated successes. A resource-curse case sits beside a sovereign-wealth-fund success; a cash-transfer triumph beside a subsidy reform that sparked riots. The aim is a balanced evidence base, not a highlight reel.

• Real interventions. Every case is an actual programme, policy, or reform — not a hypothetical or a model.
• Documented outcomes. Each case reports outcomes with at least one verifiable source, and usually several.
• Honest about uncertainty. Where the causal evidence is weak, the study says so rather than overclaiming.

The evidence tiers

Not all evidence is equal. Each study is automatically tagged with an evidence tier based on the number of citable references and whether the underlying research uses credible causal-identification methods (randomised trials, quasi-experiments, or systematic reviews).

The tier is a signal, not a verdict. A "strong" tag means the evidence is well-identified — it does not mean the programme was good. Read the study.

How to use the library

• Search & filter by topic, region, or decade — or ask a plain-language research question and let the matcher surface relevant cases.
• Save studies to a personal reading list that lives in your browser (no account needed).
• Take Cornell notes directly on each study page and export them as Markdown for your own work.
• Follow the citations. Every reference is copy-ready, and related studies are surfaced as a small literature map.
• Explore the Insights dashboard on the home page to see how the collection is distributed across topics, regions, and decades.

How it's built

The library is deliberately simple and durable: static HTML, CSS, and vanilla JavaScript with no runtime framework, so it loads fast and will keep working for years. Behind that sits a small, zero-dependency Node build pipeline that treats each case study as a single source-of-truth JSON file and generates everything else — the card index, a search index, the statistics dashboard, and the sitemap — automatically. A continuous-integration check validates every study against the schema and fails the build if the generated data drifts out of sync.

The full data model and build commands are documented in the data README.

Contributing

Corrections, better citations, and new case studies are all welcome. Each new study needs a programme name, country and period, a 200–400 word description, evidence of outcomes, and at least two verifiable sources. Open an issue or a pull request on GitHub.