---
title: "Intervention Database"
group: "Flagship references"
order: 3
summary: "A searchable intervention reference with evidence and depth filters, cost context, side-by-side recorded fields and source-linked research dossiers."
status: live
statusNote: "15 sourced intervention records with six filters, side-by-side field comparison and research dossiers. Costs retain their original outcomes and caveats; review remains pending."
conditions: ["measurement-gap", "single-lens"]
firstMilestone: "Every intervention with a Cochrane or Campbell review is in the database."
help:
  - "Add an intervention with sources: copy any file in src/content/interventions/."
  - "Update an evidence grade when a new systematic review appears."
  - "Add a cost-per-outcome figure only from a published, credible estimate."
related: ["/tools/impact-calculator/", "/tools/compare-tool/", "/programs/evidence-reviews/"]
---

An *intervention* is a specific thing done to make life better: a bed net, a cash transfer, a law against lead paint. The Database records what each one does, where it sits on the [depth scale](/method/#depth), how strong the evidence is and how long its effects last.

Cost per outcome appears only where a credible source publishes one, with its date and caveats. The starter database does not yet contain a comparable cost estimate for source-level work. How to build one is part of [Open Problem 3](/open-problems/measuring-the-source/). The [Impact Calculator](/tools/impact-calculator/) and the [Compare Tool](/tools/compare-tool/) read straight from this database.

## Use this reference base

Search mechanisms and sources, then filter by cause, depth, evidence grade, time to effect, cost-data availability or editorial review. Select two to four entries to compare their recorded fields. Download a dossier to retain the full text, source list and estimate caveats.

## Coverage and review

The comparison never turns the records into a ranking. Cost estimates may refer to different outcomes, settings, dates and models. Unknown is not zero, and an evidence grade is separate from human editorial review.