How we study

Each claim carries one of seven labels: primary source, secondary source, official statistic, interpretation, normative claim, unverified, or synthetic. Unverified material can never be published as current (§64, §71).

Every item passes nine stages — proposal, draft, evidence, rights, accessibility, technical review, approval, publication and monitoring — each with a stop condition. Legal, statistical, translation and community reviews are triggered automatically by what the text contains (§173, §184).

The person who writes a version never approves it (§178). Where a small team makes that impossible, the exception is recorded and shown on the public version history.

We do not produce composite scores, league tables, red-amber-green ratings or default orderings by value. Areas may be compared side by side; they are never ranked (Volume 8 §110).

Every publication states what it does not tell you. Aggregate figures for a village or panchayat conceal differences of caste, gender, hamlet and landholding; we say so on every profile (§155).

If software helped draft any part of an item, the item says what was used and how it was checked. Nothing about law, statistics or communities written with machine assistance is published without human review (§70, §146).