ImpactMojo DataverseIndian Data Navigator
Most development questions in India can be answered — at least partly — with data that already exists, for free. The hard part is knowing which source, what it actually measures, how fine-grained it is, how often it comes out, and where it will mislead you. Search and filter India's major public datasets, learn each one's strengths and gotchas, then build and export a data-sourcing plan for your project.
India's major public datasets
Tap a card to expand its detail — unit of analysis, granularity, frequency, access and caveats. Use Add to plan to shortlist the sources you'll actually use. Everything here is illustrative and current to the platform's last review; always confirm the latest round and terms on the source's own site before you rely on it.
My data-sourcing plan
The datasets you shortlisted, plus space to note what you'll pull from each and how you'll triangulate them. Export it as a plain-text plan to drop into a concept note or research protocol.
Browsing, filtering and shortlisting are free. Exporting the data plan is a Premium feature (Practitioner plan and up).
Matching a question to a dataset
Before you download anything, run your question through these five checks. They save you from the most common mistake in secondary-data work: using a source that looks right but can't actually answer what you asked.
1. What is your unit and geography?
Do you need a number for an individual, household, village, district, or state? A survey representative at the state level cannot give you a credible district figure, and a district-representative survey cannot give you a village one. Match the dataset's lowest reliable level to your question — not the level it happens to publish.
2. Census, survey, or administrative data?
| Type | Strength | Watch out for |
|---|---|---|
| Census (e.g. Census of India) | Covers everyone — supports the finest geographic detail. | Infrequent (decennial) and slow; India's last full count was 2011. |
| Sample survey (NFHS, PLFS, NSS) | Rich detail, regular, scientifically representative at its design level. | Sampling error grows as you slice finer; not reliable below its design level. |
| Administrative / MIS (HMIS, U-DISE+) | Near-complete coverage, high frequency, cheap. | Reflects what the system records — reporting gaps, incentives to over/under-report. |
3. Level, or change over time?
To measure a trend you need the same thing measured the same way at two points. Beware comparing across sources or across a change in method or questionnaire — a jump can be an artefact of the instrument, not real change. Panel datasets (IHDS, CMIE-CPHS) follow the same units over time and are strongest for change.
4. How recent must it be?
A decennial census tells you structure, not this year's crisis. If you need something current, lean on high-frequency administrative or panel data — and accept its coverage caveats. Always check the reference period: "2019–21" data collected before a shock says nothing about after it.
5. Who collected it, and what do they miss?
You know where to look
The right dataset, at the right level, read with its caveats in mind, is the fastest route to credible evidence — and usually free. Pair this with the RCT Readiness and Impact Evaluation labs to turn data into a defensible claim.