How we audit government data

The Data Audit desk reads the data that India’s ministries, departments and states publish, and reports what it shows. The aim is an eye-opener, never an attack: what the official numbers say, where they do not add up, how they could be better, and where they are done well. Our findings are about data, never about a person, and the same tests run on every ministry and every state, whoever governs it.

Five tests on every dataset

1. Adds up. Totals equal the sum of their parts; state figures sum to the national one; no percentage above 100, no negative counts, no duplicated rows, no lakh written for crore.
2. Two numbers. The same quantity in two official sources, or twice in one: a scheme’s claim against a survey, a report against a certification, a budget estimate against actual spending.
3. Plausible. Claims of exactly 100 per cent, growth no real process could produce, figures piled on round numbers, a district far outside its neighbours, a series that jumps when a method changes.
4. Money. Allocated against released against spent, unspent balances, and spending crowded into the last weeks of the financial year.
5. Available. Data that is late, missing, broken, withdrawn without notice, or published only as a scanned PDF.

Three labels

Every finding carries one label, shown at the top of the audit.

• Data error: the data contradicts itself (a total that does not match its parts, an impossible value). Something in the published file is wrong.
• Needs explanation: the data is internally consistent but raises a question it cannot answer by itself, such as two official figures that disagree. We show the gap; we do not claim to know which figure is right.
• Confirmed by a second source: an independent official source shows the same thing.

A mismatch between two official figures is reported as a mismatch. We never call it fraud unless an official finding does.

How a figure gets onto the page

• We record where and when each dataset was read: the address, the date and the time. Official dashboards change daily, so each audit is a dated snapshot.
• Every figure is calculated twice, by separate code. If the two results differ, the finding is dropped until we know why.
• Figures on the page are computed from the stored data, not typed in by hand.
• The data behind each audit is published as a download, with its source.

Fairness

• No party names, symbols or colours. Ministries and states are compared on the same measures.
• Each audit says what the data does not show, as plainly as what it does.
• Where data is timely, clean and complete, we say so: credit is part of every audit.

Corrections

If any figure in an audit is wrong, send the correcting document to editor@oquilia.com. We correct it and record the change on the audit’s page. If a ministry revises its data after an audit, we note that as well.

Changes to this method

25 October 2026: first published.