Seven Questions to Ask of Any Number Before It Reaches Print

Introduction

A number in print looks authoritative. It may have travelled through a ministry bulletin, a news agency, an aggregator, a blog and a social media post before it reached you. At each stage it can lose its unit, its date or its qualifier. This checklist is meant for writers, editors and anyone who quotes figures in a report, a presentation or a post.

Questions 1 to 4: Source, Arithmetic, Measurement, and Period

1. Where did it come from? Trace the number to its first publisher. If you are reading a news report of a ministry bulletin, look for the bulletin. If you are reading a summary of a consultancy report, look for the consultancy's own page. In our Education package, a secondary summary of the national higher-education survey gave a gross enrolment ratio of both 7% and 30.0%. The official report was the reference, and we told readers so.

2. Does the arithmetic work? Do the parts sum to the whole? Does the percentage match the two numbers it describes? PhysicsWallah's segment revenues (Rs 105 crore and Rs 490 crore) did not add up to the consolidated Rs 1,054 crore. The Ministry of Tourism summary said foreign arrivals fell 8.1% to 91.49 lakh from 99.52 lakh: (99.52 − 91.49) / 99.52 = 8.07%, which checks out. Spending two minutes with a calculator is the cheapest check there is.

3. What exactly is being measured? Name the measure, with its qualifiers: net or gross, nominal or real, calendar year or financial year, consolidated or standalone, people or visits. The Ministry of Tourism data show foreign tourist arrivals falling 8.1% while foreign tourist visits to states rose 21.6%. Both are correct because they count different things.

4. What period does it cover? A year-on-year number and a quarter-on-quarter number differ. A figure from February 2026 is not a figure for the year. When our hospitality piece described the Indian hotel sector as weaker than the previous month and stronger than the previous year, both claims were true.

“Most statistical errors are copies of copies.”

Questions 5 to 7: Second Sources, Headlines, and Inferences

5. Is there a second source? Look for an independent measure. Cushman & Wakefield and CBRE both reported record nine-month office leasing, though their totals differed (64.6 and 66.4 million square feet). Agreement in direction with disagreement in level tells you the trend is robust and the exact figure is not.

6. What is the headline doing? Headlines compress. Check that the body supports the headline. One article headlined hotel investment as a "record" while its body did not say so. We did not repeat the word.

7. What did you infer? If you filled a gap, label it. Where we guessed that 30.0% was the intended ratio, we said it was our inference and why. Inference is legitimate; unlabelled inference is not.

A Worked Example (Invented Numbers, Ours)

An invented report says: "Exports of widgets from Eastland rose 25% to $5 billion." Apply the checklist: 1. Source: a news summary. Find the customs authority's release. 2. Arithmetic: $5 billion after a 25% rise implies $4 billion before. Does last year's published figure show $4 billion? 3. Measure: widgets only, or all manufactured goods? Goods only, or goods and services? 4. Period: calendar year, financial year, or the latest quarter annualised? 5. Second source: does the trading partners' import data show something similar? 6. Headline: does it say "record" or "surge" without a base? 7. Inference: have we assumed the unit is dollars, not local currency?

The country, product and figures are invented. In practice, one of the seven questions usually finds a problem.

Special Cases & Handling Unverifiable Data

Special cases: • Forecasts. Attribute them to the forecaster and the date. A forecast is not a fact. • Company claims. Label them as claims. "Best ever quarter" is a claim. • Quotes. Use the exact words, short, with the speaker and the date. • Percentages of small bases. A 120% rise from a small base may be a few units. Give the base. • Round numbers. Be suspicious of exact round figures with no source.

When you cannot verify: Say so. A sentence such as "we could not independently confirm this figure" is more useful than silence. If the number matters to the argument, find another route or leave it out.

A one-line version for the desk: Source, sum, measure, period, second source, headline, inference.