Read averages, risk changes, samples and charts critically before accepting a numerical claim.
For incomes of 20, 22, 24, 26 and 200, the mean is 58.4 and the median is 24. The mean includes the influence of the highest value; the median describes the middle position. Neither is automatically wrong.
If an event rate moves from 2% to 1%, the absolute reduction is 1 percentage point and the relative reduction is 50%. Reporting only 50% hides the starting frequency. The same distinction applies to conversion rates and discount comparisons.
A result among 100 volunteers is not automatically representative of an entire country. Ask who was eligible, who responded, when data were collected and what was excluded. A percentage without its denominator can conceal a very small count.
Check whether the vertical axis starts at zero, whether time intervals are equal, and whether values are totals or per-person rates. Correlation alone does not establish causation. An observational comparison can reflect confounding variables.
English edition: 16 September 2026.