Identify the unit of comparison
Read the title, labels and notes before interpreting the shape. Determine whether the chart shows counts, rates, percentages or an index. A higher count can reflect a larger underlying population rather than a higher rate. Ask who or what belongs in the denominator. If that information is unavailable, state the uncertainty. A compelling visual impression cannot compensate for not knowing the quantity being compared.
Inspect the scale
Check where the axis begins, whether intervals are equal and whether the scale is linear or logarithmic. A truncated axis can exaggerate the visual size of a difference, particularly in a bar chart where length carries meaning. It is not automatically illegitimate in every display. Judge whether the choice is clearly labelled and appropriate to the purpose, then describe the numerical difference separately from its visual drama.
Check the time window
A trend can look different when the start and end dates change. Ask whether the selected period is justified by the question or merely convenient for the narrative. Look for missing intervals and changes in reporting practice. Do not assume that connecting two observations establishes a continuous path between them. Where possible, compare the chart with its underlying table or official data description before drawing a strong temporal conclusion.
Compare like with like
Confirm that categories use compatible definitions and coverage. Two series can share a label while counting different events or populations. If definitions change, a break may be needed rather than an uninterrupted comparison. Read uncertainty markings and footnotes, not just the largest number. A chart can be mathematically accurate yet invite an unfair comparison when the reader is not told how the categories differ.
Write the claim in words
State the measured quantity, comparison, period and limitation in one short paragraph. Then ask whether the chart supports that sentence or only a weaker one. For practice, redraw a small hypothetical dataset using two clearly labelled scales and explain how perception changes while values remain the same. The lesson is to inspect the design, not to assume every striking chart is deliberately deceptive.
