A chart can be perfectly accurate and still wildly misleading, and the usual culprit is not the data but the axis: a truncated y-axis routinely turns a two-percent change into a visual cliff. Statistical authorities have catalogued these techniques for decades — Darrell Huff's 'How to Lie with Statistics' documented the truncated axis trick as far back as 1954. Confirmed, dated, still the number-one move in viral charts today.
The discipline is simple: audit the axes before you react to the shape. The shape is an argument; the axes are the evidence.
What does a truncated y-axis do?
It deletes context and magnifies drama. A line chart of values between 94 and 98 looks like a mountain if the axis starts at 90, and like flatlands if it starts at zero. Neither view is false; only one is honest about scale. The check: find where the y-axis begins. If it does not start at zero for a bar chart, the bar heights are no longer proportional to the values — that is not a judgment call, that is how bar charts work.
Why do dual axes mislead?
Two series on two different scales can be stretched until they align or diverge however the chartmaker likes. Correlation becomes decoration: move the right-hand axis and the lines cross where you want them to. When a chart shows two y-axes, treat the apparent relationship between the two lines as an editorial choice, not a finding. The honest question is whether the chartmaker publishes the scaling — unconfirmed scaling means an unconfirmed relationship.
What is the trick with time ranges?
Cherry-picked windows. A trend that is flat over twenty years can look like a rocket over a window of fourteen months. The tell is the x-axis: where does it start, and does that date have a reason? Charts that begin at an unusual baseline — just after a crash, just before a peak — are chosen, not found. Ask what the same series looks like with five more years on each side. If the chartmaker did not ask that, the chart is advocacy.
How do 3-D and pictogram charts distort?
Volume lies. A 3-D bar drawn twice as tall is roughly eight times the visual volume; perspective tilts make rear bars look smaller. Pictograms that grow in both dimensions break proportionality the same way — a money bag drawn twice as big reads as four times as much. Huff covered both in 1954, and they still circulate because they still fool. Flat, proportional, labeled: the boring chart is the trustworthy one.
What should you check in ten seconds?
Four things, in order. Does the y-axis start at zero, and if not, is the truncation disclosed? Are both axes labeled with units and dates? Is there a source named — a named institution, dataset or agency? And does the headline describe the data or inflame it? Reuters' fact-check desk debunks viral charts regularly, and the failure is almost always in that list, rarely in the arithmetic. Numbers are cheap to verify; framing is where the lie lives.
What does a missing source tell you?
Everything you need. A chart with a named dataset can be checked, argued with and reproduced; a chart with no source line cannot be confirmed or corrected — which is sometimes the point. Fact-checking desks consistently find that viral charts failing review lack sourcing far more often than they contain arithmetic errors. The register writes itself: no source, no status above unconfirmed. Watch near-sourcing too — 'data from a study' without the study's name is the same absence wearing a lab coat.
How do colors and maps mislead?
Choropleth maps and heat colors carry hidden choices: where the color breaks sit, and whether the scale is linear. Moving a breakpoint can flip a region from 'average' to 'alarming' without changing a single number. The check is the legend and its intervals. If the intervals are unequal or undisclosed, the color is an opinion. Same rule as the axes: the picture argues, the labels testify.
When is a truncated axis legitimate?
Sometimes it is. Stock charts, temperature series and other bounded measures routinely use non-zero baselines because the meaningful variation is small relative to the offset. The standard is disclosure and consistency, not zero at all costs. What is never legitimate is a truncated axis plus a breathless caption about an unprecedented surge. The axis sets the magnification; the caption sets the alarm. If they disagree, believe the axis.
For more context, read How to trace a viral claim to its original source.
For more context, read crowd size estimate.
For more context, read Famous quotes nobody ever said.