The most dangerous lie in data visualization is the one that contains no false number at all. The figures are real, the labels correct, the source cited — and still the chart pushes the reader toward a conclusion the data does not support. Let us walk through three classic tricks on one and the same series: four quarters with almost unchanged values — 98, 99, 101, 103.
Trick one: the truncated axis
Here is the honest chart. The axis starts at zero, and you see it at once: there is growth, but a tiny one, about five percent over the year.
The same four values (98, 99, 101, 103). With the axis at zero, the difference is what it really is.
Now the same series, but the axis starts not at zero but at 96. Not a single number changed — only the lower bound of the scale did.
Exactly the same numbers. The truncated axis turns 5% of growth into a several-fold visual jump.
The difference is startling. A modest 5% rise has turned into a dramatic surge, with the last bar looking several times taller than the first. A reader who sees the second chart without the first walks away with a false impression — and cannot accuse the author of lying, since every number is correct.
Trick two: the dual scale
Put two lines on one panel with independent axes and you can make them “cross” or “diverge” at any point you like — simply by tuning each axis. Two utterly unrelated quantities will look correlated. The trick is so flexible that it is almost always worth asking: what happens if you nudge the right axis by a couple of ticks?
Trick three: choosing the interval
A trend can be “created” or “erased” by picking a convenient start and end. Show a series from a local minimum to a local maximum and you have a “rapid rise”; show it from the peak and the same series “falls”. There is one defence against this: a long series. The more context on the chart, the harder it is to hide a convenient segment inside it.
If the conclusion changes when you shift the axis by two ticks, it is not a conclusion — it is an artefact of styling.
The reader’s checklist
Three questions for any chart. Where does the Y axis begin — and if these are bars, does it begin at zero? How many axes are on the panel — and are their scales tuned to make the lines “coincide”? What interval is shown — and is the series clipped to start from a convenient point? If all three answers are calm, the chart can be trusted. These are exactly the rules we hold to across every Meridian piece; more in the Methodology section.
