How Charts Lie: Eight Manipulation Techniques
A chart is an argument wearing the costume of a fact
Numbers feel objective. A chart feels more objective — it's the numbers, drawn, right there in front of you. That feeling is exactly what makes charts such effective liars.
Because a chart is not the data. A chart is a set of choices about how to show the data — which axis, which range, which comparison, which time window — and every one of those choices can be made honestly or made to steer you. The data can be completely true and the chart completely misleading, at the same time, with nothing technically false in it.
This is the more useful sibling of PS-04. That lesson was how to present data honestly. This one is the same knowledge pointed the other way: the eight moves people use to mislead you with real numbers, so that a chart designed to steer you stops working — because you can see the machinery.
Almost none of these require lying. That's the unsettling part. Every technique below can be done with accurate data, which is why "but the numbers are real" is no defence at all.
The eight techniques
1. The truncated axis — starting the y-axis above zero
The most common by a wide margin. A bar chart's vertical axis should start at zero, because a bar's length is the comparison. Start it at, say, 90 instead, and a change from 91 to 93 — a 2% difference — looks like a bar three times taller than its neighbour.
How to catch it: look at where the y-axis starts. If it's not zero on a bar chart, mentally redraw it from zero and watch the dramatic difference shrink to what it actually is.
(Nuance: line charts showing a trend can legitimately be truncated — you're reading the slope, not comparing lengths. The dishonesty is truncating bars, or truncating without saying so.)
2. The inverted or unlabelled axis
Flipping an axis so up means down, or leaving it off entirely so you can't tell the scale. A famous example inverted a gun-deaths axis so that a rise in deaths looked like a fall. And a chart with no numbers on the axis is asking you to trust the shape without checking the size.
How to catch it: read the axis direction and confirm there are actual numbers on it. No numbers = no chart, just a shape.
3. Cherry-picked time window
Showing exactly the stretch of time that supports the claim and cropping the rest. A stock that crashed and partly recovered can look like a triumph if you start the chart at the bottom. A hot year looks like runaway warming or reassuring stability depending purely on where you begin.
How to catch it: ask "why does the chart start and end exactly there?" Then look for the longer series. The honest question is always "what happened just before the window?"
4. Cumulative totals hiding a decline
Plotting a running total instead of the actual periodic values. A cumulative chart can only ever go up or stay flat — so "total downloads to date" keeps climbing impressively even as new downloads collapse. The rise is real and completely uninformative.
How to catch it: ask whether it's a total-so-far or a per-period number. If adding each period can only make the line go up, it can't show a decline even when one is happening.
5. Misleading area and 3D
Making one thing twice as big as another by doubling its width and height — which quadruples the area, so it looks four times bigger, not two. Bubble charts and pictographs do this constantly. 3D does its own damage: perspective makes near bars look bigger and the tilted base makes values genuinely hard to read off.
How to catch it: be suspicious of any comparison done by area, icon size, or 3D. Ask what the actual numbers are — the visual is usually exaggerating them.
6. The dual axis — engineering a correlation
Two lines, two different y-axes, scaled so they appear to move together. Because you choose both scales independently, you can make almost any two series look correlated — and the crossing point, which viewers read as meaningful, is a pure artefact of the scaling.
How to catch it: two lines with two different axes is an instant red flag. The apparent relationship was chosen, not discovered. (This is the same reason PS-04 says never to make one.)
7. Correlation dressed as causation
Not strictly a chart trick, but charts sell it. Two lines that rise together imply one causes the other, and a well-drawn chart makes the implication feel proven. It isn't — they might both be driven by a third thing, or the link might be coincidence. (DT-03 is the whole lesson on this.)
How to catch it: a chart can show two things moving together. It can never, by itself, show that one causes the other. Ask what else could explain the pattern.
8. Missing context / the denominator trick
A number with no baseline. "10,000 cases!" means nothing without out of how many and compared to when. Raw counts where you needed a rate; a scary total with no population underneath it; a percentage change on a tiny base ("up 300%!" — from one to four). The missing denominator is where a huge share of misleading numbers hide.
How to catch it: always ask "out of what?" and "compared to what?" A number alone is not information — it's information waiting for a denominator.
Why real data can still lie
The through-line worth holding onto: not one of these eight requires a false number. Every chart here can be built from perfectly accurate data. The manipulation lives entirely in the framing — the axis, the window, the comparison, the missing baseline.
This is why "the data is real, I checked" is not a defence, and why fact-checking a chart's numbers misses the point. The numbers are usually fine. The lie is in the choices around them, and those choices are invisible unless you know to look for them. Knowing the eight is what makes the invisible choices visible.
What this means for you
- A chart is an argument, not a fact. Read it like something trying to persuade you, because it is.
- Check the axis first, every time — where does it start, does it have numbers, which direction. This one habit catches the most common lies.
- Ask "out of what?" and "compared to when?" The missing denominator and the cherry-picked window are everywhere.
- Real numbers can produce a false picture. Verifying the data is not the same as verifying the chart.
- The more a chart makes you feel something instantly, the more worth slowing down to check its construction — a chart engineered for impact is a chart that made choices to get it.
Try it: find three lying charts (30 min)
Detection is a muscle; build it on real examples.
- Find three real charts — news articles, social media, an advert, a company's own investor deck. (Company results presentations and political graphics are reliable hunting grounds.)
- For each, work through the eight and identify at least one technique in play. Write:
CHART: (where it's from) ......................................
TECHNIQUE(S): .................................................
WHAT IT WANTS ME TO CONCLUDE: .................................
WHAT AN HONEST VERSION WOULD SHOW: ............................
(redraw it in your head — axis from zero, full time window,
the denominator added — what changes?)- For at least one, sketch or describe the honest version and note how much the story changes.
✅ Finish check: three real charts, each with a named technique, what it was steering you toward, and how the honest version would differ.
Summary card
- A chart is an argument in the costume of a fact. It's a set of choices, and choices can steer.
- Almost none of these require a false number — "the data is real" is no defence.
- 1. Truncated axis — bars not starting at zero. Redraw from zero.
- 2. Inverted / unlabelled axis — check the direction and that there are actual numbers.
- 3. Cherry-picked window — ask why it starts and ends exactly there.
- 4. Cumulative totals — a running total can't show a decline. Ask: total-so-far or per-period?
- 5. Area & 3D — doubling width+height quadruples area. Distrust size and perspective.
- 6. Dual axis — two scales can fake any correlation. Instant red flag.
- 7. Correlation as causation — a chart can show togetherness, never cause.
- 8. Missing denominator — always ask "out of what?" and "compared to when?"
- Check the axis first. It catches the most.
Sources
- Cairo, A. — How Charts Lie, 2019
- Huff, D. — How to Lie with Statistics, 1954
- Wainer, H. — Visual Revelations, 1997
- Tufte, E. — The Visual Display of Quantitative Information, 1983
Next lesson: DT-03 — Correlation Is Not Causation (L1) Related: DT-01 Reading Data · DT-09 Verifying a Source · PS-04 Presenting Data · MN-10 The Anatomy of a Scam Path: Hard to Fool — 2/6