How a graph learns to lie.

Meet the bones of a chart.

Before we catch a chart lying, let's name its parts — because the lie almost always hides in one of them. Here's a plain bar chart with everything labelled.

• Axis — a number line along the edge of the chart. The y-axis runs up the side (how much), the x-axis runs along the bottom (which thing, or when).
• Scale — how much each step on an axis is worth. A fair scale takes even, equal steps (0, 2, 4, 6…). Bend that and the picture bends with it.
• Baseline — the value where the bars start growing from, sitting at the bottom. For most bar charts it should be zero.
• Truncated (chopped) axis — a y-axis that doesn't start at zero. It snips off the bottom of the chart so only the tippy-tops are left — and the tops are where the differences look biggest.

Hold onto that last one. It's the most common trick in the world, and you're about to drive it yourself.

Chop the axis. Watch a molehill become a mountain.

Below are four lemonade stands and how many cups each sold. The real numbers barely differ — the busiest stand sold only about 10% more than the quietest. But drag the slider to lift the baseline off zero, and watch that tiny gap explode into a "shocking" chart. Then hit Make it honest and watch the shock shrink back to the truth.

This is the whole con in one move. The bars are honest about their heights; the floor is the liar. By starting the axis at 100 instead of 0, you throw away everything the stands have in common and zoom in on the sliver where they differ. A 10% gap can be made to look like a 6× landslide — same data, opposite story.

For bars, the baseline belongs at the bottom.

A bar chart works because the length of each bar stands for the number. Twice the length should mean twice the value. That only stays true if every bar is measured from the same honest floor: zero. Chop the floor off and the lengths stop matching the numbers — a bar that's "twice as tall" might be carrying a value that's only a hair bigger.

Bars are about length. If you don't start at zero, the length is a lie.

Picture two phones side by side. One lasts 19 hours on a charge, the other 20 hours — basically a tie. Start the axis at zero and the two bars look like twins, which is the honest story. But start it at 18 and the 20-hour bar is suddenly twice the height of the 19-hour one. The advert can now scream "lasts twice as long!" with a straight face, because it never said the words — it just drew them. Same one-hour difference, two completely different feelings, decided entirely by where the floor sits.

So is a chopped axis always cheating? Not quite — and that's the honest part of the story. A truncated y-axis is fine when:

• The chart is a line graph tracking small wiggles over time — like daily temperature — where the whole point is to see the changes, not the total height. (Here, length doesn't stand for the value, so the rule relaxes.)
• Tiny differences genuinely matter, and the chart shouts about it — a clear note, a visible break mark (a little zig-zag) on the axis, and labels you can't miss.

The trick isn't chopping the axis. The trick is chopping it quietly, on bars, and hoping you read the picture instead of the numbers.

Uneven scales: a ruler with cheating marks.

A fair scale takes equal steps for equal jumps: the gap from 0→10 looks the same as 10→20 and 90→100. Break that promise and the chart can stretch the parts it likes and squash the parts it doesn't — without changing a single number.

Look at these two y-axes for the very same readings. On the left, the numbers are spaced evenly. On the right, someone crammed the small numbers together and spread the big ones out, so a steady climb suddenly looks like an explosion at the end.

The same bending can happen sideways. A fair x-axis that shows time should space the dates evenly — one year, one step. But squeeze a decade of slow change into a thumbnail and spread the last few months across half the chart, and a gentle slope rears up into a wall. If you ever see a time chart whose dates aren't evenly spaced — 2010, 2015, 2018, 2019, 2019.5 — someone is choosing where to stretch and where to squash.

There's also a fair kind of stretched scale called a logarithmic scale, where each step multiplies (1, 10, 100, 1000…) instead of adds. Scientists use it on purpose for things like earthquakes or sound, and it's honest — as long as it's labelled. The sneaky version is the unlabelled one that hopes you'll assume the steps are even when they aren't. Same rule as always: read the numbers on the axis, don't just feel the shape.

3D charts: where slices grow a belly.

Tilt a chart into fake 3D and something sneaky happens: the slices and bars near the front get a thick, shaded side that adds bulk your eye counts as "more." Nothing in the data got bigger — the chart just leaned toward you and put its closest piece in the spotlight.

Here are two pie slices that are exactly equal — each is half the pie. But tilt the pie and add depth, and the front slice looks like it ate the back one for lunch.

Bars catch the same disease. Stand a row of bars up in 3D and the ones in front get a chunky shaded box that the back ones don't, so the eye scores them as taller. Worse, the perspective can make it genuinely hard to tell where a bar's top lines up with the scale — you're guessing at the value instead of reading it. The chart stopped being a measuring tool and became a stage set.

The cure is dull and powerful: flat charts read true. A plain 2D pie or bar has no front and no back to play favourites, so each slice is just its honest share. When a chart shows up wearing a 3D costume, ask what the tilt is distracting you from.

Pictograms: when "twice as much" looks like four times.

A pictogram shows numbers with little pictures — coins, people, lemons. Done right, two coins mean twice one coin. The trap is making the picture bigger instead of using more of them. Here Group B is genuinely twice Group A. Drag the slider from "taller only" to "scaled up whole" and watch what your eyes actually see.

Here's the maths behind the gut-punch: if you scale a picture's height and width by 2, its area doesn't double — it goes up by 2 × 2 = 4. Triple both and the area is 9 times bigger. So a designer can show a number that doubled while the picture balloons to look four times as mighty, all without writing a single false number. Fair pictograms keep every icon the same size and just use more of them.

Spot the trick.

Four little charts, four sneaky moves. Look hard at each one and guess what's off, then tap the card to reveal the trick. (Each is a deliberately broken example — no real data.)

Notice the pattern? Every fix was the same: stop reading the shape and start reading the numbers and labels.

You're the fact-checker now.

A poster claims: "Our smoothies have DOUBLED in popularity!" next to a giant cup icon that's twice as tall and twice as wide as last year's cup. What's the honest read?

Six questions for any chart.

You don't need to memorise every trick. You need a habit: before you believe a graph, run it past these six. Most lies fail at least one.

1. Where does the y-axis start? If it's a bar chart and the baseline isn't zero, be suspicious of how big the differences look.
2. Are the scale's steps even? Check that equal gaps mean equal jumps. Watch for a log scale (and whether it's labelled).
3. Is everything labelled? Axes, units, a title, a source. No labels means no claim you can check.
4. Why this range? Ask what's outside the chart — before the start date, after the end date. Cherry-picking lives at the edges.
5. Is it dressed up in 3D or giant icons? Tilts and resized pictures distort area. Imagine it flat and plain.
6. Does the picture match the numbers? Read the actual values. If the feeling and the figures disagree, trust the figures.

Why this matters past the maths test.

Misleading charts aren't a textbook curiosity — they're a daily sales pitch. Once you've got the eye for it, you'll start spotting these everywhere:

• Adverts. "4 out of 5 people prefer ours!" beside a bar that towers over the rival — on a chopped axis hiding that the real gap is tiny.
• Headlines. A line that "plunges" or "soars" often turns out to be a normal wiggle on a zoomed-in, baseline-chopped axis.
• Social media. A screenshot of a chart with no source, no labels, and a suspiciously perfect trend — shared because the shape is exciting, not because anyone checked it.
• Even good-faith reports. Plenty of misleading charts aren't evil — they're rushed. A default 3D setting or an auto-chopped axis can fool the person who made it, too.