Reading Charts Without Being Misled

What you'll learn

Run the three-second scan: axes, units, and time window before the shape
Spot truncated axes and cherry-picked windows — and know when truncation is actually fair
Recognize cumulative curves and dual-axis charts as the theater they usually are

A chart walks into a meeting with an advantage no number has: you feel it before you read it. The line rockets up and the room is impressed before anyone checks what the axis says. That feeling is the product being sold. Most misleading charts contain no false numbers at all — the data is real; the drawing is doing the persuading. This module is the field guide to the four tricks that account for most of it, plus the three-second scan that beats all four.

The words that matter

Axis: The ruler a chart is drawn against. The y-axis (vertical) sets how big changes look; the x-axis (horizontal) sets which slice of history you’re shown.
Truncated axis: A y-axis that starts above zero, so small differences fill the whole chart and look enormous.
Time window: The start and end dates someone chose. The window picks the story: start at the crash, everything since looks like triumph.
Cumulative chart: A running total. It can only go up — even a collapsing business produces a rising cumulative curve.
Dual-axis chart: Two different y-scales in one picture. By stretching either scale, two unrelated lines can be made to “move together.”
Data-ink: The parts of a chart that carry information, as opposed to decoration. 3D bars, glows, and exploding pie slices are decoration — and often camouflage.

Trick one: the amputated y-axis

Take four quarterly satisfaction scores: 82, 83, 84, 86. Drawn from zero, that’s what it is — a gentle drift. Now start the axis at 80 and the same four numbers become a moon launch, with the last bar three times the height of the first. Nothing was falsified. The chart just amputated the part of the picture where “barely anything changed” was visible.

82, 83, 84, 86 — a flat story from zero, a rocket from 80. The axis chose the feeling for you.

Text description of this diagram

Two bar charts, drawn from the same four numbers: 82, 83, 84, 86. On the left, the axis starts at zero, and the four bars are visibly almost the same height — a calm, honest picture of a small improvement. On the right, the axis starts at 80, so only the top sliver of each value is drawn: now the last bar towers at three times the height of the first, and the identical data reads as a dramatic surge. The amber banner underneath says it plainly: same four numbers, two feelings.

Is a truncated axis always a crime? No — and this is where literacy beats cynicism. For data that never approaches zero (body temperature, stock indexes, exchange rates), zooming in is legitimate and sometimes essential; a fever chart drawn from 0°F would be useless. The rule isn’t “never truncate.” The rule is: notice the axis before you feel the shape, and ask whether the zoom level matches the real-world importance of the change.

Tricks two, three, and four

The convenient window. “Revenue is up 40% since March” — and March was the worst month in company history. Every trend line has a start date somebody chose, and the choice can manufacture growth or hide decline. The defense is asking to see a longer window: real trends survive zooming out; manufactured ones don’t.

The cumulative curve. “Total customers ever acquired” rises even while the business quietly bleeds users — additions get counted forever, departures never subtract. Cumulative charts are the favorite of struggling products because they cannot go down. When you see one, ask for the per-period version: new customers per quarter tells the truth cumulative totals are hired to hide.

Two axes, one insinuation. A dual-axis chart overlays two lines on two different scales — and by stretching either scale, a presenter can make almost any pair of lines dance together: marketing spend and revenue, hiring and outages, anything. Charting professionals treat dual axes as the number-one chart mistake, because the apparent correlation is an artifact of scale-fiddling. And even when two lines genuinely move together, hold that thought — whether moving together means anything is exactly what the next module is about.

Common misunderstanding

“The data is real, so the chart is honest.” A chart makes dozens of choices the data doesn’t make for it — axis start, time window, cumulative vs per-period, scale pairing, even 3D perspective that makes the front pie slice look bigger. Real numbers plus loaded choices equals a misleading chart with a clean conscience. Judge the choices, not just the numbers.

Spot the trick

Try this at work

Install the three-second scan: before letting a chart impress you, read (1) where the y-axis starts, (2) what units and scale it uses, and (3) when the time window begins. Axes, units, window — then the shape. It’s the visual version of “from what, to what, out of how many?”, and after a week it becomes automatic.

The bottom line

Charts persuade before they inform. The four usual suspects — truncated axes, convenient windows, cumulative curves, dual axes — all use real data and loaded drawing. Read the axes before you trust the shape.

Why it matters

Dashboards run the modern company — and every dashboard is a gallery of somebody’s charting choices.

Between QBR decks, board slides, and the BI dashboards you met in Data & Cloud Platforms, you’ll read more charts this year than memos. The people who make good decisions from them aren’t faster readers — they’re just immune to the four tricks, because they scan the frame before the picture. Next module, we tackle the deeper illusion charts love to suggest: that two lines moving together means one causes the other.

Quick check

1. A bar chart of scores 82, 83, 84, 86 looks explosive. The most likely reason?

2. Why do struggling products love cumulative charts?

3. A dual-axis chart shows spend and revenue moving in perfect sync. What's the sharp response?

Answers explained

B is correct — starting the axis at 80 amputates the unchanging part of the picture, so a 4-point drift reads as a tripling. (If you picked A: no falsification needed — the drawing does the work. If you picked C: numbers have no drama until an axis gives them some.)
A is correct — additions accumulate forever and departures never subtract, so the curve rises regardless of health; the per-period version is the honest one. (If you picked B: ease isn’t the motive — the guaranteed upward slope is. If you picked C: no auditor requires cumulative marketing charts.)
C is correct — dual axes let a presenter stretch either scale until the lines dance; a shared scale or indexed view removes the trick while keeping the data. (If you picked A: you’d be approving an artifact of scale-fiddling. If you picked B: the numbers may be perfectly real — it’s the two rulers that manufacture the story.)