How can a chart mislead without using fake data?

Two bars, one chart. Q1 at 97.2, Q2 at 98.1. In the pitch deck, the second bar towers over the first at roughly four times its height. The data is real, the labels are correct, and the impression is false: the axis starts at 96.

None of the tricks below require fake data. That is exactly why they work: there is no wrong number to point at, only a wrong impression.

Here are the five, one by one, each with its tell.

Trick 1: the truncated axis

Start the axis at 96 and a rounding error becomes a rocket:

Bars encode value by length, so the reader's eye compares lengths, and a truncated axis breaks that encoding at the root. The rule has no exceptions: a bar axis starts at zero. Every time. Lines are allowed to zoom in on a range, but the label must say so.

Trick 2: the dual axis

Revenue on the left scale, signups on the right scale, two lines on one chart. The problem: with two independent scales, the crossing point of the lines is a design choice, not a fact. Stretch one scale and the lines diverge; compress it and they converge. The author picks the story.

If one series needs its own scale, it needs its own chart. Two panels, same width, aligned x-axis. The reader still sees both trends, and nobody chose where the lines cross.

Trick 3: the cherry-picked window

Zoom in far enough and any trend tells the opposite story. A series that climbs from January to December almost always contains a few weeks where it dips; frame the chart on those weeks and growth becomes decline.

Always ask: what happens just outside the frame? Default to the full period, then zoom with a label that says you zoomed.

The defense when you are the one making the chart: show the full period as context, even small, next to any zoomed view. A zoom with context is analysis; a zoom without it is framing.

Trick 4: the cumulative curve

Cumulative totals only ever go up. Even while the business goes down. Plot cumulative sales and you get a smooth curve climbing up and to the right, while the per-month bars underneath fall the whole time.

This is a famous keynote trick. Nothing on the chart is false; the shape of the curve simply cannot show a decline, because a running total has nowhere to go but up. If the metric is a running total, ask for the per-period version.

Trick 5: the missing denominator

The fifth trick does not even need a chart. "Signups doubled" is technically true when one user became two users. Any percentage or growth claim presented without its base is a headline, not a measurement, and it belongs in the same audit as the four visual tricks above.

The 5-second chart audit

Five questions before you believe any chart:

Five questions, five seconds, and every trick on this page gets caught. The point is not paranoia; it is that the audit is cheap enough to run on every chart, including your own before you present it.

The takeaway

A chart does not need fake data to lie. It only needs a helpful axis, a second scale, a convenient window or a running total. Run the five questions before you believe anything in a deck, and pin the print-ready PDF above next to your screen for the next all-hands.