Spotting Misleading Charts
Rohan's slide for the board shows two bars: February conversion 2.4%, March 2%. The February bar towers over March's, five times as tall, and the room gasps. Nobody lied; every number on the slide is correct. The drama comes from where the axis starts. As the BI analyst, you are the person who catches this before the meeting, and who never builds it yourself.
The rule
Check the axis, the window and the ink before you believe.
A chart misleads when its design makes a difference look bigger, smaller or different from what the numbers say. The usual culprits:
- A bar axis that does not start at zero. A bar is read by its length, so cutting off its base exaggerates differences. Rohan's axis started at 1.9%, so the bars were 0.5 and 0.1 points tall.
- A cherry-picked window. "Conversion rose three weeks in a row!" can be true after the worst month of the year. The period shown should match the question, not the hoped-for answer.
- Ink that does not match the number. In 3D charts and bubbles, area carries the message. A circle with twice the radius has four times the area, so a value twice as big looks four times as big.
- Two axes on one chart. With a left and a right scale, the designer chooses where the lines cross. Almost any story can be drawn.
- Running totals presented as performance. A cumulative line always rises, even in a week when every day was worse than the last.
Lines are different from bars. A line is read by its position and slope, so zooming the axis to 80–100% is fair as long as the axis is clearly labelled. Zooming is a trick only on bars and areas.
The shortcut
Give every chart three ten-second checks: Where does the axis start? Why this time window? Does the amount of ink match the size of the number?
Where does the axis start?
At 1.9%. The bars are drawn from 1.9, so their lengths are 2.4 − 1.9 = 0.5 and 2.0 − 1.9 = 0.1.
The honest version is less dramatic and more useful. The room can now discuss the real question: why did 0.4 points of conversion disappear?
Drills
Answer fast — trust the rule.
A line chart of weekly on-time rate uses a clearly labelled axis from 80% to 100%. Is it misleading?
Jaipur's bubble has twice the radius of Kolkata's. How many times bigger does it look?
Rohan wants to show conversion rising for the last three weeks, right after a bad month. What is the fair version?
Remember
- Check the axis, the window and the ink before you believe.
- Three ten-second checks: axis start, time window, ink versus number.
- Trap: bars that do not start at zero. Lines may zoom; bars may not.