Charts That Mislead: Interview-Ready Guide to Truncated Axes, Dual Axes and Bad Scales
A business news panel flashes two bars: last quarter looks tiny, this quarter towers over it, and everyone starts talking about a “surge.” Then you notice the y-axis begins at 95, not zero - the change is real, but the drama is manufactured by the scale.
- A chart misleads when visual distance is not proportional to data difference. The eye sees shape first and reads labels later.
- Truncated axes are most dangerous in bar charts because bar length implies magnitude from zero.
- Dual axes can create fake relationships by scaling two unrelated series until their lines seem to move together.
- Bad scales include inconsistent tick intervals, unlabeled log scales, cherry-picked ranges and compressed timelines.
- Line charts may use non-zero axes if the goal is to show small variation, but the truncation must be visible and justified.
- The quickest audit: check baseline, tick spacing, units, source, time period and whether the visual effect matches the actual percentage change.
- Interview line: “I would not reject the chart immediately; I would test whether the encoding supports the business question without exaggerating the effect.”
The Big Picture
Misleading charts are rarely about one “bad” chart type. They happen when a chain of choices - data selection, scale, encoding, annotation and narrative - squeezes a complex reality into a visual takeaway that overstates, hides or reverses the truth.
The Three Scale Traps You Must Spot Fast
Trap 1 - Truncated axis: the axis starts above zero or ends early, making small differences look large. It is especially risky with bars because the bar itself represents quantity.
Trap 2 - Dual axes: two different y-axes are placed on the same chart. Because each line gets its own scale, the designer can make unrelated variables visually appear correlated.
Trap 3 - Bad scale choice: the chart uses inconsistent tick intervals, an unlabeled log scale, unequal time gaps, cherry-picked dates or a range that hides the denominator.
Worked Example: How a Tiny Change Becomes a Giant Bar
Suppose a company’s revenue rises from ₹100 crore to ₹108 crore.
The true increase is:
(108 - 100) / 100 = 8%
Now imagine a bar chart with the y-axis starting at ₹95 crore. The first bar has visible height 5 units, and the second has visible height 13 units.
The visual increase becomes:
(13 - 5) / 5 = 160%
The chart has turned an 8% business improvement into a 160% visual jump. That is the essence of axis manipulation.
The Dual-Axis Trap: Correlation by Design
Dual axes feel efficient because they fit two metrics into one chart. The problem is that each axis can be stretched independently, so the slope and overlap of the lines are partly design artefacts.
A safer alternative is to index both series to 100 at the starting period, or use small multiples - two aligned charts with the same x-axis but separate panels. This keeps comparison honest without forcing false visual similarity.
Definitions You Can Say in One Breath
Graphical integrity: “The representation of numbers, as physically measured on the surface of the graphic itself, should be directly proportional to the numerical quantities represented.”
- Truncated axis: an axis that does not start at the natural baseline, usually zero for magnitude charts.
- Dual-axis chart: a chart that plots two measures using separate y-axes on the same x-axis.
- Log scale: a scale where equal visual steps represent equal ratios, not equal additions.
- Lie factor: the size of effect shown in the graphic divided by the size of effect in the data.
Chart Integrity Measures: 5 Checks With Formulas
Use these when you need to move from “this looks misleading” to a rigorous audit.
NPCI publishes UPI transaction data that analysts often use to discuss digital payments growth in India. If monthly UPI volumes are shown as bars with a y-axis starting near the lowest month, a normal month-on-month change can look like a dramatic jump. The strategic so what: in high-growth Indian categories, scale discipline matters because the narrative can influence investor, policy and product decisions.
When a Non-Zero Axis Is Actually Acceptable
Do not give a simplistic answer like “all truncated axes are wrong.” The rule depends on the chart type and question.
Reuters Graphics: Making Scale Choices Visible During a Crisis
Reuters Graphics showed how careful scale design can make pandemic data understandable without hiding uncertainty or exaggerating trends.
Situation: During the COVID-19 pandemic, readers needed to compare outbreaks across countries, including India, where raw counts, testing changes and reporting delays made simple charts easy to misread.
The move: Reuters Graphics leaned on visible sourcing, clear units, rolling averages, country-level pages and scale choices suited to the question. For exponential spread, log views helped compare growth rates; for public comprehension, linear views kept magnitude intuitive. The power was not one technique - it was a system of clear labels, consistent time axes, explanatory notes and source transparency.
Outcome or lesson: The lesson for managers is not “use log charts” or “avoid log charts.” It is to match scale to decision. If the decision is hospital capacity, absolute magnitude matters. If the decision is speed of spread, growth rate matters. A responsible chart makes that choice visible.
Takeaway: Trustworthy visualisation is not plain-looking visualisation; it is visualisation where the viewer can see how the chart earned its conclusion.
How AI Changes Charts That Mislead
1. AI-generated dashboards can automate bad defaults. Natural-language BI tools can convert “show growth” into a chart with an aggressive axis range or an unnecessary dual axis. The risk is speed without visual audit.
2. AI can also become a chart reviewer. LLMs can inspect a chart specification, calculate percentage changes, flag truncated axes and suggest small multiples or indexed charts. This is useful before presentations, especially when dashboards are built quickly.
3. AI raises the bar for accessibility. Tools can generate alt text, summarize trends and detect low-contrast colours, but they may miss whether the scale itself changes the meaning. Human judgment still owns the business interpretation.
Paste the chart data and a screenshot description into ChatGPT or Claude and ask: “Audit this chart for truncated axes, dual axes, bad scales, missing labels and lie factor. Suggest a safer chart for an MBA interview answer.” Then verify the math yourself.
Interview Relevance
“You are shown a dashboard where sales and ad spend are plotted on a dual-axis chart and appear to move together. How would you evaluate whether the chart is misleading?”
A strong answer does not sound like a chart police complaint. It sounds like a manager protecting decision quality: “Before acting on this dashboard, I would separate the real movement from the design effect.”
Common Mistake
The costly mistake is saying “truncated axes and dual axes are always wrong.” That sounds rigid and junior because some analytical questions require focused ranges or ratio scales. The one-line fix: judge the chart by whether the scale honestly supports the decision being made.
What to Revise Next
Once you can spot scale manipulation, move to the two skills that make your own charts presentation-ready: readable visual design and knowing when not to chart at all.