How to Lie With Charts

Charts are the closest thing modern life has to a Jedi mind trick. A single bar going up and to the right can make
a budget look heroic, a product look miraculous, or a problem look “basically solved” (spoiler: it’s usually not).
And the wild part is that the chart can be made from real numbers and still leave you with a completely wrong idea.

This article isn’t here to teach you to scam people. It’s here to teach you how chart-based nonsense happensoften
accidentally, sometimes “strategically”so you can spot it, correct it, and avoid being fooled by the loudest
rectangle on the screen.

What “lying” looks like in data visualization

A chart “lies” when it creates a strong impression that the underlying data doesn’t actually support. That can
happen through design choices (scale, color, shape), selective framing (time window, categories), or missing context
(sample size, uncertainty, definitions). Sometimes it’s deliberate. Often it’s just someone racing to make a slide
deck before lunch and accidentally committing a crime against axes.

The goal of honest charts is simple: make the viewer’s “fast” interpretation match what a careful reader would
conclude after checking the numbers. If the chart’s first impression is dramatically different from the truth, the
chart is doing propaganda, not communication.

12 classic ways charts mislead (and how to spot each one)

1) Truncating the y-axis to exaggerate differences

A bar chart invites your brain to compare lengths. If the y-axis starts at 0, those lengths are proportional to the
values. If the y-axis starts at, say, 90 instead of 0, tiny differences can look enormous.

Example: Imagine Product A has a satisfaction score of 92 and Product B has 96. If the axis runs
from 0–100, you’ll see “pretty close.” If the axis runs from 90–100, B looks like it’s running laps around A.

Spot it: Check whether a bar chart’s baseline is zero (or clearly labeled as a deliberate,
explained exception). If the baseline is missing or “zoomed,” treat the visual drama as suspicious until proven
innocent.

2) Omitting axis labels (or units) so you can’t sanity-check

A chart without labels is like a recipe that says “add some amount of stuff.” It might taste fine, but you can’t
verify anything.

Missing units are especially sneaky: dollars vs. thousands of dollars, per month vs. per year, percentages vs.
percentage points. One of these can be “a modest change,” and the other can be “call your accountant.”

Spot it: If you can’t answer “what exactly is being measured and in what units?” from the chart
itself, the chart is incomplete at best and manipulative at worst.

3) Dual axes (double y-axes) to imply a relationship

Dual-axis charts plot two lines with two different scales on the same graph. You can make almost anything “move
together” if you’re allowed to stretch one axis like taffy.

Example: One line is “marketing spend” and the other is “sales.” Scale them just right andwow!
Your ads are clearly causing revenue. Or maybe both are rising because it’s the holiday season. The chart won’t
tell you. It’ll just wink aggressively.

Spot it: If you see two y-axes, ask: “Would this look correlated if the scales were fixed in a
neutral way?” Better yet, separate the charts or use a scatter plot if the relationship matters.

4) Playing games with aspect ratio to change the “steepness”

Line charts feel like speedometers for change. Make the chart tall and narrow, and trends look steep and urgent.
Make it wide and short, and the same data looks calm and manageable.

Example: A 10% increase over a year can look like a rocket launch in one layout and like a gentle
hill in another.

Spot it: Ask whether the visual steepness matches the actual change. Look for y-axis range and
compare the change to the full scale.

5) Cherry-picking the time window

If you get to choose where the timeline begins and ends, you can turn randomness into destiny.

Example: Show “crime is rising!” by starting the chart at a multi-year low. Or show “inflation is
falling!” by starting at a peak. Both can be technically true while still misleading about the overall trend.

Spot it: Ask “What happens if we zoom out?” A credible chart often shows context: multiple years,
a baseline period, or a comparison to typical variation.

6) Using cumulative totals to make everything look like it only goes up

Cumulative charts are great for “how much in total so far.” They’re terrible for “is the situation improving?”
because cumulative totals almost never decrease.

Example: A cumulative signups chart can keep climbing even if weekly signups are shrinking. You’re
not lying; you’re just letting the shape do the PR work.

Spot it: If the question is about rate (per week, per month, per capita), a cumulative
line can hide what you need to know. Look for rate charts or add a second chart that shows changes per period.

7) Choosing a chart type that overstates precision or hides variability

A smooth line can suggest a stable trend even when data is noisy. A bar chart can hide distribution details.
A single average can erase important differences between groups.

Example: “Average income rose” can be true while many people saw no gainif the increase happened
mostly at the top. The chart isn’t wrong, but it’s incomplete.

Spot it: Ask whether the chart should show spread (box plots, distributions, percentiles) rather
than only a mean.

8) Histogram/binning tricks to reshape the story

Histograms depend on bin width and bin boundaries. Change the bins and the distribution can look “mostly normal,”
“wildly skewed,” or “two separate populations.” Sometimes all three impressions can come from the same dataset.

Spot it: If a histogram is making a dramatic claim (like “there are two groups!”), check whether
bin choices are explained. When in doubt, try multiple bin widths or show a density plot alongside.

9) Percentages without denominators (a.k.a. “Percent of what?”)

“X increased by 50%!” can mean “from 2 to 3” or “from 2 million to 3 million.” Both are 50%. Your life choices
should not hinge on which one the presenter forgot to mention.

Spot it: Look for the baseline value and sample size. If they’re missing, the percentage is a
headline, not evidence.

10) Log scales without explanation

Log scales can be the right choice for values that grow exponentially or span huge ranges. But they can also confuse
people who read “equal spacing” as “equal change.”

Spot it: If the y-axis labels jump by factors (1, 10, 100, 1000) or otherwise look non-linear,
you’re likely on a log scale. A responsible chart labels this clearly and explains why it’s used.

11) 3D effects and “chartjunk” that distort perception

3D pie charts are the final boss of misleading graphics. The front slices look bigger. Perspective changes area.
Your eyes get tricked while your brain nods along like, “Yes, geometry, surely.”

Even without 3D, unnecessary decoration (heavy gradients, dramatic shadows, icons, fake textures) can distract from
the data or imply importance where there is none.

Spot it: If you notice the styling before you notice the values, the design is doing more work
than the data.

12) Color scales that smuggle in emotionor hide differences

Color is powerful. Red often means danger. Green means safe. Darker colors feel “more.” A poorly chosen palette can
create alarm or calm without changing a single number.

Example: A map shaded from “light pink” to “deep red” can feel terrifying even if the numeric
differences are smallor if the underlying metric should be shown per capita rather than total count.

Spot it: Check the legend. Is the scale linear? Are categories evenly spaced? Is “white” actually
zero, or just “low-ish”? If the palette implies a moral judgment, the chart needs extra scrutiny.

A 60-second chart audit (use this before you believe anything)

1. What’s the claim? Say it out loud in one sentence.
2. What’s measured? Units, definitions, and timeframe.
3. What’s the baseline? Especially for bar charts and “% change.”
4. Is the scale honest? Zero baseline where appropriate, no hidden log scale, no mysterious breaks.
5. Is context missing? Sample size, uncertainty, comparison group, or “what happened before?”
6. Is the chart type right? Bars for comparing categories, lines for change over time, distributions for spread.
7. Could design be influencing emotion? Color choices, 3D, heavy decoration, selective labeling.

How to be persuasive without being misleading

Sometimes you really do need to zoom in on small differences. Sometimes log scales are the most truthful way to show
growth across orders of magnitude. The difference between “clarity” and “con artistry” is transparency.

When it’s okay to zoom the axis (and how to do it responsibly)

• Explain why you’re zooming (“values are tightly clustered; we’re showing detail”).
• Label the axis clearly, including the starting value.
• Consider adding an inset or small companion chart with the full 0-based scale for context.
• Avoid using zoomed axes on bar charts unless you’re very explicitbar length comparisons are especially sensitive.

Replace dual axes with better options

• Use two aligned charts with the same x-axis (small multiples).
• Normalize both series (index to 100 at a starting point) if the question is “move together” rather than absolute values.
• Use a scatter plot if you’re truly examining correlation.

Show uncertainty like an adult

If your data comes from samples, models, or noisy measurements, uncertainty isn’t a footnoteit’s the plot twist.
Confidence intervals, margins of error, and ranges can prevent readers from over-interpreting tiny differences as
meaningful.

How to fix a suspicious chart (without starting a feud)

If you’re the person who has to say “this chart is misleading,” you’re basically the designated driver at the
world’s least fun party. Here are ways to correct it without getting banned from Slack:

• Ask a neutral question: “Can we show the full scale so people don’t overread the change?”
• Offer an alternative: “What if we add a second panel with rates per capita?”
• Keep the goal the same: “We still want the takeaway, just in a way that’s defensible.”
• Turn it into a checklist: Make the team’s default “chart QA” process catch issues early.

Real-world experiences people have with misleading charts (and what they learn)

The funny thing about misleading charts is that most people don’t meet them in a textbook. They meet them in places
where emotions, money, and reputation are involvedpresentations, headlines, dashboards, and social media posts that
show up five minutes before you have to make a decision. Below are common, very human “chart moments” that happen
in the wild, and the lessons people walk away with.

The Pitch Deck Rollercoaster: A small startup team is preparing for investors. Someone drops in a
bar chart showing revenue growth month over month. It looks amazinglike a ski slope, but in a good way. Then a
teammate asks why the y-axis starts at 980 instead of 0. Suddenly the “explosive growth” is a change from 990 to
1,010. Not meaningless, but not a rocket launch either. The team learns a hard truth: if you need a weird axis to
make the story sound impressive, the story might need more time to cook. They also learn the opposite truth:
investors don’t hate honest chartsthey hate surprises later.

The Dashboard Panic Attack: A manager opens a KPI dashboard and sees a red line trending sharply
downward. Everyone panics. Meetings are scheduled. Fingers are pointed. Then someone notices the chart is showing a
two-day window, and the “drop” is just normal daily variation. When they expand to 90 days, the metric is flat.
Lesson learned: always check the time window before you declare a crisis. A dashboard should make trends clearer,
not turn routine noise into a weekly horror movie.

The Viral Social Post: A chart goes viral claiming a dramatic shift in behavior. It’s a line chart,
clean and confident. The problem is the scale: the y-axis runs from 49% to 51%. The change is real, but the chart
makes it feel like the world flipped upside down. People who learn to spot this trick develop a new habit: they
look at the axis labels before they look at the line. It’s the visual equivalent of checking the nutrition label
before believing the “healthy” packaging.

The “Percent Increase” Advertisement: Someone sees an ad that says “50% more effective!” with a
chart that looks like a skyscraper next to a shed. When they dig in, the improvement is from 2 out of 100 to 3 out
of 100. The ad wasn’t necessarily making up numbersit was choosing the framing that sells. The lesson here isn’t
“never trust charts.” It’s “never trust charts that refuse to tell you the denominator.” A percentage without a
baseline is a magic trick with extra steps.

The Dual-Axis “Proof”: In a meeting, someone shows a dual-axis chart where two lines move together
perfectly. The presenter concludes, “Clearly, X causes Y.” A skeptical analyst replots the data with separate
charts and discovers the lines only match because the axes were scaled to match. The team learns two things: dual
axes are dangerously persuasive, and correlation still isn’t causation even when the lines are holding hands.
After that, they start using small multiples and adding notes about what the chart can and cannot prove.

The Map That Blames the Wrong Places: A choropleth map shades states by “total incidents.”
Big states look darkest. Viewers conclude those states are the “worst.” Then someone remakes the map using rates per
capita and the pattern changes completely. The lesson: maps are emotional. They can look like moral judgments when
they’re really just population maps in disguise. People who go through this experience start asking, “Is this
normalized?” before they accept any geographic conclusion.

These experiences build a kind of chart street-smarts. People stop treating visuals as decorations and start
treating them as arguments. And like any argument, a chart deserves cross-examination: What’s the scale? What’s the
baseline? What’s missing? Once you learn to ask those questions, misleading charts don’t disappearbut they lose
their superpower.

Conclusion: The most honest chart is the one that survives questions

If you remember one thing, make it this: a chart is not a photo of reality. It’s a designed objectan argument with
fonts. Design can clarify truth, or it can cosplay as truth. Truncated axes, dual scales, cherry-picked timeframes,
and flashy styling can all push viewers toward a conclusion the data doesn’t earn.

The good news is that you don’t need to be a statistician to defend yourself. Check the axis, the baseline, the
units, and the context. Ask what happens when you zoom out. Look for missing denominators and hidden uncertainty.
And when you make charts, aim for integrity that’s boring in the best way: clear, transparent, and hard to
misreadeven by someone who’s skimming on their phone with one eye open.