Lesson 8 of 9 · about 16 minutes

Charts and Maps That Tell the Truth

You'll be able to

Match a chart type to the comparison you want readers to make
Start bar charts at zero and know when a line chart need not
Map rates per capita instead of raw counts
Pick map color breaks that do not invent patterns
Build and publish a chart in Datawrapper with a source line

A chart makes an argument faster than any paragraph. That is its power, and the reason a bad one does so much damage. Readers trust shapes. Give them a shape that matches the numbers.

The images module, "Creating and Using Images in Articles," has a lesson on infographics and charts that read in five seconds, covering headlines, source lines and color contrast. This lesson is about the data underneath: which chart, which scale, which denominator.

Pick the chart by the comparison

You want readers to see	Use	Avoid
Change over time	Line chart	Pie charts, 3D anything
Ranking of categories	Horizontal bars, sorted	Alphabetical order (unless readers look up their own town)
Parts of one whole, few parts	Stacked bar, or a pie with 2 to 4 slices	Pie with 12 slices
Relationship between two measures	Scatterplot	Two lines on two axes
Where something is higher	Choropleth of a rate	Choropleth of a raw count
A handful of exact numbers	A small table	A chart nobody needed

That last row matters more than people think. Four numbers in a sentence beat four bars in a graphic.

Bar charts start at zero

Readers judge bars by length. If the axis starts anywhere but zero, the lengths lie.

Same two numbers. On the left, cutting the axis at 40 percent makes a 3-point rise look like a fourfold jump. On the right, bars start at zero.

The arithmetic: above a 40 percent baseline, the bars are 1 and 4 units tall, a ratio of 4. The true ratio is 44 divided by 41, about 1.07. A 7 percent difference drawn as a 300 percent one.

Line charts are different. Readers judge lines by slope and position, so an axis from 40 to 46 can be honest for an approval rate that wanders between 41 and 44, as long as the axis is labeled. The danger runs the other way there: a tight axis can make random wiggle look like a trend. Ask whether the change is bigger than the noise (the uncertainty lesson) before zooming in.

Other common distortions

Dual axes. Two lines on two scales can be made to cross, diverge or move in lockstep just by stretching one scale. Use two small charts side by side instead. (Datawrapper's free plan doesn't offer dual-axis charts anyway, as of October 2026, which is no great loss.)
Cherry-picked start dates. Starting a crime chart at the lowest year in a decade makes every later year look like a rise. Show the longest range that is comparable.
Nominal dollars over long periods. Adjust for inflation (lesson 5) and say so in the chart's notes.
Uneven time gaps. Bars for 2010, 2015, 2024 and 2025 spaced evenly hide that the first gaps are five and nine years.

Maps: rates, not counts

A map of raw counts by county is, nine times in ten, a population map. The biggest county is darkest, and readers learn nothing they didn't already know.

Example: four invented counties

County	Cases	Population	Rate per 10,000
North	186	412,000	186 / 412,000 x 10,000 = 4.5
East	44	61,000	44 / 61,000 x 10,000 = 7.2
South	9	7,400	9 / 7,400 x 10,000 = 12.2
West	27	98,000	27 / 98,000 x 10,000 = 2.8

A count map makes North darkest. A rate map makes South darkest and North third. The rate map answers the question readers care about: where is the risk higher?

But South's rate rests on 9 cases. CDC WONDER would flag a death rate built on fewer than 20 events as unreliable. Options: combine three to five years, show counts in the tooltip or a table beside the map, or add a note. Do not let a tiny county's noisy rate become the headline.

Color breaks

Choropleths sort values into color steps. How you set the cutoffs changes the picture.

Equal intervals (0 to 3, 3 to 6, 6 to 9...) are easy to explain. One outlier can push everything else into one or two colors.
Quantiles put the same number of places in each color. Every color gets used, but tiny differences near a cutoff can land neighbors in very different shades.
Meaningful cutoffs, such as above or below the state average, are often the clearest for a news story.

Use one color family from light to dark for "more of something." Use a two-color diverging scale only when there is a real midpoint, like zero change or the state average. Five or six steps is plenty. And remember that big empty counties take up most of the map's ink while holding few people. If the story is about people, consider a simple sorted bar chart of rates alongside the map.

Building it free

Datawrapper is the standard newsroom tool, and its free plan lets you create, publish and embed charts, maps and tables and export PNGs, with a "Created with Datawrapper" credit on each one (as of October 2026). Paste your data from a spreadsheet, pick a chart type, write the headline as a finding, fill in the source and notes fields, and publish. It includes ready-made county maps for the United States.

LibreOffice Calc and Google Sheets both make serviceable charts for checking your own work. Sort, chart, look for the outlier, then decide whether readers need to see it too.

Checklist

Headline states the finding in a sentence.
Bars start at zero. Line axes are labeled and not zoomed into noise.
Rates, not counts, on any map of people.
Small counts flagged or combined across years.
Dollar amounts across years adjusted for inflation, with a note saying so.
Source line names agency, dataset and years.
Someone else read it cold and told you what it says. It matched what you meant.

Reporting angle

When an agency or campaign hands you a chart, ask for the numbers behind it and rebuild it yourself with a zero baseline or a full time range. Compare the two.
For any map of counts in a press release, ask for the population figures used, or get them from the Census Bureau, and redo it as a rate.
Ask what years the chart covers and why it starts where it does. A start date chosen after a record year is worth a question.
Keep the spreadsheet behind every chart you publish, with the exact rows the chart used, so you can answer a reader's challenge the same day.

Common mix-ups

Claim: Every axis must start at zero. What the evidence shows: Bar charts must. Line charts may not need to, if labeled honestly.
Claim: A darker county on a map has a bigger problem. What the evidence shows: Only if the map shows a rate, and the rate rests on enough cases to be stable.

Try it

Make the 41 versus 44 percent bar chart twice in a spreadsheet: once with the axis at 40, once at 0. Show both to someone and ask how big the change is.
In Datawrapper, map any county-level count for your state as a rate per 10,000 or per 100,000, using the same population year for every county.
Try equal-interval and quantile breaks on the same map. Write one sentence about which counties changed color and whether that tells readers something true.
Add a source line and a note flagging any county whose rate rests on fewer than 20 events.

Check yourself

1. Two bars show 41 percent and 44 percent on an axis starting at 40 percent. How much longer does the 44 percent bar look?

2. Which data belongs on a map of rates instead of counts?

3. When may a line chart's vertical axis start above zero?

4. A county map shows South County darkest at 12.2 cases per 10,000, from 9 cases. What should the caption or story add?

5. As of October 2026, what does Datawrapper's free plan require on published charts?

Terms in this lesson

Choropleth map: A map that shades areas such as counties by a value, like a rate or a percent.
Truncated axis: A value axis that starts above zero, which exaggerates differences on bar charts.
Classification breaks: The cutoffs that decide which values get which color on a map.
Dual-axis chart: A chart with two vertical scales for two series, easy to manipulate by stretching either scale.

Full glossary

Sources

Datawrapper, Pricing and plan FAQ checked 2026-10
CDC WONDER, Underlying Cause of Death, Help: unreliable and suppressed rates checked 2026-10