Writing & MediaFree
Data Story Chart Selection Brief
Send the analyzed dataset and get the chart that fits it, the value table beside it, and every framing that would have exaggerated the result.
Every chart-selection guide gives the same decision tree: a line for change over time, a bar for ranking, a map when geography is the variable. Datawrapper's own chart-type guide works exactly that way, and it is good advice as far as it goes. None of them ask you to name the chart you rejected and say what number it would have shown instead. None of them ask you to publish the table behind the picture, so a reader can check the chart against the actual values rather than trusting the shape of the line.
This brief names both. It takes the analyzed dataset and picks the chart type the data's shape actually supports, then lists every rejected framing with the number it would have produced. On a worked file, a district's chronic absenteeism rate rose from 14.8 percent in 2019 to 28.3 percent in 2024, a 91 percent rise, shown honestly as a zero-baseline line chart across all six years. A bar chart truncated to start at 10 would have shown that same rise as 3.8 times taller, nearly double the true 1.9-to-1 ratio.
Built for reporters and data desks turning one finding into a publishable graphic, especially when the number is politically live enough that somebody will look for the manipulation first. Send the analysis once the sensitivity check is done, or the cleaned dataset if the chart is the first thing built from it. Before the piece runs, finding verification puts the number itself to the subject, since a chart is only as honest as what it is drawn from.
Why the axis is doing more work than the data
A truncated y-axis is the most common exaggeration because it survives a casual glance. Northwestern's Kellogg Insight cites research finding that even readers warned about a cropped axis, and asked to write down the actual values, still judged the change as far larger than it was. On the worked file, starting the same bar chart at 10 instead of zero turns a 91 percent five-year rise into a bar that looks 3.8 times taller, when the honest ratio is 1.9 times. The axis produced the extra height, not the data.
Not every chart needs a zero baseline, and treating all of them alike is its own mistake. The University of Washington's Calling Bullshit course materials explain why: a bar's ink represents the whole value, so cropping its axis misstates the value itself. A line's ink represents the change between points instead, so a tighter range can sharpen a real trend without hiding one. The rule this brief applies is that the chart type decides the axis, never the size of the story.
A chart can mislead with an honest axis and a dishonest window just as easily. Kessler County's six-year series rose steadily except for one pandemic-disrupted year, when the rate jumped from 15.1 to 26.9 percent, a 78 percent jump in twelve months alone. Framed by itself, that single year implies a crisis accelerating at four times the honest five-year average. Framed inside the full series, it is one disrupted year followed by a plateau. The brief runs both windows and reports which one the complete data actually supports.
How it works
Send the finding
Attach the dataset, state the finding as you'd publish it, and say what shape the change takes.
River tests the shape
The data's own shape decides the chart type, not a house style or which one looks more dramatic.
Every rejection gets named
Each framing that would have shown a different number is listed with the number it would have produced.
Work it in chat
Ask for an alternate chart type, a different window tested against the same data, or the sheet re-sorted.
What you get
- The chart type your data's shape supports, not a house style or the most dramatic option
- Every value behind the chart, published as a sheet a reader can check against the picture
- A named list of the framings that would have exaggerated the finding, each with the number it produces
- A short document explaining why the chosen form fits this specific finding, in plain, checkable language
- Axis and window choices checked against the specific finding, not a generic honesty checklist repeated on every story
- Built to hand off: the sheet feeds straight into whatever layout or publishing tool the desk already uses
Common questions
Why not just use whatever chart type looks best?
Because 'best' usually means most dramatic, and the two are not the same thing. A bar chart with a truncated axis often looks more striking than an honest one, which is exactly why the choice needs a stated reason. This brief picks the type the data's own shape supports and says why, so 'best' means accurate first.
How does it decide which framings to list as rejected?
It tries the standard manipulations against your own numbers: a truncated axis, a narrower window, a dual axis chosen to suggest correlation. Each one gets computed, not just named, so the doc reports the specific number a truncated axis at 10 would have shown, not a general warning that truncation exists.
What if the honest chart makes the finding look small?
Then it looks small, and the doc says so plainly rather than reaching for a framing that would fix that. A real 91 percent rise over five years is still a large finding on a zero-baseline chart; it just will not look as sudden as a cropped axis would make it look.
Does this replace a designer or a graphics desk?
No. It replaces the guessing that happens before a designer gets involved: which chart type the data actually supports, and which axis or window choices would misstate it. A designer still builds the final graphic; this brief is the honest brief they build it from.
What format are the chart and the sheet?
The chart comes as an editable native chart inside a document, not a flattened image, so a designer can restyle it without rebuilding it from the data. The sheet is the full underlying series, one row per category or period, exactly as it would be published.
Can it check a chart someone already built, rather than build a new one?
Yes. Send the existing chart's axis range and window alongside the full dataset. It reports whether that choice matches what the data supports, or falls into one of the framings that would have exaggerated the finding, with the honest number stated beside it.
Data Story Chart Selection Brief
Fill in the form and your workspace opens with the work already underway.