shape library
Same data.
Different shape.
Different story.
Fifty-eight counties, one file, four shapes. Learn what each shape shows, what it hides, then build your own.
Start with the barEvery shape ↓
New here? For families & schools →
Every shape a chart can take, and what each one hides.
Every chart throws something away — that is what makes it a chart, and choosing what to drop is the whole skill. Four tiers, simplest shape first. Open any card for two worked examples from real public data. The ladder asks the same question harder each year, the dives work it twenty-two times, and the pool is where you choose for yourself.
cartesian · origin at (0, 0)polar · origin is the centregeographic · the map is the framenone · says whydashed rule · named, not shipped
01Basics · 7 shapes
One question, one answer, minimal prerequisite. Where every student starts.
the table
Every value, exactly as it was recorded. Nothing is summarised and nothing is hidden.
the number
One value, on its own. It is not a trend and not a spread — just the one number.
the line
A line follows one thing over time. Read it left to right and watch which way it leans — that lean is the whole story. A dashed stretch is a guess, not something anyone measured.
the bar
Bar length is the value, so it only means anything measured from zero. Start the axis anywhere else and the lengths stop matching the numbers.
the donut
Slices add up to the whole thing. Good for parts of one total, bad for ranking.
the picture graph
One little picture per thing, lined up so you can count them.
the pictograph
One symbol stands for several things, and the key tells you how many. A half symbol is half that many.
02Intermediate · 12 shapes
More than one moving part: a distribution's shape, a relationship between two variables, a many-node structure.
the histogram
The bars touch because they’re slices of one scale, not separate things. Step back and read the shape — where the data piles up, and where it thins out.
the scatter
Each dot is one thing measured twice. A tight cloud along a line means the two move together.
the dotplot
One dot per thing, stacked up wherever it lands. You can still count every dot, and the piles show where they gather.
the box plot
The middle half of the data sits in the box. It shows the spread, not how many.
the matrix
Colour is the value here, not where a square sits. The diagonal always glows brightest, and it means nothing — that is each thing compared with itself.
the dumbbell
Two dots joined by a line. The length of the line is the finding, not either dot.
the slope chart
Two moments joined by a line. Which things went up, which went down, and by how much.
the rose
Each wedge points a direction, and its length is how many landed that way. There’s no left-to-right, so don’t read it as a trend.
the calendar grid
The days laid out as a calendar, shaded by how much happened. Weekly rhythms — quiet weekends, busy Mondays — jump straight out.
the flow diagram (sankey)
Ribbon width is the amount. Whatever flows in has to flow back out.
the radar
It’s the overall shape that tells you something, not any single arm. Read the whole outline; one point on it barely matters.
parallel coordinates
Each line is one thing traced across every column. Watch for the line that cuts against the grain — that’s your odd one out.
03Geospatial · 6 shapes
Where, not just how much. Maps are charts, and they fail in their own ways.
the choropleth map
Darker means more, lighter means less — one number, shaded region by region. Big regions grab the eye whether or not many people live there.
the proportional symbol map
How much is at one exact spot. The area of the circle is the value, not its width.
the dot density map
One dot for a fixed number of things, dropped where they happen. Where the dots crowd together, so does the thing.
the flow / origin-destination map
How much moved from where to where. Line thickness is the amount.
the hexagonal binning map
When there are too many dots to read, this counts them into tiles so you can see where they pile up.
the bivariate choropleth
Two things at once on one map, blended into a single colour per area.
04Data science · 5 shapes
Many variables at once, and the question of whether what you see is real.
the bubble chart
Compare things on three or four measures at once — position, colour and size all carry one.
small multiples
The same chart repeated once per group, so you can see whether the pattern holds everywhere.
the scatterplot matrix
Every pair of measures plotted against every other, so you can spot which ones move together.
the strip / jitter plot
The raw values themselves, side by side, before anything is summarised.
is this real, or noise?
Shuffle the data many times and see whether your result still stands out. If it does not, it was luck.