Chart Chooser Tool: A Free Interactive Cheat Sheet for Picking a Chart
The chart chooser tool asks a few questions and recommends a chart, with the reasoning and traps spelled out, plus a cheat sheet you can scan on its own.
Narrated lesson · C-02
Listen to it 2:07
0:00 / 2:07
Answer one question, sometimes two or three, and a chart appears with the reasoning spelled out
underneath it.
Read the transcript
Answer one question, sometimes two or three, and a chart appears with the reasoning spelled out
underneath it. That is the whole idea behind the chart chooser above: instead of scrolling a
gallery of chart types trying to recognise the right one, the tool asks what the data is actually
for, and narrows the field from there. The first question is always the same: what does the chart need to show? Compare values across
categories, show change over time, show parts of a whole, show a distribution, show a relationship,
or show a ranking. That single choice already rules out most chart types. A follow-up question or
two, usually about how many categories or how many points in time, narrows it the rest of the way
to one recommendation. This idea is not new. A marketing professor named Andrew Abela built a paper flowchart around this
same four-purpose logic, comparison, composition, distribution and relationship, and it circulated
through consulting workshops for years afterward as a decision tree on a single page. The chart
chooser here works the same way but interactively, and it adds a sixth purpose, ranking, since a
leaderboard or a top-ten list behaves differently enough to deserve its own branch. Below the interactive tool sits a cheat sheet: a table with one row per goal, listing which charts
suit it and which charts tend to mislead or slow the reader down for that same goal. It is built to
be scanned in ten seconds by someone who already knows roughly what they want but wants the traps
named before they build the slide. A bar chart with a cut axis, a pie chart with fourteen slices,
a rainbow of unrelated colours on a comparison chart: these show up across almost every row of that
table, because they are the same handful of mistakes recurring in different disguises no matter
which chart type gets chosen. What the tool cannot do is tell a reader whether their data is actually any good, or whether the
underlying numbers are honest in the first place. It only recommends a shape for data that is
already sound. Choosing the chart is the easier half of the job.
Chart chooser
Find the chart for your point
Answer one to three short questions about what the chart has to say. The tool suggests a chart, explains why it fits, lists the traps, and names alternatives. The full table below works as a cheat sheet on its own.
Recommended
Column chart
Bars that start at zero let the eye compare lengths, which is the most accurate visual judgement people make. With few categories or few time points, one column per value stays readable at a glance.
Watch out for
Start the value axis at zero; a cut axis exaggerates differences.
Keep one colour and use a second only to highlight the bar that carries the message.
If category names are long, turn the chart sideways into horizontal bars.
Alternatives
Horizontal bar chart · Line chart (for time series)
Recommended
Grouped bar chart
Placing two or three series next to each other inside each category makes the within-category comparison easy. It works while the number of bars stays small enough to scan.
Watch out for
More than three series per group turns into a comb; switch to small multiples.
Label series directly or keep the legend right next to the bars.
Put the series the audience compares most next to each other.
Alternatives
Dot plot with two dots per row · Small multiples of bar charts
Recommended
Sorted horizontal bar chart
Horizontal bars leave room for long category names and sorting them turns the chart into a ranking the reader can scan top to bottom. It handles 20 or so items before it gets crowded.
Watch out for
Sort by value unless the categories have a natural order (age bands, months).
Start the bars at zero.
Write values at the end of each bar instead of adding gridlines.
Alternatives
Dot plot · Table with inline bars
Recommended
Dot plot
A dot per item on a shared scale uses much less ink than a bar, so dozens of items fit on one slide. Because dots do not need a zero baseline, the scale can zoom into the range where the differences are.
Watch out for
Sort the rows by value.
Show the scale clearly, since the axis may not start at zero.
For very long lists, show the top and bottom few and move the rest to a table.
Alternatives
Table with conditional shading · Sorted horizontal bar chart
Recommended
Slope chart
Two vertical axes, one per date, joined by a line per item: the angle of each line shows the size and direction of the change. It is the clearest way to show before and after for several items.
Watch out for
Label each line at both ends instead of using a legend.
Highlight the few lines that matter and grey the rest.
Only two points in time; the chart hides everything in between.
Alternatives
Dumbbell chart · Paired bar chart
Recommended
Line chart
A line connects points in time order, so the reader sees trend, peaks and turning points in one sweep. Up to about four lines stay distinguishable on one chart.
Watch out for
Label each line at its end instead of using a legend box.
Use a zero baseline only when the absolute level matters; otherwise zoom to show the change honestly.
Keep time intervals even; uneven gaps distort the slope.
Five or more lines on one chart become a tangle. A grid of small charts with the same scales gives each series its own panel, and the eye compares shapes across panels quickly.
Watch out for
Use identical axes in every panel, or the comparison is misleading.
Order the panels by something meaningful, such as the latest value.
Show a faint copy of all series in each panel if context helps.
Alternatives
Line chart with one highlighted series · Heatmap with time on the horizontal axis
Recommended
Pie or donut chart
With up to five parts, a pie shows at once that the pieces add up to a whole, and one large or small share stands out. Beyond that, angles become hard to compare.
Watch out for
Never 3D and never exploded slices; both distort the areas.
Start the largest slice at twelve o'clock and go clockwise.
Write the percentages on the slices; readers cannot judge angles precisely.
Alternatives
Single 100% stacked bar · Sorted horizontal bar chart with percentages
Recommended
Treemap
Nested rectangles show both levels of a hierarchy at once, with area standing for size. It fits many parts in a small space and makes the dominant groups obvious.
Watch out for
Area is hard to compare precisely; label values on the big rectangles.
Colour by the parent group, not by random hues.
Small rectangles become unreadable; group the tail as "Other".
Alternatives
Sorted horizontal bar chart · Sunburst chart
Recommended
100% stacked bar chart
Each bar is one whole split into the same parts, so the reader compares composition across groups. The first and last segments share a baseline and are the easiest to read.
Watch out for
Put the part that matters most at the left edge, where it has a common baseline.
Keep to about five parts.
The chart hides the size of each whole; add totals if they differ a lot.
Alternatives
Grouped bar chart · Small multiples of pies (only with very few parts)
Recommended
Stacked area chart
Layers stacked over time show how the total changes and how the mix inside it shifts. It suits long time series with a handful of parts.
Watch out for
Only the bottom layer has a flat baseline; the others are hard to read exactly.
Put the most stable or most important part at the bottom.
For shares rather than totals, use a 100% stacked version.
Alternatives
Line chart of each part · Small multiples
Recommended
Waterfall chart
Floating bars show each addition and subtraction on the way from a starting value to an end value. It answers "what moved the number" in a single view.
Watch out for
Colour increases and decreases consistently and explain the colours once.
Anchor the first and last bars on the axis so the totals are clear.
Keep steps to about eight; group minor items.
Alternatives
Sorted horizontal bar chart of the changes · Table with a running total
Recommended
Strip plot (dot per value)
With few values, every value can be shown as its own dot along a scale. The reader sees the range, clusters and outliers without any summary hiding them, and rows make groups comparable.
Watch out for
Jitter or make dots semi-transparent so overlapping values stay visible.
Mark the median with a line if a summary helps.
Keep the same scale for every group.
Alternatives
Box plot · Table of sorted values
Recommended
Histogram
Counting values into equal-width bins shows the shape of a distribution: where the bulk sits, whether it is skewed, and whether there are two peaks. An average alone would hide all of that.
Watch out for
Bin width changes the picture; try a few and pick one that shows the real shape.
Bars touch because the bins are continuous.
State the bin width and the number of values.
Alternatives
Density curve · Box plot
Recommended
Box plot
Each box summarises a group with its median, middle half and range, so several distributions line up side by side for comparison. It is compact where histograms would need a panel each.
Watch out for
Many audiences do not know how to read one; add a one-line key.
A box hides two-peaked shapes; check the data first.
Consider overlaying the individual points when there are not too many.
Alternatives
Violin plot · Small multiples of histograms
Recommended
Scatter plot
Each item becomes a point placed by two values, so patterns such as correlation, clusters and outliers are visible immediately. It is the standard chart for the relationship between two measures.
Watch out for
Say what one point represents (a city, a patient, a month).
A trend line helps, but correlation is not causation; do not write the title as if it were.
Label only the outliers you discuss.
Alternatives
Binned heatmap (very many points) · Connected scatter plot (if the points are in time order)
Recommended
Binned heatmap (2D histogram)
With tens of thousands of points, a scatter plot turns into a solid blob. Counting points into a grid of cells and shading by count shows where the data is dense.
Watch out for
Use a single-hue colour scale from light to dark.
Show the colour key with actual counts.
Cell size changes the picture, just like histogram bins.
Alternatives
Scatter plot with transparency · Hexagonal binning
Recommended
Bubble chart
A scatter plot where the size of each point carries a third variable. It suits a small number of items where one of the three values is a volume, such as population or revenue.
Watch out for
Scale bubbles by area, not diameter.
People read size far less precisely than position; put the most important variables on the axes.
Overlapping bubbles need transparency or fewer items.
Alternatives
Scatter plot with colour for the third variable · Small multiples of scatter plots
Recommended
Correlation heatmap
With four or more variables, a grid of every pair shaded by its correlation shows which variables move together. It works as an overview before you pick one pair to show in detail.
Watch out for
Use a diverging colour scale centred on zero.
Order the variables so related ones sit together.
Follow up with a scatter plot for any pair you draw a conclusion from.
Alternatives
Scatterplot matrix · Parallel coordinates
Recommended
Bump chart
A bump chart plots rank rather than value over several dates, so crossing lines show who overtook whom. It is built for league tables and top-10 lists that change.
Watch out for
Rank hides the size of the gaps; mention the values if they matter.
Highlight a few items and grey the rest.
Label lines at the right-hand end.
Alternatives
Slope chart (two dates) · Line chart of the underlying values
Chart cheat sheet
Start from the goal of the chart, not from the data type. Each row lists charts that suit the goal and charts that usually mislead or slow the reader down.
Goal
Recommended charts
Avoid
Compare categories
Column or bar chart, sorted horizontal bars, grouped bars (2 to 3 series), dot plot (many items)
3D bars, a cut value axis on bars, pies for comparing unrelated items
Change over time
Line chart, column chart (few points), slope chart (two points), small multiples (5+ series)
One pie per year, more than four overlapping lines, two value axes with unrelated scales
Parts of a whole
Pie or donut (up to 5 parts), 100% stacked bar, treemap (hierarchy), waterfall (additions and subtractions), stacked area (over time)
3D or exploded pies, pies with 6+ slices, rows of pies to compare
Distribution
Histogram, box plot, strip plot (few values)
A single average in a bar, pies, line charts across unordered bins
Answer a question or two above and a chart type appears, with the reasoning underneath it and the traps named alongside it. That is the entire premise of the chart chooser tool: instead of scanning a gallery of chart types hoping to recognise the right shape on sight, the tool asks what the data is actually for first, then narrows the field.
How the chart chooser tool works
The first question is always the same: what does this chart need to show? Compare values across categories, show change over time, show parts of a whole, show a distribution, show a relationship between two things, or show a ranking. Purpose first, chart second. That single choice already eliminates most of the field. A follow-up question or two, usually about how many categories exist or how many points in time are involved, narrows it the rest of the way to one specific recommendation, complete with what to watch out for and a couple of alternatives worth considering if the first suggestion doesn’t quite fit the slide.
This purpose-first approach is not a new idea. Marketing professor Andrew Abela built an early paper flowchart around 2009 using a similar four-way split, comparison, composition, distribution and relationship, and it circulated through consulting workshops for years as a decision tree handed out on a single page. Juice Analytics later built its own chart-picking templates crediting Abela’s classification model directly. The chart chooser tool above works from the same underlying logic, but interactively, and it adds a sixth branch, ranking, since a leaderboard or a top-ten list behaves differently enough from a plain category comparison to earn its own recommendations, including a bump chart for a ranking that moves across several dates.
The logic behind the cheat sheet
Below the interactive tool sits the chart chooser cheat sheet, a compact table with one row per goal, pairing the chart types that suit that goal against the chart types that usually work against the reader for that same goal. It exists for a different moment than the interactive tool above it: someone who already has a rough idea what chart they want but needs the traps named quickly, without clicking through three questions to get there.
The pattern worth noticing across that table’s “avoid” column is how few distinct mistakes it actually contains, dressed up differently row by row. A cut value axis on a comparison chart and a truncated axis on a time-series chart are the same error in two costumes. A rainbow of unrelated colours on a bar chart and a 3D tilt on a pie chart both distort the same thing, a reader’s ability to judge size accurately, just through different mechanisms. Learning the small handful of underlying mistakes matters more than memorising which chart type each one happens to attach to, since the same handful keeps reappearing across the whole cheat sheet.
A cheat sheet is not a new idea either
A single table that turns a complicated judgement into a quick lookup shows up well outside chart design too. The NHS’s own risk matrix for risk managers, published in 2008, scores any risk on a 5×5 grid, consequence multiplied by likelihood, and colour-codes the result from green to red so a member of staff without risk-management training can read the severity of a situation in seconds. The chart chooser cheat sheet is built on the same premise: compress a judgement that would otherwise take real expertise, or a slow scroll through a gallery of chart types, into a table that anyone under deadline pressure can scan correctly on the first look. Neither table replaces the reasoning underneath it; both just make that reasoning fast to reach when there isn’t time to work through it from scratch.
Why purpose beats appearance
A gallery approach to chart selection, scroll until something looks right, fails in a specific and repeatable way: it lets a chart’s familiarity substitute for whether it actually fits the data. Pie charts get picked constantly because they are instantly recognisable, not because a fourteen-slice pie serves the reader better than a sorted bar chart would for the same numbers. Starting from purpose instead of appearance removes that shortcut. If the goal is a ranking of twenty-five items, the tool never offers a pie chart as an option in the first place, because a pie was never built to answer a ranking question, no matter how familiar it looks on the slide.
The full reasoning behind each of the six purposes, worked through chart family by chart family with more chart type examples than fit in a quick tool result, lives in the which chart should I use guide. That page is the place to go for the “why” behind a recommendation; this chart chooser tool exists for the moment a recommendation is needed fast, with a spreadsheet already open and a deadline already close.
What the tool won’t tell you
The chart chooser tool recommends a shape for data that is assumed to already be sound. It has no way to check whether the numbers behind a chart are complete, correctly sourced, or presented with an honest axis, and a perfectly chosen chart type built on a truncated axis or a cherry-picked date range is still a misleading chart. That is a separate skill, covered directly in how to spot a misleading graph, and it is worth treating as the second half of the job, not an optional extra once the chart type is settled.
Questions
How does the chart chooser tool work?
It asks what the chart needs to show — comparison, change over time, parts of a whole, distribution, relationship or ranking — then one or two follow-up questions about how many categories or points are involved, and returns a recommended chart with the reasoning and common traps.
Is the chart chooser tool free to use?
Yes. It runs entirely in the browser, needs no account, and returns a recommendation in a few clicks, alongside a cheat-sheet table underneath that works as a quick reference on its own.
What is the chart chooser cheat sheet?
A table with one row per goal — comparing categories, showing change over time, and so on — listing the chart types that suit that goal next to the chart types that usually mislead or slow the reader down for the same goal.
Where did the chart chooser idea come from?
Marketing professor Andrew Abela built an early paper flowchart around comparison, composition, distribution and relationship that circulated through consulting workshops for years. This chart chooser tool works from the same purpose-first logic, in an interactive form, with a sixth branch added for ranking.
Does the chart chooser tool cover every chart type that exists?
No. It covers roughly twenty common chart types across six purposes, chosen because they cover the overwhelming majority of real presentation and report data. Rare or specialised chart types built for one narrow field are outside its scope.
Can the chart chooser tool tell me if my data is trustworthy?
No. It only recommends a chart shape for data that is assumed to already be accurate. Whether the underlying numbers are honest, complete and correctly sourced is a separate question the tool cannot answer.
