Dot Plot Maker

Paste your category labels in one column and their values in the other, and the tool draws a Cleveland dot plot: one row per category, one dot per value, sorted however you like. Everything runs in your browser and nothing is uploaded.

Position Beats Length

The dot plot was developed by William Cleveland at Bell Labs in the 1980s as a deliberate replacement for the bar chart in ranking comparisons, and the argument behind it is experimental rather than aesthetic. Cleveland and Robert McGill's 1984 work on graphical perception tested how accurately people decode different visual encodings, and produced a ranking. Judging position along a common scale is the most accurate task there is. Judging length comes several places lower, and angle and area lower still.

A bar chart asks you to compare lengths. A dot plot asks you to compare positions on a shared axis — the task people are demonstrably best at. Everything else about the chart follows from that one substitution:

• The bars disappear. What replaces them is a thin guide row and a dot. In Edward Tufte's terms the data-ink ratio rises sharply: almost every mark left on the page is carrying information.
• The chart stops shouting. Twenty bars fill the page with saturated blocks and the eye reads the mass rather than the ends. Twenty dots read as a shape — a curve of values you can trace down the column.
• Long labels get a home. Rows run horizontally, so category names sit in a left-hand column at a natural reading angle. No 45-degree rotation, no truncation to "Integrations dire…", no wasted vertical space.

Why the Axis Need Not Start at Zero

This is the practical difference that matters most, and it is worth being precise about. A bar encodes its value as a length measured from the axis. Move the axis and every length changes, so a bar chart whose axis starts at 140 rather than 0 genuinely misrepresents its data — a value of 150 draws a bar a third the size of a value of 170, when the real ratio is 0.88.

A dot encodes its value as a position. There is no length to distort. So the axis is free to frame the window your data actually occupies, and the chart can resolve differences a zero-based bar chart flattens into nothing.

The obligation that comes with that freedom is honesty about the frame: the axis range has to be legible, and you must not describe differences as proportional when the axis does not begin at zero. "Twice as far along the axis" is not "twice as much".

Comparing Session Duration Across Landing Pages

Suppose you measure average session duration, in seconds, for eight landing pages:

The whole spread is 34 seconds, from 148 to 182. Drawn as bars from zero, the shortest bar is 148 ÷ 182 = 81% of the longest — every bar reaches at least four-fifths of the way across, and the differences occupy the last fifth of the chart. The eight bars look, at a glance, the same. The 3-second gap between Careers and Press releases is 1.6% of a 182-unit bar: invisible.

Drawn as a dot plot with the axis running 145 to 185, that same 34-second spread fills the plot. The 3-second gap now spans 7.5% of the chart width — small, but visibly small rather than absent. And the shape becomes readable: a steady decline through the middle six pages with Careers and Press releases sitting apart at the bottom, which is a finding, not a rendering artefact.

To reproduce it, paste the labels into the left box one per line and the values into the right box in the same order, then choose Descending.

Sorting Is Half the Chart

Category order is a design decision, not a property of the data, and it is the single biggest lever on how readable a dot plot is. Leaving categories in the order they happened to arrive in a spreadsheet throws away most of the chart's value.

• Sort by value for the ordinary case. The chart becomes a ranking, the top and bottom are found instantly, and any cluster or gap in the middle shows up as a visible kink in the column of dots.
• Keep the original order when the categories have an inherent sequence — months, age bands, survey responses from "strongly disagree" to "strongly agree". Sorting those by value destroys information.
• Keep a fixed order when you are producing the same chart repeatedly and readers need to find their own row. A weekly report that re-sorts every week makes tracking one category harder, not easier.

Ascending versus descending is mostly about where you want the reader's eye to land first, since the top row of the chart is read first.

Before-and-After Comparisons

Dot plots are a natural fit for two-state comparisons. Plotted as two dots on the same row joined by a line — a dumbbell or connected dot plot — the gap between the dots becomes the change, read directly as a distance, and the direction is the direction the line points.

This tool draws one series per chart, so the practical route to the same insight is to plot the change itself. Take the landing pages above, subtract each old figure from its new one, and you have a single column of signed values:

Pricing page          +15
Blog — technical posts  +8
Integrations directory  +6
Product tour            +5
Press releases          +4
Careers                 -2
Blog — company news     -4
Customer stories        -5

Sorted descending, this chart answers "what moved?" faster than either of the two originals, because movement is the quantity being encoded rather than something the reader has to compute by eye. Negative values sit left of zero and positive values right of it — one of the few cases where a dot plot's axis genuinely should include zero, since zero is the meaningful reference point for a change.

When a Bar Chart Is Still the Better Choice

Dot plots are not a universal upgrade. Reach for a bar chart when:

• Zero is genuinely meaningful and magnitude is the message. Counts, revenue, headcount — quantities where "how much" matters more than "which is higher" — read well as bars, because the length is a fair encoding of the amount.
• There are only a handful of categories. With three or four rows, bars are easy to read, universally understood, and the ink cost is trivial. The dot plot's advantage grows with the length of the list.
• The values are parts of a whole. Bars can stack; dots cannot.
• The audience is general and the chart is unsupervised. Dot plots are less familiar, and a reader who does not notice the axis range can misread them. In a report nobody will be present to explain, that risk is real.

The reverse test is just as simple. If you have more than about ten categories, or labels longer than a couple of words, or values clustered in a narrow band well away from zero, the dot plot wins on every count.

Two Charts, One Name

Be aware that "dot plot" names two unrelated charts. The one this tool draws is Cleveland's: one row per category, one dot per value, used for ranked comparison. The other — the Wilkinson dot plot, and the one most people meet in school — stacks one dot per observation above a number line to show how a set of measurements is distributed. They share nothing but the word.

If what you have is a list of raw measurements rather than one value per named category, you want the distribution, not a ranking: a histogram shows its shape, and a box plot summarises its centre, spread and outliers.