This entry is part of the Chart Index, the reference library for the Chart Design Field Guide.

A polar bar chart wraps a bar chart around a circle. Categories occupy angular positions, values extend outward from the centre, and the result reads as a rose, a clock face or a compass. The form is old — Florence Nightingale's 1858 diagram of army mortality is a polar chart, and the wind rose has been drawn by mariners for longer than that — and it survives because some data genuinely goes round in a circle: hours of the day, months of the year, points of the compass.

The trade is accuracy for shape. A bar on a straight baseline is the most precisely readable mark in the repertoire; the same bar bent around a circle loses its common baseline, sits on a curved scale, and covers more area the further it reaches. The polar bar is the right choice only when the circular arrangement says something true about the data that a straight line would hide.

What it is

Each category receives an equal angular slot. A quantitative value determines how far the bar reaches out from the inner radius along a shared radial scale, and several measures can be stacked within the same slot. Concentric rings mark the value ticks; the category names run around the outside.

The hero chart above is the canonical case: a wind rose. Sixteen compass directions, each slot the share of hours in the year when the wind blew from that direction, stacked by speed band. North is at the top, so the chart reads like a map: the summer north-easterly sea breeze is the lobe at the upper right, the southerly changes are the lobe at the bottom, and the westerlies of winter push out on the left. No straight bar chart could show that the two strongest lobes are almost opposite each other. Randomise the shares and the rose reshapes, but the geography stays put.

When to use it

Polar bar charts are the right choice when:

  • Categories have a genuine cyclic order — hours, weekdays, months, compass bearings — where the last category is next to the first.
  • The overall rhythm of the cycle matters more than the ranking of individual categories.
  • The circle carries domain meaning: bearings on a wind rose, the clock face for a daily pattern.
  • A compact radial overview sits beside a conventional detail view that carries the precise values.
  • You have 8–36 slots. Fewer and the circle is mostly empty; more and the slots become too thin to label.

When not to use it

  • Categories have no circular meaning. Products, regions, teams: a wrapped ranked list is harder to read and says nothing a straight bar does not.
  • Precise comparison is the task. Use a bar chart with a common straight baseline.
  • Values are close together. Small differences vanish on a curved scale.
  • There are many series. Stacking three bands in a slot works; stacking eight does not, and grouped bars within a slot are unreadable.

Design principles

Earn the circle

Use the polar layout only when the domain is cyclic. The pair of figures below plots the same twenty-four hourly counts of emergency department arrivals. On the circle, midnight sits next to 11 pm and the overnight trough reads as one continuous quiet arc; on the straight axis, the same trough is torn in two, half at each end of the chart. That is the one thing the polar form buys — and for this data it is worth having.

Arrivals by hour, on the clock
Emergency department arrivals per hour, averaged over a year, midnight at the top. The quiet stretch from 1 am to 6 am is a single unbroken notch, and the busy afternoon is one broad lobe. The shape is the message.
The same arrivals as a bar chart
Identical counts on a straight axis. Every bar can now be read against the gridlines — 17:00 is 88, 05:00 is 20 — but the overnight trough is split between the two ends of the chart and the cycle has to be imagined.

Do not wrap a ranked list

The failure mode is the opposite case: nominal categories forced onto a circle because the circle looks interesting. The next pair takes eight product lines with no cyclic order at all. On the polar chart the reader has to walk around the ring comparing curved bars of different widths; on the sorted horizontal bar the ranking is simply handed over.

Revenue by product line, wrapped
Eight product lines around a circle. Is Platforms bigger than Decision tools? The bars start on a curve, end at different angles, and the biggest bar is also the widest. The circle has added nothing but doubt.
Revenue by product line, ranked
The same eight values as a sorted horizontal bar. Common baseline, one straight scale, the ranking read top to bottom. This is the chart the polar version was hiding.

Start from a meaningful direction

Midnight at the top and north at the top are conventions readers already carry; use them. Where the cycle has no conventional origin — a financial year, a shift roster — put the start of the reporting period at the top and say so. The two figures below hold the same hourly arrivals; the second begins the ring at 6 am, so the afternoon peak lands on the left and the chart no longer looks like a clock.

Midnight at the top
The conventional clock. Morning on the right, evening on the left, and the eye can read times without consulting the labels.
Six in the morning at the top
The same data starting the ring at 06:00. Nothing is wrong with the numbers, but the shape has rotated ninety degrees, the peak sits where the reader expects the evening, and every reading now needs a label.

Use radial grids sparingly

Two or three reference rings are usually enough. Dense rings turn the background into the most prominent feature, and because every ring is a full circle, each one adds far more ink than a straight gridline would.

Ten rings: the grid is the chart
Monthly rainfall with a value tick every 10 mm. Fourteen rings and fourteen tick labels stacked up the axis: the labels collide into an unreadable column and the rings add nothing a reader can use. Every ring is a full circle of ink.
Three rings: the bars are the chart
The same rainfall with three reference rings. The tick labels are legible, the autumn-to-winter peak reads immediately, and the rings are there when a value is needed, not before.

Keep the hole small and the bars honest

The inner radius exists to give the category labels and the radial axis somewhere to sit, not to make a doughnut. A large hole moves every bar outward, where the same value covers more area, and it hides the fact that bars near the centre are drawn thinner than bars of equal value further out. Keep the inner radius under a quarter of the chart, and start the radial scale at zero — a truncated radial scale exaggerates differences exactly as a truncated bar axis does, with the curve compounding it.

Prefer direct annotation for important values

Tooltips can carry detail, but the key pattern and exceptions should remain visible without interaction. Where the slots are wide enough, print the value inside the arc; where they are not, name the peak and the trough in the caption. The value axis along the start angle is a reference, not a substitute for the numbers the reader will want to quote.

Stack for composition, not for comparison

A stacked polar bar — the wind rose — works because the bands within a slot are parts of one whole and the reader compares slot totals, not bands. Grouped bars within a slot do not work: two curved bars side by side in a twenty-degree wedge cannot be compared. If the question is which of two series is larger in each slot, draw two roses or use a line chart.

Anatomy

The composition of a polar bar chart
An anatomical guide

Angle encodes category, radial distance encodes value, the inner radius sets the baseline that every bar grows from, and circular gridlines provide the value references. The radial axis along the start angle names the scale; the category labels around the outside name the slots.

  • Bar chart — the more accurate default for categorical comparison; use it unless the circle is doing work.
  • Radial bar chart — the other way to bend a bar: series on concentric rings, values as angular sweep.
  • Radar chart — several variables on spokes, connected into a profile rather than drawn as bars.
  • Polar range bar — bars that span a low and a high value rather than growing from the baseline.
  • Coxcomb / polar area chart — Nightingale's variant, where the area of each wedge encodes the value; harder to read than the polar bar and usually best avoided.

Reading list

  • Cleveland, W. S. & McGill, R. (1984). Graphical Perception: Theory, Experimentation, and Application to the Development of Graphical Methods. The ranking of visual judgements that explains why curved bars are hard to compare.
  • Nightingale, F. (1858). Notes on Matters Affecting the Health, Efficiency, and Hospital Administration of the British Army. The polar area diagram in its original setting.
  • Wilke, C. (2019). Fundamentals of Data Visualization. On coordinate systems, and when polar coordinates are appropriate for periodic data.
  • Munzner, T. (2014). Visualization Analysis and Design. On radial layouts and the perceptual cost of angle encodings.