24

This is not a donut chart.. not showing distribution here. These are different attributes contributing to the total score (in the middle)

Chart Type

Another example: enter image description here

The thickness of the wedge is proportionate to the severity of the alert type.

Is there any other chart that works well for this type of data?

  • 1
    do you intend to use this chart? or are you open to exploring a few alternatives to solve the problem you're working on? – Mike M Jun 14 at 15:01
  • 3
    Looking for alternatives. – essdeepee Jun 14 at 15:38
  • 44
    It's the answer to a bad joke about a doughnut chart and bar chart getting together. – 習約塔 Jun 14 at 18:28
  • 24
    @xiota So a donut walks into a bar ... – kapex Jun 14 at 22:00
  • It looks kind of like a bastardised one of these charts, but I have no idea what to call it. – Someone Somewhere Jun 15 at 8:34
10

Since the attributes are not distribution, you could show the initial score, and place a severity breakdown below. Allow hovertips and details for drilldown.

Instead of focusing on the current chart (the implementation, or 'how'), maybe we can address the possible questions in the users mind (the problem space).

  • What is my current threat/security score? Should I be concerned?
  • What's contributing to it?
  • Which attributes are at dangerous levels?
  • How can I get more details on this? (if there's a problem)

Here's a first sketch to prompt a discussion. I'll update my initial answer as needed.

It seems like you have:

  • a total score
  • its attributes (composition)
  • a severity level for each attribute, which looks like it's tied to the count of a specific attribute.

Understanding attributes and severity levels

For a severity level, it appears that some attributes can have a higher count, but still have a lower severity level. Is that correct?

enter image description here

A tabular breakdown might seem less 'visual', but with a limited amount of attributes, you can order by severity, and show a decent amount of information in a relatively small space.

And unlike a polar chart, the labels aren't spaced out too far, so your eyes don't need to move much to take in all the attributes, their counts, and severity level.

If so, you could have a column that measures the severity, with a 'no threat' for attributes that are 'in the clear'.

The difficult part could be in making a universal Severity scale, not necessarily tied to the count.

  • Simple and nice representation! Wouldn't the label "Score breakdown" confuse the user? Risk score is derived based on some calculation/formula where these events are attributing to it. – essdeepee Jun 17 at 9:25
  • @essdeepee you might be right. Also, have users tried to add the count around the chart example you originally posted? – Mike M Jun 17 at 13:04
56

This is an absolute 'No no' in creating charts. This is a very bad way of representing data, to make it look like Donut chart but non-functional. It doesn't matter if the thickness is proportional to the value, that is a secondary dimension. The primary dimension must be proportional to the primary value. Having a regular Donut chart is adequate for this problem.

  • 11
    Even a "proper" donut chart is often not the best choice for data visualization, so these examples are actively failing at being sub-optimal! – maxathousand Jun 14 at 15:58
  • 6
    what's the point of a donut chart? are they less fattening tan pies or something? – Jasen Jun 15 at 8:14
  • 9
    It looks kind of like one of these, but squished down. The data is in the radial axis, which makes sense when you have things that don't necessarily sum to 100%. – Someone Somewhere Jun 15 at 8:36
  • @Jasen I think it's just graphic designers getting bored and wanting new things – Redwolf Programs Jun 16 at 15:25
  • 2
    @Celos maybe I should have said “failing at even being sub-optimal”, meaning they’re not even doing a poor job correctly. – maxathousand Jun 17 at 11:53
34

I wouldn’t call this a “chart”, but rather numbers with decorations. The numbers don’t add up to the total, and the pie sizes are decorative rather than informational. Sizes don’t make sense either. Using this will create more confusion than answers.

The most approximate chart to the provided images is a Donut Chart. Refer to this link about how to use donut charts.

  • 3
    "The numbers don't add up to the total" what total? The numbers aren't even fractions of some total for them to add up to. – AJMansfield Jun 15 at 15:27
  • Not sure what the numbers are representing. – Nicolas Hung Jun 15 at 16:28
  • 1
    @AJMansfield: The number in the center looks like it ought to be a total of something, but it is not. Perhaps it is a different kind of summary statistic, but if so, it's rather unclear what it represents. I could imagine multiplying each quadrant by a numerical weight and then summing the products, but there is no indication of such a scheme on this "chart." – Kevin Jun 16 at 19:03
  • @NicolasHung The number in the center is the Risk score, which is derived based on some calculations/formula where different events are contributing to it. It is not the sum of events count. – essdeepee Jun 17 at 13:56
21

Topologically, it looks the same as a Spider chart with 4 variables, but each variable is displayed as a quarter-circle instead of a point.

comparison to spider chart]

17

That looks like a form of polar area diagram.

Also known as a ‘rose diagram’, this is almost 200 years old, and was popularised by Florence Nightingale to highlight death rates in different months:

enter image description here

It uses equal angles, with the radius indicating the magnitude.  (Though that means the area isn't proportional.  Alternatively, you could use area to indicate the magnitude, but then the radius wouldn't be proportional…)

It's most often used with sequential data, often larger sets, but I found this example with 4 separate values:

sample polar area diagram

That doesn't use the full circle, and isn't annular (with a hole in the centre) like the chart shown in the question.  But I don't think those are very important.

Personally, I don't like the lack of proportionality in the areas, nor the use of a sequential chart for isolated values, nor the difficulty of comparing non-adjacent values.  So I'd probably suggest something more like a bar chart in this case.

But there's ample precedent (some of it historically important) for this sort of diagram!

  • For me the one in the question is not a polar chart because it has a different scale for each variable. Easily seen because the value 1 has the thickest line associated with it. In the question the radius doesn't indicate magnitude, it indicates... well, I don't know really. – Some wandering yeti Jun 17 at 9:12
  • 1
    @Somewanderingyeti OP has said that thickness is magnitude of severity – Baldrickk Jun 17 at 10:40
2

It looks like a 2x2 matrix diagram combined with a donut chart in the center that shows the total count. A bar chart or heatmap table would be a better form of communicating the relationship between the numbers, e.g:

enter image description here

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.