Small Multiples or Panel Charts [Chart Review]

While more companies have decreased or suspended dividends in 2008 than in recent years, the majority still has announced dividend increases.

(source)

This series of charts is just plain awful.  It’s pretty bad when I think that pie charts would have been an improvement to the effectiveness of the data visualization.  There are much better ways to show this type of data.  A series of small multiples or panel charts (mini charts) either as a line or column chart is probably the best approach. 

Here are a few resources to see some excellent examples of panel charts or small multiples.

Processtrends – Kelly O’Day

Peltier Tech – Jon Peltier

Perceptual Edge – Stephen Few

Most Commented Posts



 Subscribe to main feed



11 Responses to “Small Multiples or Panel Charts [Chart Review]”

  1. Jon Peltier Says:

    You don’t consider these small multiples? It’s a series of small ineffective multiples.

    I was going to get the data and mess with different visualizations with it. I went to the source, and that was only an image file, no data. I was going to try to manually digitize the data, which is easy to do with bar or XY charts, any kind with linear measurements. But charts such as pies and donuts have two problems which are not specifically called out, though they are related to other problems:

    1. To digitize an XY or bar chart, I get out my trusty ruler, measure the scale and the position of the points or lengths of the bars, and write a few formulas in Excel. But there’s no easy way to do this with circumferential dimensions. I could use a protractor, but I don’t actually own one; maybe one of the kids has one somewhere that they used in fourth grade. Lacking that, do I try to somehow measure arc length?

    2. The values are really just fractions. There is no mention of the total number of companies, and I don’t think it’s valid to assume it’s constant. If I solved problem 1, I could use the values they give (# of companies that decreased or suspended dividends and back out the other numbers.

    Since it’s just too much work, I’m just not going to follow up further.

  2. Tony Says:

    I think the values in the black bar below the small multiples are more effective and tell the story better than the donuts, CD charts, 45 Charts, pies, tire charts, Olympic ring charts, life preserver charts, Cheerio charts, lifesaver charts, well you get the point.

    If I just add another black bar with the numbers for the companies that increased their dividends, we would have a great visualization… a table.

    Click the link below to see the rule where small multiples can’t be pie charts.
    http://tinyurl.com/5y7w8u

  3. Jon Peltier Says:

    Is that the link you wanted to post? It links back to this page.

    I’m not sure I totally agree with this assertion (without reading the rule you cited). If the small pies only contain two values (like the pies in http://www.fivethirtyeight.com, or if you use simplistic pies like those in some specialty fonts (with onlly 0%, 25%, 50%, 75%, and 100%), I think they could be effective.

  4. Tony Says:

    Yes, I just made up the rule that outlaws pie charts as small multiples…that is until I looked at the website you posted.

    I think they can be used like you stated, if done very simplistically.

    Then we go back to the harvey ball type font or a spot matrix chart shown in the following post: http://tinyurl.com/5m6s9d

    Let’s not stray too far from the original point – they pie charts are very ineffective.

  5. Jon Peltier Says:

    In general pies are less effective than their users believe. For a quick glance, a pie with two wedges can be effective. The pie icons from the harvey ball font are as effective as any icons for showing a limited set of categorically ranked values. I’m not convinced the pie icons are any worse than the spot matrix symbols; which I have to concentrate on more fully to interpret.

  6. Peter Walker Says:

    I just don’t get the donut. Is it just to be different than a typical pie?

    But to the point, this is another instance where a bar chart is more effective. One could argue the presentation uses less real estate than 5 bar charts. However, the gaol is communicating actionable information, and these donuts just don’t do it well.

  7. Tony Says:

    @Jon – I agree that it definitely takes some concentration and thought to figure out the spot matrix symbols.

    @Peter – I do think the donuts are just something different. However, I think they can be even more misleading than a pie because of their layout.

    If it was me, I would have created a panel (line) chart for each segment over the 5 years.

  8. Jon Peltier Says:

    Peter -

    A donut chart is like a pie chart, only it allows more than one concentric ring, therefore more than one series. The distortion is even worse than a pie chart, because the area of each segment is a function of its inner and outer diameters and of the included angle. Leave the donuts for the cops and the pies for the diners, and visit a nice bar.

  9. Tony Says:

    I LIKE IT Jon! I’m writing that one down…

  10. Peter Walker Says:

    Jon -

    I’ll drink to that!

  11. DSA Insights » Blog Archive » Misleading and Unreadable Data Visualization Says:

    [...] Jon recently said, “Leave the donuts for the cops and the pies for the diners, and visit a nice [...]