I don’t want to get into a heated discussion because I think we may be at an impasse.
I’d rather have our collective readers take up the debate and tell us which positions are more correct or incorrect.

(I’d REALLY love to have Stephen Few give his thoughts!)

That being said let me offer a couple facts and an concession.

#1. The thumbnail graphic for the Media Cloud (according to my handy-dandy Developer plugin in Firefox) is about 200W x 300H pixels for an area of 60K pixels. The original graphic (which you have to screenshot out of the webpage) is roughly 370W x 657H pixels for an area 243K pixels – that’s more than 4 times the size of the thumbnail.

#2. Stephen Few has this to say about chart junk, “…a term coined by Edward Tufte to describe visual content in an information display that serves only as decoration.” (“Information Dashboard Design” by Stephen Few, (c) 2006, p.185) And Tufte’s longer diatribe, “The interior decoration of graphics generate a lot of ink that does not tell the viewer anything new. The purpose of decoration varies — to make the graphic appear more scientific and precise, to enliven the display, to give the designer an opportunity to exercise artistic skills. Regardless of its cause, it is all non-data-ink or redundant-data-ink and it is often chart junk.” (“The Visual Display of Quantitative Information” 2nd Ed. Edward Tufte, (c) 2001, p.107) Some more info is available at Wikipedia.

I could concede on the area of the circles if your graphic provided the reference point. In other words, If I am judging the size of the circles in relationship to a standard reference (a la the standard kilogram) then ok, but since your graphic does not include the legend for size, there is no standard reference to judge the encoded areas against.

1. No link to site: Very fair, and an honest mistake. I always try to link out to the source (you’ll see this throughout the Tableau blog) and in this case thought I had linked the image. But I had not. This is now fixed in the original post.

2. I used the image that was on their site, and covered up the parts that did not pertain to the argument. Since I was focusing on the BBC, I made sure to keep that clear. But placing the viz police logo on their graphic was out of line.

3. The largest graphic I could find even when I clicked into their site was 370 x 190 pixels, so there’s not much larger to go. And yes, you can see the color difference if you stare long at the graphic, but isn’t the point to communicate quickly and effectively without struggling with the data?

4. Re using a different data set: I couldn’t get the original data from the site, so I used a representative sample of one of our graphs. Re-creating the original data set would have been extremely error-prone so I chose to use a different set altogether.

5. While humans do have trouble comparing areas, I don’t quite agree that using areas for measurements is categorically bad. On maps, it can be especially useful. When all areas are relative to a common reference point, as in the circles, they are representative of the data. When they are relative to different quantities that are not being compared, such as the areas of countries on a map, they distort the data. In this example we double-encoded the information with color and size, and focused on the color with the legend. But your point is valid: we could have left size off altogether.

6. The border on the points is unobtrusive and is also consistent across all the points, so in my view it’s not chartjunk. Chartjunk is giant Santa clipped into a viz of Christmas gifts. But this point might be a matter of taste.

7. Again I must disagree- circles on a map, if encoded in color, can represent data well. The consistent placement of these in the center of each state, their size and color don’t lead me to believe they’re cities. And as I argued above, by making each data point relative to a common scale, it’s easier to compare the data.

In any case, thanks for reading and for policing the viz police. We do want to have a great blog that highlights good visual design, so it’s fair to call us on it when we fall short.

Best,

Ellie

Any bounded area suffers from the area bias problem so I agree with you there. Cartograms are actually an excellent example of breaking the rules on purpose. The mental model of the map is broken in a specific way so that the new area is representative of the data more than the geographic boundaries in play.

I wouldn’t mind the dots so much (generally or specifically in the case of their blog post) if they did not encode data on the size and leave out the legend that explains what it is – poor form.

And yes, Tableau in static does not show how powerful a tool it really is.

From personal experience, I have found that using country boundary polygons misrepresents the underlying data because of the area displayed. Cartograms are better with the area issue, but warps the boundaries excessively. I find that Tableau’s method of centering a sized mark of the center of a boundary area a wise compromise.

Also, one cannot see the true power of the Tableau viz without opening it in at least a reader; a screen capture of a Tableau viz is the tip of an iceberg. I am talking about the tool-tips that pop-up on hover, the selecting and drilling down into data, advanced filtering options, and so on. A Tableau viz is not a picture to only look at, but a data viz to interact and have a conversation with.

Sorry, I may have gotten off topic, but I too feel the “Viz Police” missed the mark on all the errors you pointed out and more.