Reflection on two design quirks

This post is for paid members only. Log in or subscribe to see it. Thank you members for supporting my work.

When I first saw this slanted column chart on Visual Capitalist (link), I feel this may be another case of questionable design distorting data representation.

It isn't so simple.

Slanting the columns does not actually distort the encoding of the data. Take a look at the last column on the right- where the designer drops a perpendicular from the top rung of the ladder to the "floor" of the chart. In so doing, a right-angled triangle has been outlined. The length of the slanted side is the hypothenuse, the length of which is (height)*sin θ where θ is the angle of the slant at the floor. Thus, the ratio of lengths of two slanted sides (x1/x2) = (h1/h2) after the sin θ factor cancels out as each column is given the same slant.

For this chart, readers are mostly interested in the year-on-year change: on a conventional column chart, this is reflected in the difference in heights between successive columns. Now, (x2-x1) = (h2-h1)*sin θ so the measured difference in the lengths of successive slanted sides is proportional to the measured difference in heights of the columns. The observed ratio is a constant multiple of the actual ratio of the data. For this usage, there is an absolute distortion but not a relative distortion.

In sum, data distortion is not a strong enough reason to disapprove of the slanting feature.

***

I'm also fascinated by the designer's end run around the start-at-zero rule for column charts. While not explicitly stated, the floor of each ladder can be thought of as starting at zero. The use of the broken scale essentially resets the scale to start at 18 so the chart in reality starts at 18 rather than 0.

(Such distortion of the data encoding impacts the calculation I did in the section above, because I have assumed starting at zero. But the culprit would then be violating the start-at-zero rule.)