Similarly, when there are gridlines and zebra stripes together (the gridlines running neatly along the boundary between white and gray), then one of them is surely junk. But which one is a matter of taste.

Great discussion here, as usual…

What about the chart on revenue and plan?

The data table is misleading. I don’t think there is a 95% variance, rather about 5% below plan.

Cheers,
Bob

Regarding the unequal intervals: I’ve already discussed using a different axis scale, such that the horizontal distance between points is proportional to the time between measurements.

Regarding the potential for missing observations: One can only hope that the chart was not distorted by omission of data that did not support the claim (i.e., that there are increasingly more drivers in Texas each year). Leaving out some years’ data because collecting it is a chore is not a crime; leaving out some years’ data because it would show a drop while other years show an increase should be.

Regarding omission of Mike’s Xcelsius book: It would seem that I’ve violated the rule in the preceding paragraph! Let me put forth these excuses:

1. I used Xcelsius for a couple projects several years ago, found it not to my liking, and ignore it now.

2. I would not like to mention the book and inadvertently cause some readers to purchase the book and apply its teachings.

But the real reason I omitted the Xcelsius book is that I simply forgot about it.

BTW, I see that you’ve “conveniently” left out Mike’s Xcelsius book from the list, which is filled with examples that happily violate tip #1 :D

I see my decision to include chart axes in the category of Chart Junk has met with some opposition. ; )

I do see the points you have all made in reference to Tip #2 where I say “Remove Axes that Don’t Provide Value”. I do not mean to suggest that any axis be thrown away cavalierly.

All I suggest here is that you should remove anything that does not contribute to the message in your chart. In some simple charts (such as my example), the data points, the x axis and general trend give you (I believe) the landmarks you need to not only determine the validity of the trend, but to garner the information presented without the need to refer to an axis. Axis labels are not sacred in my book and can be removed if not needed.

I remember reading somewhere that (paraphrasing here) a designer is not done when there is no more to add, but when there is nothing more which can be removed. Tip #2 comes to you in that spirit, nothing more sinister than that.

Instead of
Neither Approve nor Disapprove 50% : ##########
Approve 30% : ######
Disapprove 20% : ####

Do:
Approve 30% : ######
Neither Approve nor Disapprove 50% : ##########
Disapprove 20% : ####