> This looks suspiciously like logarithmic scaling which, > as XKCD readers will attest, is a pretty typical > technique for representing statistics which h

> This looks suspiciously like logarithmic scaling which, > as XKCD readers will attest, is a pretty typical > technique for representing statistics which have > differences measured in orders-of-magnitude. It is, but never implicitly. Sure, you can get the actual proportions from reading the markings at the bottom, but then what's the point of visualizing in the first place? Either make your logarithmic scaling explicit and clear (as is done in scientific graphs), or use generally assumed linear scaling. > it's not as if Nielson only bumped Nokia, RIM and > Microsoft's column widths. No, but it amounts to the same thing.