Why the NYT Spiral Graph Is a Failure. And a Success.

If you’re on Twitter and follow anyone who’s even remotely involved in data viz, you’ve already seen this recent graph from The New York Times which shows COVID case counts in the U.S. since the beginning of the pandemic. In this graph, time follows a spiral path, with each 360° rotation representing a year:

This chart basically burned data viz Twitter to the ground, with everyone and their dog either praising or trashing it.

My turn.

Among the chart’s critics, the common concerns seem to be that…

  • The spiral layout makes it unnecessarily hard to precisely estimate and compare case counts at different points in time, and to perceive the overall pattern of change during the past two-plus years.

  • The spiral layout distorts the data because, for example, 2021 looks “bigger” than 2020 because the 2021 circle within the spiral is bigger than the 2020 circle (and they aren’t really circles).

  • Most readers don’t think of months of the year as points on a circle, as they do with hours of the day (on a circular clock) or cardinal directions (on a compass).

Ultimately, they accuse the designer of choosing the spiral design solely as an attention-grabbing gimmick, or as artistic license run amok at the cost of clarity and precision.

Among the chart’s fans, the main praises seem to be that…

  • It’s an unusual chart and so draws readers in.

  • It captures how the pandemic feels like a never-ending cycle.

  • It makes it obvious that the number of cases is higher now than it ever was during past surges.

Which side am I on? Both. Kind of.

From where I sit, this debate is a textbook example of failing to take the purpose of a chart into account when evaluating it. The exact same chart can be terrible in the context of one purpose, but excellent in the context of another. I wrote a blog post about this last April which, I still believe, is possibly the most important article that I’ll ever write about data visualization, precisely because of situations like these.

In order to evaluate this (or any) chart, the first step is to determine what the designer hoped to accomplish by publishing it. I wasn’t in the room for that discussion at the NYT, of course, so I have to guess what they hoped to accomplish with this chart. Unless you were in the room, you’re guessing too, BTW.

If I assume that their goal was to get as many readers as possible to notice the chart and read the associated article, then this chart was probably a better choice than a more conventional, “straight” line chart. In order to know that for sure, though, we’d need run an A/B test, simultaneously publishing two versions the article (one with the spiral chart, one with the straight chart) and then see which one prompts more people to read the article.

If, on the other hand, I assume that their goal was to enable readers to perceive precise insights within the case count history, then I can say with pretty high confidence that this chart is a failure. A controlled experiment barely seems necessary to confirm that a “straight” line chart of some type would better accomplish that goal.

If I had to guess which goal the NYT had in mind when designing this chart, I’d guess that it was the first one. If that’s the case, and if the A/B test that I described would result in more people reading the version of the article with the spiral graph, then it is, objectively speaking, a better chart than a “straight” line chart.

Luminaries such as Amanda Makulec, Neil Richards, and Ben Jones have grasped this crucial subtlety on Twitter, but most others, it seems, haven’t. Most just argue that this chart isn’t “the best way to visualize this data” without ever mentioning what the chart is for. This, however, reflects what I consider to be the biggest misconception in data visualization, i.e., that there’s a “best way to visualize a given data set”, regardless of why the designer was creating the chart in the first place. The reality, however, is that there’s only ever a “best way to visualize a given data set for a given purpose”. If we fail to take a chart’s purpose into account when critiquing it, we’ll always be flailing and disagreeing with one another.

By the way...

If you’re interested in attending my Practical Charts or Practical Dashboards course, here’s a list of my upcoming open-registration workshops.