On Monday of this week the New York Times graphics department gave a talk at the AAG conference to a standing room only audience. The talk, titled Mapping the News in the Age of Visualizations: the Art and Science of the NY Times Graphics Department, was obviously a popular one at the conference so I’ll attempt to summarize some of the things that I got out of it in this post.
Of the 22 people in the NYT graphics department, four presented during this talk: Matthew Bloch, Matthew Ericson, Archie Tse, and Jeremy White. They started by going through a case-study of some of the decision making that has to go on when presenting data visually for the news by describing various ways that the 2008 elections results could be mapped.
First, there’s the typical blue/red by state mapping technique that most everyone is familiar with where states that voted primarily for Obama were colored blue and states that voted primarily for McCain were colored red.
Then they went through slides showing other ways of mapping the data in order to better reflect the population distribution such as taking out all counties where the persons per square mile was less than 3 (colored white), which created a map that showed a more balanced red/blue scheme than the typical map. Also, they tried to extrude cities as 3D bars reflecting population but the New England cities and Los Angeles overwhelmed this map. They tried cartograms but these distorted the geography in the middle of the country so much that you couldn’t tell what color belonged to what city, in, say, Texas. One of the more successful maps was one that depicted party shift from the 2004 election to the 2008 election, shown here:
Now, things were moving quickly in the talk and I’m not sure if I’ve got this down right, but there was a map that got a lot of critical acclaim, and I believe it was this party shift map. However (and again, I hope this is the map that they referred to) some did not like the map because it was not the typical election map, supposedly causing some confusion. Thus we arrive at learning point #1:
If you do something differently when people are expecting something that has always been done a certain way, make sure you make it very clear that the map is DIFFERENT. There was a remark that perhaps some day we will teach people that all maps require a certain amount of time for interpretation by the reader, but until then, people may draw the wrong conclusions if they read the map as though it were a standard election map.
Before this post gets too long, what I’m going to do is just summarize the rest of what I thought were the major learning points. A lot of these came from the question/answer part of the talk. If you were at that session and want to chime in with other bits that I missed please do so. This is certainly not a comprehensive report on the talk since that wasn’t what I set out to do.
- They don’t keep a strict style book. It did eventually come out that they do, indeed, have a color book and a typography book with standard color schemes and 15-20 typefaces, but it sounded like they may not adhere (at least to the color styles – I’d guess they don’t have a lot of leeway with fonts) to these very strictly. Regarding the common style found in all their graphics from print maps to interactive maps and data-displays, it sounds like they do quite a bit of review and eventually everyone winds up with about the same aesthetic. It was noted that they strive for a clear and simple presentation.
- Someone asked what tools they use. I jotted down as many as I heard, though I may not have got them all, here are a few: ArcGIS, Illustrator for the endpoint of all print maps, map publisher, r, various APIs, OSM, TileMill.
- Their design process consists of a lot of preliminary sketching. TAKE NOTE!!!
- When asked how they deal with data of high uncertainty, they jokingly answered that they don’t know. Their real answer was something along the lines of omitting that kind of data from the display. What they mean by this, I assume, is if parts of the country (for example) have uncertain election returns, those areas are colored white instead of red or blue.
- There was some discussion on how they serve up their interactive maps. Not being a complete expert in this arena, what I got out of their answer was that they can’t do it the normal way because they have too much traffic. Therefore, they heavily cache the map so that not every single user is getting a fresh call from the database, maybe only 1 out of 20 users is, and the rest are getting snap-shots.
For a few examples of great New York Times maps, see this post on Map Elegance: Putting the Data First.