Check out the 5-minute Ignite Seattle talk I gave on this project:
As you might gather from my blog's archives, I really enjoy maps. They are a widely understood and broadly engaging way to convey information, especially when overlaid with other data.
One of my favorite data-maps is a well known piece by Stephen Von Worley: The Contiguous United States Visualized by Distance to the Nearest McDonalds. The furthest you can get within the USA is ~107 miles, incidentally. It's a brilliant post; fun, personal, and on a subtle level is discussing a deep part of Americana.
Another chain restaurant by which we might measure our lives, particularly in the PNW, is Starbucks. This Seattle-based-behemoth has played an integral role in the coffee culture around the world, and along with the Nerd Triumvirate (Microsoft, Boeing, and Amazon) has secured our city's place on the global stage.
Starbucks has been used in the past to gauge economic health, and I've found it used as a standard metric in geography classes. The closer you live to a Starbucks the higher your rent is likely to be, and in NYC the density of locations can reach as high as 150 with a radius of 5 miles!
I wanted to look at not only how Starbucks were distributed across the USA (like in Von Worley's McMasterpiece) but how we are distributed around Starbucks.
Here is the USA as mapped by Starbucks-owned locations (with thanks to my friend David B), connected using a Delaunay triangulation. As with Mc D's, these latte-slingers are clustered around major cities and highways. As an aside, I wonder if anyone has tried to calculate the optimal path for visiting every location...
Related to the Delaunay triangulation, here is the Voronoi diagram for every Starbucks location. This beautiful grid of spaces can be used to tell you the furthest point from a company-owned Starbucks (about 170miles 140miles, though note that other franchise locations are relatively nearby, e.g. inside of grocery stores)
Finally: by comparing Starbucks locations to everyone's favorite data source (the US 2010 Census) we can make a graph with a truly impressive result (in my opinion). By counting the number of people who live within a given distance to each Starbucks, we can measure how well centered Frappuccinos are to the US citizenry. In other words: draw a 1-mile circle around every store, then add up the % of the population living within the circles. Repeat for 2, 3, 4....100 miles. What I found left my jaw hanging...
There are ~311 million people living in the USA, with 82% living in urbanized areas. One might define urbanization in the modern era as the distance to the nearest Starbucks. An "urban" environment would therefore be anyplace within a 20 mile radius. Yes, more than 80% of the USA (that's 250,000,000 people) live within 20 miles of a Starbucks.
While it might seem silly to drive 20 miles for a cup of Pike Place Roast, you would definitely drive that far to Lowes or Costco or Ikea... and you might get thirsty while you're out! And thus the burr grinder of progress spins on.
Update: You can now buy a poster version of this map. Neat!
Note: This article has been updated thanks to a bug pointed out in the comments. I always believe in being forthright, and have been loving the dialogues this post has generated. Keep 'em coming!
[ Leave a comment ]
Very nice! Now, draw a T-Rex on my latte' please.
ReplyDeleteWhat data source is this from?
ReplyDeleteThis post was picked up by the Atlantic Wire!
ReplyDeleteWhat did you use to generate the Voronoi and Delaunay triangulation visualizations?
ReplyDeleteThe log scale on the chart implies 10% live inside a Starbucks, with at least a few percent swimming in coffee their whole lives.
ReplyDeleteI notice the maps above show only the 48 contiguous states. While Hawaii has its share of Starbucks, and ocean around the islands to keep you from getting too far from them, I'm curious what the results would look like for Alaska.
ReplyDeleteThis is great stuff, thanks! In considering what you included here I actually wondered aloud what a similar map would look like for office locations of the USPS. Would that map look more evenly distributed that the SBUX map? Would you be willing to do that map too and overlay them? It might be very interesting... ;-)
ReplyDeletePike Place is an actual place in seattle, a market
ReplyDeletenot "pikes"
Right you are! That's a (sorta) typo on my behalf, and as a Seattle resident is inexcusable. However - while Starbucks correctly refers to their beans as "Pike Place Roast", their website URL calls it Pikes Place: http://www.starbucks.com/menu/drinks/brewed-coffee/pikes-place-roast
DeleteColloquially many people call it Pikes Place.
Typo Corrected!
Interesting post. I'm curious about your "furthest point." That point, according to Google Maps, is only 95 miles from Casper, Wyo. Casper is home to a pair of Starbucks locations. (There is a third, but it's in a grocery store.) Where are you getting the 170 mile figure? (Also, you have it listed as 107 in the second graph.)
ReplyDeleteYou raised a great point here, and I went back and checked all my math. Turns out I had a trivial indexing error, which made my maps look slightly wrong, and as a result the distance/position was off. The now corrected location is in Grand Teton National Park, and the distance is ~140 miles.
DeleteThanks for reading so carefully! I'm always happy to correct things!
Thanks for the explanation and update.
DeleteUntil last year, my town was 74 miles from the nearest Starbucks, despite being home to a university with 6,000 students as well as a separate medical school. But then the regional hospital added a limited selection of Starbucks drinks to its gift shop's menu, so I guess we technically have one now?
ReplyDeleteI'm intrigued by the second starbucks map but I don't understand it (or maybe I'm not understanding Veroni diagrams). Why does it appear that there are a number of Starbucks out in the ocean? I ask this in the spirit of the only dumb question is the unasked question. thanks!
ReplyDeleteYou should do Subway next. I'm betting that it's worse than McDonalds and Starbucks.
ReplyDeleteI see an issue with your methodology.
ReplyDeleteSimply drawing a circle around areas with Starbucks locations is all right for estimating how far we all live from one - as the crow flies. However, we do not fly to Starbucks (okay, maybe the ones in the airports); we drive. Predicting mileage from Starbucks, if it were to be modeled on real-life behavior, would not be a circle; it would be a polygon, with the points forming 20 miles by road away from the coffee shop.
In flat areas this may not affect your data much. However, in mountainous areas, your data would be very much altered. In my work in the rural West I've been less than 20 miles as the crow flies from a Starbucks... but as much as 2 hours and 60+ miles by road. If you are on the other side of a mountain range, your driving habits and times are radically different, as all roads do not leads to a Starbucks.
You can (somewhat) painlessly refine your data to be more accurate by instead using a map radius tool and setting it to your 20-mile drive. This would also better illustrate the urban-rural divide in the Mountain West and how chains enter our cultural zeitgeist - and tend to leave the rural or remote cultures in the dust.
Thoughts?
this is beautiful
ReplyDeleteThank you =)
DeleteThat second chart is interesting, but I don't think it is a voronoi diagram. A voronoi diagram would have the Starbucks points at the center of each cell.
ReplyDeleteThe SB's aren't the center. The center is the point that is furthest away.
DeleteThe vertices are the SB's. To get the furthest point, you take the cetroid of the voronoi polygons
This comment has been removed by a blog administrator.
ReplyDeleteI agree with Caitlin Dempsey (GIS Lounge) that this is one of the year's most interesting maps. Just shared it with my readers here:
ReplyDeletehttps://www.facebook.com/KKHConsulting
Thanks for sharing it, Katherine!
DeleteMay I use your map in a class presentation?
ReplyDeleteMay I use your map in a class presentation?
ReplyDeleteas Jennifer Goodland pointed out, you might be interested in the Graph Voronoi diagram
ReplyDeleteMartin Erwig. The graph Voronoi diagram with applications. Networks, 36(3):156-163, 2000.
Kurt Mehlhorn. A faster approximation algorithm for the Steiner problem in graphs. Information Processing Letters, 27(3):125-128, 1988.
Hello Jim, I fell in your blog investigating about Voronoi tesselation,.. I am trying to do this for a 3d distribution of points, each one represent a galaxy clusters, with the aim of find the local density cluster. However, I achieved get the Voronoi neighbors and calculate the Voronoi vertices using geometry package of R. I am using the volume of the convex hull for find the volume of individual cells. For me is intuitive that the volume of the convex hull of the individual cell is equal to the volume cell itself, because I can not find examples against it in my mind, but, of course, I am not
ReplyDeletereally and completely sure about it assumption. Do you now something about this? Could you help me in this matter?? I will be grateful about it..