Twitter's full data stream—their "firehose"—is a very detailed thing. Access to raw tweet upon raw tweet lets brands know what customers think and allows first responders to instantly tabulate hurricane damage. The firehose is also full of metadata which discloses personal, geographic, and technological information on Twitter's tens of millions of users. Gnip, one of the best known Twitter firehose resellers, just turned a raw sample of metadata from 280 million tweets into an amazing example of data visualization.
The fully scalable and searchable visualizations, created by Eric Fischer and MapBox for Gnip, uses metadata from 280 million tweets collected from a data sample going back to 2011. Gnip's Ian Cairns told Fast Company in a phone conversation the sample was pruned to remove multiple tweets from the same geographic location in order to emphasize geographic distribution rather than tweet frequency. Gnip and MapBox only selected tweets with location metadata attached, which ranged from 2% to 4% of the total tweets in Twitter's firehose. When posting messages to Twitter, users can choose whether to embed geographic location metadata. According to Cairns, the percentage of tweets with location metadata attached is decreasing over time.
One visualization of Twitter use by locals and tourists uses Twitter metadata to break down the percentage of residents and tourists in a specific geographic area. "Tourists" are defined as Twitter users who sent tweets from a specific city for less than a month. In New York City's Financial District, Fast Company's home grounds, most tweets come from locals—except around the World Trade Center site, the Statue of Liberty/Ellis Island Ferries, and the Staten Island Ferry where tourists all predominate. Further north in the Bronx, the location info from Twitter metadata is clear enough to make out Yankee Stadium.
A second visualization shows language use on Twitter. Metadata from Twitter messages also tells the language the tweet was sent in. In the United Kingdom, most tweets are in English...except in London, where a cacophony of foreign-language tweets dominate the visualization. Across the channel in France, a blip of color shows up from the tourists and non-Francophone migrants in the seaside city of Calais.
Similarly, in bilingual Montreal, the visualization gives an instant snapshot of the city's demographics. Montreal's downtown is Francophone with visible English-speaking pockets surrounding Mont Royal park, and the city's suburbs are neatly segmented in English- and French-speaking segments.
The last visualization illustrates what mobile devices are used to post to Twitter. Separate colors are coded to BlackBerry, Android, and iPhone. In Toronto, for instance, there is a BlackBerry-using urban core in the city's financial quarters surrounded by a iPhone-loving periphery.
Different countries also prefer different technology. In the Middle East, iPhones dominate in Israel. To the north, iPhones are also preferred into the Lebanese capital of Beirut... but Androids are more common in the hinterlands. Meanwhile, Jordan is equally split between phone formats while BlackBerry has a stronghold in Egypt and Android is most common in the Gaza Strip.
There were surprises to be found in the map also. Cairn noted that he was taken aback by the relatively small number of BlackBerry users in D.C. compared to the iPhone and Android. Could the federal government's march toward non-BlackBerry BYOD be having an effect?
But for this observer, the most interesting takeaway is the way phone ownership is neatly tied to socio-economic class in the United States. In the city of Philadelphia, the phone OS visualization paralleled the city's class divisions to a T. Prosperous Center City is a mix of BlackBerry and iPhone users. Economically struggling North Philadelphia and the neighboring city of Camden are solid Android territory. The posh suburbs of the Main Line are dominated by iPhone owners, while more working class suburbs such as Norristown and Upper Darby are Android strongholds.
So why do these amazing visualizations? Gnip's stock-in-trade is offering Twitter's firehose and resulting analytics to enterprise customers. As Cairns put it, "We're comfortable with our scope and the business data we offer—this is just us showing off."