Travis Swicegood, director of technology at Texas Tribune, spoke this week at the latest Hacks/Hackers Chicago Meet-up about the challenges of working with public data — real world data, as Swicegood calls it.
There are plenty of challenges in collecting, managing and presenting data from a state the size of Texas — 26 million people, 254 counties, five major cities and a gross state economy of $1.2 trillion. Swicegood shared just a few of the challenges of managing and wrangling data:
“Data frequently, frequently, frequently disappears.”
Swicegood described “real world data” as being dirty and unpredictable. Despite those negative attributes, lost data is even worse. He has a copy of every dataset he works with at the Tribune — a habit he formed from after seeing data go missing from local locations as well as government websites. He also distinguished the Texas Tribune’s civic datasets from “big data.” While Tribune datasets may include millions of records, Swicegood noted data collections of, say, social media organizations are much larger and constantly growing.
“Making assumptions about the data is something that can get you in lots of trouble.”
Tribune beat reporters act as domain experts for data and in-depth “explorers.” The dedicated data folks might be best suited to manipulation and vizualization, but a beat reporter can help a developer understand datasets and explain apparent discrepancies. They've also got an ability to understand what makes them really interesting thanks to time on the beat.
“This is why more citizens don't grab big government datasets.”
Swicegood recounted the difficulties of collecting inconsistent school data from the 254 Texas counties for the Tribune's Public Schools Explorer. Not to mention the occasional 700-column wide CSV file and the reports from 141 agencies to compile the Tribune's famous state employee salary database.
Swicegood also shared his data tools and techniques for working with dirty, inconsistent data. He recommended analysts capture their calculations in scripts, document their data munging and use version control so their operations can be replicated when new data is released.
"Our CEO likes to say we're The Boy Who Lived."
The Tribune launched within months of two other non-profit news organizations, Chicago News Cooperative and Bay Citizen. Four years later Texas Tribune is the only one still running as originally envisioned — Chicago News Coop shut down in 2012, and Bay Citizen has partnered and rebranded a few times. Swicegood said the Tribune’s business model and focus on Texas (a state with plenty of wealthy donors) have helped it succeed.
“I like to say we're a technology company that produces a journalism based product.”
Texas Tribune has always been an online only publication. Swicegood’s slowly winning converts to his position … to the chagrin of some in the newsroom, he said.
