Four visions of data journalism, moderated by Kathryn Corrick, digital media consultant.
Bella Hurrell, specials editor on the BBC News website
The BBC specials team produces a whole range of added-value content for the BBC website. They’re becoming part of a visual journalism team at the BBC. Data journalism can be long slow projects, but not all of them. You should pick subjects that have a shelf life – road traffic accidents, unemployment, that sort of thing, Update the data and people will keep coming back. Make tools they will want to keep using. Build sharing into it.
They did a project plotting road traffic fatalities through FOI requests – the map was the most popular element, because it allowed people to understand what the situation is where they lived. They also visualised some of the most interesting data – for example, bikers are 21% of fatalities, compare to 1% of traffic. They liveblogged every accident on one day to bring publicity to it, to help amplify the data and give it more life. It was really popular and followed nationally. Their military deaths in Afghanistan gets traffic every time there’s a new death. Their unemployment tracker gets updated monthly and gets s ready stream of traffic.
Visualisation helped bring a dry subject like the Eurozone debt web to life. They had comments open, and responded to the issues raised. “People really appreciated it”. They key seems to be a double-whammy of personally applicable information that is also globally relevant.
Claire Miller, senior reporter and data journalist, Media Wales
From global to very, very local… The bread and butter of what Media Wales do is government data. The focus is still stories for the paper, so they’re reacting to what is released by government, and finding stories in that. Beyond the day to day, it’s a lot of FOI data used to create stories. With FOI you can get the data you want at the level you want. For example, she asked exactly when and where all the parking tickets where handed out in Wales. They visualised it using Tableau. A&E visits by location, not surprisingly, increased nearer the hospital you lived, and that showed up well on a map. Mapping empty homes allows quick identification of hotspots.
With open data, more and more stuff is being published, so there’s lots of potential.
People look for specific things, as uses of the data store show – local elections, the Olympic torch relay and sport. Education bubbles along all the time… They ended up making their own Olympic Torch map, because they couldn’t embed the official one easily. It went crazy. It was the most popular thing on the site. Wales lacks the same easy access information on school performance in the UK. Media Wales gathered everything they could find, gathered it into one app, and let people access it.
And anything with rugby in it is popular…
Damian Kimmelman, CEO, Duedil
Everything that consumes electricity will inevitably be connected to the internet. And that means it will leave a data trail. And they are, in his words, “data whores”.
Duedil is a site for examine the state of companies using available data (Martin did an excellent write-up of their Hacks/Hackers talk). They’ve had acquisition offers for £20m – and they’ve turned it down. They’re still seed funded. And they’re still finding new ways of making data more interesting and useful. They’re planning on launching a facility called lists. Create your lists of types of companies, and use the data to find new ways of tagging, categorising and analysing the companies. But they need more information. Mapping the companies around your social graph – will that show whose companies have changed dramatically over the last few years?
Heirecrchy of needs for data: it needs to be clean – deduced and usable. It needs to be findable – and linked.
Users need to know the provenance of data – who touched it, who keyed it in? Did the accountant make a mistake? The more people touch data, the more imperfect it. It’s important to understand the authority of a dataset.
James Ball, data journalist working for the Guardian investigations team
He’s a reporter, dammit Jim, not a designer. Whatever you’re trying to do – there’s a dataset you can buy, open, assemble or FoI. But that’s a bit like saying there’s someone who knows the key to your story – how do you find them? He wants to challenge the idea of “from data to story”.
There’s all sorts of caveats when you’re using data from surveys and censuses. Investigating the stat used as the basis of a Diane Abbot comment piece lead to exposure of a biger story – one disproportionate rise in young black male unemployment – which hit the front page. Sometimes readers will simulate investigations – people claiming in comment threads that the vacancies claimed in jobs centres are not real, or zero hours contracts, or the like. They tried to scrape the relevant data from a government site – but it had protections in place. So he had to FOI it – and got the data, albeit heavily encrypted (but the phoned the password over). It was very messy, inconsistent data.
This wasn’t a story from digging around in data. This was questions from the readers and a comment piece which they could answer with data. Do you ring around your sources and ask them for a story? If you’re doing that, you’re doing something wrong. Don’t do that to data, either. Talk to humans, look at news, and then ask the right questions of your data. This argues for not having data journalisms in silos. Don’t just keep them in offices looking at spreadsheets…