“Ediscope” has a whiff of the Alexander Graham Bells about it. The idea was to create a system, with a means of automatically spotting new articles published on news sites, of finding out how much social activity had happened around those articles and in some cases, of scraping page view counts. It was supposed to be available for public use and that would have been grand. It’s a pity Ediscope was doomed from its inception.
Lifshits, working at Yahoo! Labs, did create the tools and conduct one study (The Like Log study) of 45 news web sites. He gathered information over three months 2010-2011 from more than 100,000 articles. The results are still available online, and they are interesting.
What makes the Ediscope research valuable is that it is the first quantatitive examination of social signals (or SOGs) by article across news sites. SOGs are Facebook likes, Tweet counts, Google +1s and any other cumulative measure of social network activity associated with a piece of content. The most important SOG at the moment for mainstream news sites is Facebook likes (that could change rapidly in the face of Google+).
Lifshits’ paper describing Ediscope is flawed. His biggest mistake is assuming the SOG count is directly linked to reader satisfaction or value. The data-driven approach led him astray. Value is not simple – it has private and public components – and private value, the kind that a reader won’t own up to, is not reflected in social networks. On the other side, “liking” something publically is sometimes done for reasons divorced from personal value or satisfaction.
Lifshits found what he called a “surprisingly low” correlation between page views and SOGs. In other words, you don’t learn much about the page view count by knowing how many times an article has been liked. The SOG/PV ratio actually goes down the more popular a story is – on average, the more page views a story has, the lower its social count ratio. Lifshits assumed this exposed a flaw in news site production processes that could be corrected.
This is wrong. Mainstream news orgs push massive streams of traffic to articles through exposure on homepages. Lifshits is assuming that the editors have made a mistake – have exposed the wrong story – if the SOG count doesn’t reflect the PV count. But what people “like” publically is only a subset of what they will click on. There is a bunch of stuff that falls into the “that’s not me but I still want to read about it” category. The dirty little secrets.
But Lifshits’ main point, that news businesses can increase audience size and engagement by researching social signals counts on content types, is spot on. As social networks become an increasingly important distribution channel, it makes a lot of sense to know what type of content is going to work in that channel. Homepages remain centrally important. They give you character, voice, and seed the stories that get shared. There’s no getting away from them. But they are vulnerable to behavioural change in the audience and technological change in the platform.
Every news site I can think of would rather have a “balanced portfolio” where a good chunk of traffic came from search and an even bigger chunk came from social networks. It is safer as a content provider to have your stategies based around entries direct to articles, or the unit of content, because it’s a more fundamental unit than index pages or anything else. The same can go for commercial stategies (sell the article page, not the section page).
Apart from Lifshits’s main insight – a demonstration of how to study sharing across publishers – the Ediscope project also turned up a fascinating list of the most-shared articles over that three-month period. It is eclectic. Quirky articles sit cheek-by-jowl with serious, hard news. The most shared story was the Wall Street Journal’s “Why Chinese mothers are superior”, with a whopping 342k Facebook likes. It’s good to get the emotions involved if you want people to share – that’s a theme you’ll see a lot more of in Share Wars.
And if Lifshits needed any convincing that social signals don’t always equate to readers valuing content, he can find proof in number 5 on his own list of top shared stories – a story about a minor track-and-field event published by Reuters that was shared more than 150,000 times. Headline: “Tired Gay succumbs to Dix in 200m”. Lol.
So why was Ediscope doomed? There was no way something as valuable as a suite of social analytical tools was going to remain free, not when there are bucketloads of money to be made collecting data, analysing content and advising media companies. Lifshits has started a private project and the Ediscope tools are no longer available.
We will interview Lifshits as soon as we can pry him away from his new project to find out what he is up to.