Optimize web analytics for return visits, not page views: Josh Schwartz

Sometimes, data-backed best practices create such a flood of usage that its impact is lost quicker than some adopt the act at all — see any number of Facebook posting tips.

Nevertheless, scientists at tech firms are using research methods to find the next new best practice and exploit it early, while also figuring out ways to build a loyal following. These were the sort of topics under inquiry at Wednesday’s Data + Content Variety Show.

Downtown Brooklyn’s NYC Media Lab organized a forum at Bloomberg on Wednesday morning, in which data scientists discussed how science is informing content creation, helping it spread (or at least understanding how it does or doesn’t).

The panelists at the event included Brian Eoff, lead data scientist at Bitly (and a Brooklynite); Ky Harlin, data scientist at BuzzFeed (in August, we wrote about his talk at the Brooklyn Tech Meetup); Mor Naaman, Chief Scientist at Seen.co (a company that started here in the borough); Simon Smith, Senior Vice President of Platforms at Newscorp; Josh Schwartz, lead data scientist at Chartbeat and Lisa Strausfeld, Brooklynite and data visualization designer at Bloomberg. Justin Hendrix, the Executive Director of the NYC Media Lab (we covered his appointment), moderated the morning event.

• Eoff gave a number of examples of how any best practice that gets shared will get imitated by lots of people, pointing out that the first person to tweet something twice was smart, but that inevitably led to certain content makers tweeting the same content dozens of times.
• Harlin explained that Buzzfeed’s analytics go very deep, including changing editorial output. For example, he gave an example of a list with 45 items on it that later shrunk to 27 after watching which of those items seemed to get the strongest response from users.
• Harlin also explained that Buzzfeed has APIs watching broad content trends and feeding information to editors, in order to help them guide assignments.
• Naaman seemed to get an ah-ha from the crowd when he explained the basic idea of Seen. He said that news is often generated by events, and in the Internet age, as events happen, people say things about them online. There is often a decent amount of space between regular people saying things and reporters filing stories. His product aims to fill that divide.
• That divide between a happening and a report is, in Seen.co‘s thinking, an opportunity. So they generate web pages based on top performing content that comes out of events. Such as this one that they made from #datashow hashtag yesterday.
• Smith gave his perspective on analytics, saying that page views are a proxy for ad dollars, but from the perspective of a business with multiple streams of income, all page views are not created equal. A page view from a subscriber or an influential decision maker is worth more to them than one from a drive by reader.
• Smith also gave broad context, saying that one of the things that makes the news business addictive for people in it is the way that all your assumptions about what will work and what will interest people can be completely upended in a moment by something completely unpredictable, like a hurricane, assassination or scandal.
• Schwartz from Chartbeat made the argument that analytics should be geared more toward retaining readers. So, for example, the sources that might give a spike in content probably don’t yield return readers. Where the reader comes from is the strongest indicator of whether they will come back.
• Readers are more likely to return after reading a long post than they are after reading a short one, Schwartz said, citing Chartbeat data.
• Once you have a reader coming by more than three times in a month, you can basically think of them as yours.
• Strausfeld walked through the different kinds of data visualizations, showing that there are both exploratory as well as explanatory visualizations. She also showed how an exploratory visualization (such as Bloomberg billionaires) can be turned into an explanatory one.
• Strausfeld also announced the release of a giant walk through of the housing crisis, “Bubble to Bust to Recovery,” that her team at Bloomberg created.

In a Q&A that followed up, Smith made the comment that if you want to increase readers’ engagement on a post then put a crossword puzzle on it. He was very confident that it worked, that said, it also elegantly illustrated the point that not every method of upping engagement is appropriate.

The event occured on the same day that the NYC Media Lab was able to announce that News Corp had become its 8th corporate member. Its next event is “Mobile Futures” on April 23.

Half life of content on Twitter is just two hours according to bitly’s @bde #datashow

— Adam Ostrow (@adamostrow) February 26, 2014

Chartbeat’s Josh Schwartz: what if we threw out all the other metrics and only focused on audience retention? // INDEED! #datashow

— Indu Chandrasekhar (@indumania) February 26, 2014

NYC Media Lab #datashow is aces. BuzzFeed’s Director of Data Science shared an epidemiology equation he uses to accelerate content shares.

— Mark Mulvey (@MarkMulvey) February 26, 2014

Full house at @nycmedialab‘s #datashow, & an engaging discussion on best practices in data analytics. pic.twitter.com/xFKhHyCQFf

— NYCEDC (@NYCEDC) February 26, 2014