Huffington Post: Crawling with data addicts

The Huffington Post has recently passed the Washington Post in traffic. It got 410 million page views last month (and 35 million on the iPhone alone). Data is a big part of the site’s success.

At the media panel at the Webtrends conference this week, Huffington Post’s chief technology officer, Paul Berry, described some the methods. Every hour, editors see how the traffic that hour compared to the same hour a week ago. The site is a laboratory for so-called A/B testing, where stories are played against each other to see which draws more traffic–and how long each story should stay on the front page.

“…We’ve built a lot of internal tools…” Berry said. …”A lot of us are addicted, like crack addicts, to these stats….”;

HuffPost also tracks readers’ shifting moods by carrying out automated “…sentiment analysis…” on the two million comments the site generates every month. (In other words, machines look for key words and report on whether the comment was favorable or not. If you look at individual results, they’re fairly primitive. A sentence like, “…I’m not saying I’m not crazy about it……”, can throw a machine for a loop. But they get the big …

The
Huffington Post has recently passed the Washington Post in traffic.
It got 410 million page views last month (and 35 million on the iPhone
alone). Data is a big part of the site’s success.

At the media panel at the Webtrends conference this week, Huffington
Post’s chief technology officer, Paul Berry, described some the methods.
Every hour, editors see how the traffic that hour compared to the same
hour a week ago. The site is a laboratory for so-called A/B testing,
where stories are played against each other to see which draws more
traffic–and how long each story should stay on the front page.

“…We’ve built a lot of internal tools…” Berry said. …”A lot of us are
addicted, like crack addicts, to these stats….”;

HuffPost also tracks readers’ shifting moods by carrying out automated
“…sentiment analysis…” on the two million comments the site generates
every month. (In other words, machines look for key words and report on
whether the comment was favorable or not. If you look at individual
results, they’re fairly primitive. A sentence like, “…I’m not saying I’m
not crazy about it……”, can throw a machine for a loop. But they get the
big picture.)

Berry said that many large advertisers are still eager for traditional over-the-fold real estate. But they get more clicks when their ads accompany stories about them. Clicks on Bing ads, “…go through the roof…”, when the story’s about Microsoft, he said. (Traditionally, at least some magazines have worked to separate advertisers from stories about them, but those days are disappearing fast…)

Berry said that in its editorial layout, Huffington follows the, “…Mullet
Strategy:…”, Business in front, party out back.

Link to original post