The homepage of my Trove.com account.
Describing the personalized aggregation service Trove as a “next generation” news experience, The Washington Post’s Vijay Ravindran figures it probably won’t save journalism on its own, but it’s a start.
The site, which will aggregate and personalize news from among 10,000 online sources, launches into public beta next month. It will be free at Trove.com, and on mobile apps for the iPad, iPhone and Android devices.
The effort drops the Post into the middle of a crowded field chasing the elusive goal of news personalization. Several — LiveStand from Yahoo!, News.Me, originally a New York Times creation, and Ongo, a project the Post itself is a partner in — have been announced just within the past month.
Trove, with a reported $5 to $10 million investment by the Washington Post Co., also faces a challenge from tablet-only competitors such as Flipboard and Pulse, and older Web-based services ranging from Google News to My Yahoo!.
Ravindran, senior vice president and chief digital officer at the Post, and leader of the WaPo Labs, acknowledges the challenges of getting consumers interested and keeping them engaged with customized news sites. He said he invested a couple of hours personalizing his My Yahoo! page in 1998, and 13 years later it still “has the same background.”
Two goals for the project
But, he said, the public launch of the site will be the start, not the end of Trove’s development. And, he stressed, the project has two long-term objectives: revenue and innovation.
Trove's channel pages provide more in-depth coverage of a given topic.
“We hope Trove.com is eventually a great commercial success, ” he said, creating a great consumer experience, and a great business. But as importantly, the project is aimed at helping the entire company “evolve rapidly to what that next generation news experience is going to be.”
Ravindran said that evolution will be affected by taking the tools, technology and experiences learned from Trove and adapting them to WashingtonPost.com and Slate.com. That, as much as the success of the personalization site itself, is the reason for the Post’s interest in the project.
Ravindran would not talk numbers but said WaPo Labs, and its projects, were a long-term investment for the company. That investment was redoubled with the purchase of news aggregator iCurrent last summer, reportedly for between $3 and $5 million.
It is the technology and team from iCurrent, now part of WaPo Labs, which powers Trove. Its news personalization begins with sophisticated text analysis techniques to identify content topics and categorize stories automatically.
The Trove story picker lets users choose between a series of of stories to help the site learn their news preferences.
The site calculates reader interest based on feedback gathered implicitly through observed behaviors and explicitly through a topic selection tool. Articles are also further organized by an algorithm that accounts for relevancy and reader engagement with a topic, among other factors.
I had been a casual user of iCurrent prior to its purchase by the Post Co., and tested Trove in private beta for a few days last week before speaking with Ravindran.
Separating noise from signal in algorithmic preferences
At this point it is too early to judge the effectiveness of Trove’s algorithms. My current Trove homepage represents its best guess at my interests based on 71 “channels” I voted up. That process, which feels similar to ranking genres and movies on Netflix.com, lets you pick between two specific articles — informing the system if you like one or the other, both, or neither.
And of the 75 or so stories it currently proposes, most are of interest to me, but only a few are of urgent “read this now” interest. Based on the stories I have expressed a preference for so far, my channels include The NFL, Social Media, Mobile Apps, AOL and Twitter. So far, so good. But, it is also tracking for me Tunisia, Solar Energy, Orbitz and Heart Disease — topics which may interest me based on a specific story, but not necessarily for the broader subject.
And, because I did self-select most of the topic-level channels, it is likely I am missing out on interesting articles that fall outside my preference settings.
That is an issue Ravindran’s team understands it must confront.
Ravindran suggested a few approaches might help resolve that question as the site continues to develop. First, Trove already highlights a selection of “editor’s picks” in the left-hand rail of the homepage. He said that section is meant to highlight top news that may not appear in a reader’s channel selections.
Also central to the strategy, Ravindran is looking toward the ability to supplement computer-selected stories with individual suggestions proposed by expert editors. As an example, he noted, a reader interested in the Republican presidential primary race might see a daily collection of articles curated by The Washington Post’s political writer Chris Cillizza.
He sees the eventual best solution laying somewhere along a continuum between computer-generated algorithm, and human expert editors.
A former executive at Amazon.com, he told me recommending news is decidedly more challenging than books.
“A purchase is a very high quality signal of interest,” he said referring to the act of buying book from Amazon. “You can’t compare that to a click-through on a headline.”
Additionally, most retail products have a long shelf life he said. “Therefore a recommendation of a book published four years ago can still be highly relevant.” Conversely, a news story published four days ago is likely irrelevant.
Personalization needs a social element
But part of the challenge of personalization seems to lie beyond simply recognizing stories that are of recent general interest. Using Trove, I realized that much of my news does “find me.” I rely on Twitter, Facebook and other social media channels to alert me to stories that are “trending” among my friends.
Without some of that “social signal” in Trove, the site feels like an interesting weekly magazine, but not like an active, timely news source.
Ravindran noted that in the private beta some features — such as social integration — were not yet activated. Over time, he expects topic channels to become very social, especially if the site begins to capture data on which articles are shared among your network of friends.
Users can adjust their channel preferences with a fine tuning tool.
He said that when it comes to news, what your friends find interesting is a “high value indicator” of potential relevance to you. “Social is the most important phenomenon of this decade,” he said. “Being able to develop apps that are social at their core is important.”
“We are in an exciting time right now in the online news industry,” he said, with a lot of experimentation taking place. What everyone can agree on, he argues, is that the current news experience, “needs to change fairly radically.”
Trove, The Daily, Flipboard and Pulse, “each represent someone with a passionate thesis” about the future of news, he says.
Trove.com will launch as a public beta product in a few weeks Ravindran said, and it includes some “very important” elements of what that future might look like. But, it is all part of the learning process of figuring out “what the other elements are, so we can develop a better consumer experience.”
Comments