Gravity Personalizes News Feeds For Publishers


With screens becoming smaller and the amount of news on the Web becoming more vast, “personalization” is becoming the buzz word of the year as publishers try to find ways to make their content more relevant to readers.  Many publishers are turning to tools like Gravity Personalization, which recommends stories based on past articles you’ve read as well as topics and stories that are important right now.

Gravity, not to be confused with the recent Sandra Bullock movie, uses interest-based graphs to track what a user is reading.  As a result, publishers are able to personalize their site to offer users a better of selection of relevant stories.

“[Publishers] have this basic geometric problem, they need to produce more content to get more involved, but they have a smaller and smaller area to come up with that content and engage peoples’ interest,” said Robert Leon, vice president of sales.  “So this notion of personalization has become increasingly important for publishers because of the stakes of building audiences.”

Unlike other programs on the market, like Taboola and Outbrain, which use collaborative filtering models that “recommend” stories to readers based on their previous clicks, Gravity uses artificial intelligence and refined technologies to map an interest graph on each individual, making their personalized content more relevant.  As users click on more stories, Gravity actually becomes smarter at recommending content.

“There’s a difference between personalization and doing recommendations- the difference is that collaborative filtering models rely on how many people clicked on that article, that’s how their system knows,” Leon said.  “If you put an article in front of it, it doesn’t know what that article is about, they don’t use AI or machine learning, they are just using patterns and pattern matching, but not actually personalizing the individual interest.”

Personalization, while it may seem straightforward, is not.  People’s activity on the internet tends to be difficult to analyze because, while people do read a lot of articles on the Web, they don’t give off that many “explicit signals,” such as sharing an article.  Less explicit signals make it harder for programs to understand what an individual’s interests are.

“Building interest graphs on individuals is a lot more difficult than it may seem,” Leon said.  “It’s really hard to know off of a couple of articles someone reads what they their truly interested in unless you have a really refined technology and layers them on appropriately.”

Gravity, which fully launched in February and opened up its suite of APIs to publishers, currently has almost 50 publishers using its technology, including TechCrunch and CNN Money.  The company, which was started three years ago by a group of former Myspace executives, has raised more than $20.6 million in funding.

“If Gravity is successful, then people will find more stuff that they love online,” Leon said.  “And that, I think, would be something where we would all be happy as a company.”