Taboola had a banner year in 2012, serving nearly 1.5 billion video recommendations per day by year’s end. With all of its growth in video recommendations, the company also grew revenue by more than 750 percent in 2012. The startup now sees a huge opportunity to move on from recommending videos to tackling content of all types.
- Kaltura: The Israeli Up-And-Comer That Is Revolutionizing Online Video
- Looking For A Recommendation? ‘Snoox’ It!
The new product might seem a little counter-intuitive, considering recommendations for text articles have been around for decades, and there is no shortage of companies trying to tackle that problem. The difference, according to Taboola CEO Adam Singolda, is that it’s gone beyond just providing contextual recommendations to now also bringing in technology that it used to improve its video product and applying it to text articles.
Video recommendations can be a tough nut to crack, especially compared to text. That’s due mainly to the lack of good metadata around video. Without it, Taboola was forced to use other information, like viewer demographics, social sharing, and user behavior, to determine which videos users are most likely to watch.
Applying video recommendation algorithms to text
Now Taboola is applying those algorithms to article text and getting impressive results. Singolda estimates that the company’s technology can yield three times the clickthrough rates and higher CPMs than other article recommendation offerings out there, depending on the type of content that it’s looking at.
Taboola already has some clients using full-text recommendations, including Bloomberg, Businessweek, Meredith and Jerusalem Post. It signed up some as a result of already doing business with them on the video side, but it is also offering full-content recommendations as a standalone service. Like its video recommendations, it allows publishers to connect users with other content on their own sites, while also enabling them to have their content highlighted on related articles on other sites.