• Taboola Aims to Crack Video Recommendations Nut

    For Internet-based content, recommendations have to be one of the toughest nuts to crack. There are just so many variables in play - analyzing and characterizing the initial piece of content consumed, building a large enough database of content to match it against, understanding individual user's peculiarities, presenting results in a meaningful way, etc. Still, effective recommendations are powerful because they enhance the user's experience, increase consumption and drive more ad inventory.

    One company trying to crack the recommendations nut for video content is Israel-based Taboola. I recently caught up with their CEO/founder Adam Singolda to learn more about their approach and progress. The company's ViDiscovery system analyzes both the content provider's video and its users' behaviors. The combination of the two then drives Taboola's recommendations, which can be presented in a number of different formats depending on the content provider's preferences.

    You can see several examples of the recommendations in action. At CNN.com, a little tab in the bottom left of the video window prompts for "Videos Like This" which in turn opens a horizontal scroll bar with recommended videos. In some cases the recommendation work very well, matching specific news stories with one another. But in other cases the experience was mixed. For example, when watching a video about the "Zooz Beat" music app the first recommended video was an update about Hurricane Gustav from August '08 and the second was about a refinery workers' strike in the U.K. Hmm, if there's a correlation between the three videos, I'm not sure I see it.

    Still, Taboola's team has an impressive pedigree and has raised $6M to date, plenty for a small team to continue refining its algorithms and results. The company's primary model is ad-based, with it receiving a revenue share for incremental video viewing that it drives. That kind of success-based approach will endear it to resource-constrained content providers eager to generate additional usage and revenues without extra expense. It's easy to implement Taboola by just adding a line of code to the player or web site.

    High-quality recommendations are not easy to pull off, but if Taboola can get its system really humming with demonstrable case studies of success it could gain very quick traction.

    What do you think? Post a comment now.