Taboola, a service that powers many of the video recommendation widgets you find across the web, has raised a $9 million Series B funding round. The round is being led by Crescent Point, which is based in New York and Asia, as well as Evergreen VC, which has participated in previous rounds.

This round brings Taboola’s total funding to $15 million (it previously raised a $4.5 million round in November 2008, and a $1.5 million round in November 2007).

Taboola gives publishers a turn-key product for offering recommendations on whatever video platform they’re using (it’s integrated with Brightcove, Ooyala, and other services). This is actually easier said than done — surfacing related videos is hard enough, and then there’s the matter of gauging whether they’re videos people will actually want to watch. That’s Taboola’s speciality, and it’s landed plenty of notable clients, including the New York Times, USA Today, CNN, and Bloomberg.

The service does extensive analytics on the performance of each video, generating a score for each called ‘EngageRank’  (it also tries to tailor the recommendations to each user).  And that often leads to more clicks — the company says it drives as much as 2.5 times as much video watching on Bloomberg.

Alongside the funding, the company is announcing a new milestone: the service now generates 250 million recommendations per day across its publisher partner sites.

Video recommendation startup Taboola has been having quite a year, signing up major Web publishers such as the Huffington Post and the online editions of Bloomberg and The New York Times. The driving force behind the company’s continued traction is their Text2Video product.

In fact, its most recent coup, Revision3, has seen a whopping 90% video uplift since adding Text2Video to their non-video pages.

Adam Singolda, Taboola’s CEO, explained to me that while Taboola’s initial product was purely Video2Video recommendation, the company realized that although video CPMs are higher and publishers want more video views, most Web traffic is still mostly textual. Put simply, folks on the Web read much more than they watch video.

To use Text2Video, publishers must also add the Video2Video module, allowing Taboola to crawl the publisher’s video inventory for title, thumbnail, description, etc. Then, the Text2Video module analyzes page content to offer video recommendations. Revision3 saw a 90% uplift in video views when Text2Video was added to its non-video pages, such as DiggReel. See it in action on Bloomberg.com as well, here.

Singolda went on to explain that Text2Video’s ‘secret sauce’ is its ability to measure user engagement after the click on a video recommendation. This means that unlike purely contextual/semantic analysis solutions that can tell that page content happens to be about the iPhone, and consequently ‘pull’ iPhone videos, Taboola can tell that for iPhone page content, only short videos keep users on the site. Taboola can also determine that for iPhone page content, users are likely to be more engaged with off-topic videos (travel, for example).

It seems the benefits of increased engagement and additional inventory publishers can monetize at higher CPMs are not falling on deaf ears. Taboola tells me that it’s now running on hundreds of millions of pages a month, doubling traffic every couple of months. Perfect timing for the video advertising ‘Frenzy Point‘ we recently wrote about.

In general updates, since we last covered Taboola, the company has moved its headquarters from Israel to NYC. It has also brought on Lior Golan to run Product/R&D. Golan co-founded Cyota, which was sold to RSA for $145M. Liz Hughes also joined the company as VP Business Development. She previously held the same position at RealNetworks.

Taboola has kept a low profile since its founding in 2006. I am told that the gross majority of the employees are technologists with strong mathematical and Israeli military backgrounds.

Taboola’s flagship product ViDiscovery is described as follows on the company’s site:

Taboola ViDiscovery™ moves beyond “one-size fits all” answers and matches videos one viewer at a time. Utilizing it’s breakthrough patent pending mathematical technologies that was built from the grounds up to deal with online videos, Taboola predicts what are the best subsequent videos for each viewer to watch on your site . Taboola’s proprietary patent pending technology works in three steps:

1. Video Context – Analysis of video context independent of any associated text.
2. Viewer Dynamics – Anonymously studying the viewing patterns of viewers while they view videos on your site.
3. Personalized Matching – Using the aggregated knowledge from the analysis of your videos and your viewers’ viewing pattern to match every viewer with personalized video recommendations.

To see it in action on CNN.com, go here and press the “Videos Like This Tab” on the video window.

If there indeed a deal with CNN it would serve as a major indicator that the company’s video recommendation/discovery technology packs a real punch. CNN would then join the company’s growing client list which includes the likes of aniBoom, 5Min and Sclipo.

We’ll keep a close eye on Taboola and update on any official funding announcement.

Taboola’s made integrating their system a breeze. All it takes is a single javascript tag added to your player’s code for them to start recommending videos in your site’s player (or ads, as seen right). You can currently see their service live on 5min. Taboola says that 5min has seen a 23% increase in watched videos and 53% increase in session length.

Their engine makes recommendations based on a rather complex algorithm that combines multiple methods into a single “Match Rank” — collaborative filtering (user ratings/reviews), a viewer’s history, and analyzing the video’s meta data. Their claim is that each one alone is prone to problem, but together can cover for each other’s short comings. The system even takes into account small details, like the length of a video or how much of the video people actually watched.

It’s a pretty beefy algorithm under the hood. However, the general idea is that their system sorts through your video content, then recommends them to users based on what they and users like them have watched in the past. If you want to get into the details, you’ll have to request a demo. However, it looks like a good solution for websites not looking to reinvent the wheel, but instead focus on their core product.

I see the launch of services like Taboola as a sign that the online video space is really maturing. We’re past the days of video sites having to hack away at all their problems alone. New services are popping up to meet the individual needs of publishers looking to make the transition online more easily. There are services for serving video advertising, trans-coding, and even video hosting.