Analytics/Big Data

Brits Rev Up Recommendation Engine

A U.K.-based startup has made notable progress with a content recommendation engine that mines metadata and tracks usage patterns to help customers seek out and find content they might actually like.

Considering the almost bewildering array of linear and on-demand content now available to viewers, perhaps it's not surprising that the tool, developed by ThinkAnalytics Ltd. , is gaining traction with pay-TV providers.

The company is trying to solve a freedom of choice paradox (apologies to Devo) faced by consumers -- they like having the breadth and depth of content but still need some help finding the content they might want to watch.

Recommendation engines are already popular among over-the-top (OTT) video providers such as Netflix Inc. (Nasdaq: NFLX). But ThinkAnalytics says its engine goes beyond the "light touch collaborative filtering" that some OTT players use by tapping into metadata and other information gathered not just from VoD usage and live TV channel surfing, but also by factoring in their preferences for non-video content such as books and games. The result of that work creates a "unique preference profile" of the customer, says Chairman Eddie Young.

And that process isn't limited to the set-top box. ThinkAnalytics says its technology can be built into guides driving navigation systems on tablets, smartphones, PCs and other connected devices that have become key to TV Everywhere strategies.

And the idea is gaining traction with pay-TV providers. The U.K.-based company claims that its technology currently is built into navigation systems that are in front of 70 million customers, and counts Sky , Virgin Media Inc. (Nasdaq: VMED), Telenet and Unitymedia GmbH (now part of Liberty Global Inc. (Nasdaq: LBTY)) among its major customers.

It's also working with three unnamed service providers in North America, including two "very large cable companies" that are just about to deploy the technology, as well as a smaller IPTV operator, Young says.

It's maybe worth noting at this point that Comcast Corp. (Nasdaq: CMCSA, CMCSK)'s IP-capable X1 platform, which is set to launch soon in Boston, will sport a recommendation engine.

Why this matters
As cable continues to expand its content libraries and TV Everywhere efforts, the use of slick, cloud-based navigation systems probably aren't enough to help video customers discover content without going through a laborious process.

The use of recommendation engines will give pay-TV operators access to the sort of navigation tools that have helped to popularize OTT video services and should help to improve customer satisfaction levels.

And tying a recommendation engine to these new interfaces could help cable operators improve VoD sales levels well beyond what they've been able to generate from older, more limited menu-driven guides that require users to search for content on their own.

For more

— Jeff Baumgartner, Site Editor, Light Reading Cable

COMMENTS Add Comment
Jeff Baumgartner 12/5/2012 | 5:28:59 PM
re: Brits Rev Up Recommendation Engine

I've found no problems when I know what I'm looking for (ie. the latest episode of Game of Thrones) , but cable Vod's not really set up well for browsing to the point that i never start up that menu just to poke around.  but suppose recommendations and a little smartening up of the service wouldn't hurt.  JB

shygye75 12/5/2012 | 5:28:59 PM
re: Brits Rev Up Recommendation Engine

There's a reason it's called the "idiot box" by some.

AESerm 12/5/2012 | 5:28:58 PM
re: Brits Rev Up Recommendation Engine

Interesting to hear Think Analytics describe collaborative filtering as so-last-year, or as a "good but early" step, as I reported here. Used to be that Netflix or Amazon was the model. Could this be a leap-frog moment? Granted, with so many idiots both watching and appearing on the screen, these algorithms could be the smartest game in town. 

shygye75 12/5/2012 | 5:28:57 PM
re: Brits Rev Up Recommendation Engine

It looks to be another step in the direction of, "Let's give people what we can" rather than "Let's give people what they want."

shygye75 12/5/2012 | 5:28:56 PM
re: Brits Rev Up Recommendation Engine

"We notice you've watched three episodes of Jersey Shore. Based on your viewing habits, we recommend that you walk seven blocks south to the Liquorama outlet, buy the gallon-sized bottle of Grey Goose, walk seven blocks north back to your home, and drink the contents of the bottle you purchased." Welcome to Nirvana.

craigleddy 12/5/2012 | 5:28:56 PM
re: Brits Rev Up Recommendation Engine

Much depends on how well the app is designed and how well it works. I think a well-done recommendation engine (as well as better search tools and bookmarking) could be helpful to find something to watch amid the vast wasteland of TV. Such tools could be an asset for on-demand recording and viewing, sampling new shows and movies, and tracking your favorite actors and celebrities. 

Think of it this way, Dennis: You could set it up so you would never miss Lady Gaga on TV again. Doesn't that sound appealing?...Well, OK, maybe not the best example. 



Nikki Ralston 12/5/2012 | 5:28:49 PM
re: Brits Rev Up Recommendation Engine

Collaborative Filtering is more like 'so 10 years ago' for entertainment discovery. When Amazon first started using it, it was the best thing available. CF is still one of the best techniques to use when there is no meaningful information available about the items being recommended. It is essentially statistical 'blind guessing'. So engines that know absolutely nothing about the content beyond generic metadata are still using this, because  they don't have anything better.

TA claims they use 'systematic metadata analysis'. Not sure what this means. Metadata is superficial and contains no nuanced details about the plot, mood, style, etc. So this is a classic 'garbage in, garbage out' situation. Discovery that relies on generic underlying metadata is never going to deliver much beyond the metadata.

You and I may both like comedies, but you prefer offbeat, cynical, witty comedies about dysfunctional families and I like sentimental, gross out humor about teenage life and obsessive quests.  Metadata, no matter how you analyzeit, will never understand tastes on this level.

Nikki Ralston 12/5/2012 | 5:28:48 PM
re: Brits Rev Up Recommendation Engine

VOD is hugely undermonetized because the discovery is so bad (It's often like browsing the shelves of a library organized by the Dewey decimal system.

Making it a little better will increase profits a little, but VOD has huge revenue potential- HUGE!

Discovery is the dynamite that will blow this mine open.  

shygye75 12/5/2012 | 5:28:48 PM
re: Brits Rev Up Recommendation Engine

Hi, Nikki -- Rather than mess with collaborative filtering, I think we've hit on a better way to make this stuff work: Train users to simply follow the recommendations of their favorite devices. It won't work with everyone, but I think we can reach critical mass with this strategy. We've had a few generations of prep work, courtesy of mass-market advertising.


Sign In