Boxfish aims to change TV with really big data
Boxfish has been processing, in real time, every single word spoken on 1,000 different TV channels worldwide, compiling a database of more than 86 million(!) topics that can be used to categorize and sort all broadcast content. The database is so vast and so versatile the company can barely figure out what aspect of the video industry revolutionize first. Program guides? Audience analysis? Targeted advertising?
“We’re talking to everybody,” said Marios Stylianou, Boxfish’s vice president of marketing & product. “My gut feeling is that people who distribute content are the ones that will be interested first.”
That is turning out to mean search, discovery, and recommendation. Boxfish has no intention of creating specific commercial products, including guides or recommendation engines. The company intends to sell its analytics.
That said, the company has a tool, called Personal Feed, that is essentially a user interface. Created specifically to demonstrate the capabilities Boxfish has, Personal Feed enables viewers interact with a video platform that takes advantage of Boxfish’s analytics.
Using a UI similar to Personal Feed, any company could enable its viewers to search not just for specific shows, or even types of shows, but the entire universe of available content by topic.
It’s the difference between looking specifically for “Bonanza,” versus the ability to look for (or recommending) westerns in general, versus the ability to find every single relevant show or clip, including TV serials, films, documentaries (e.g. a doc on the Pony Express on the History Channel), clips from news programs (e.g. a piece on a new archeological find concerning The Alamo), plus anything else that might be associated with the subject (e.g., Johnny Carson’s interview with John Wayne).
“EPGs? They’re now archaic,” said Stylianou. “This will change the way people discover and engage with content.”
"Our vision is to make television content relevant, accessible and personal" said Eoin Dowling, CEO and co-founder of Boxfish. "Our patented contextual feed, can be used by MVPD's to enable users follow topics they are interested in, and in the background create a personal channel based on this action. This channel can be consumed at a time shifted fashion across any device"
Now imagine collecting data on what people search for. The level of specificity could be a bonanza.
It could change the game for ratings companies. “You look at news shows, they might do 10 or 12 pieces. Which ones did people tune in for? Neilsen doesn’t know. Rentrak can’t tell you,” Stylianou said. “They don’t know what drove viewership. But if you overlay it with our data, you can see what drove them to tune in, what sent them to drop off. Ratings companies can benefit from this.”
The data can be mined to determine, in real time, what topics trending. That might be of use to programmers devising entertainment content; it could help news programs decide what subjects to cover. (The company has an interactive table on the bottom of this page on what’s trending that is hard to stump.)
You could do an on-screen text crawl of what topics are trending, and let viewers link from there if they’re interested in any of those topics, Stylianou said.
Now think of what that mean for advertising. Someone who watches ESPN might be a candidate for an ad for general sports paraphernalia. Someone who has built their own Pacers channel might be a candidate for an ad for a Roy Hibbert jersey.
Stylianou said the company processes 84 billion words a month, and has identified 86 million unique topics of discussion. The system can cross-reference with topics trending in other media, including Twitter and web sites such as IMDB or Wikipedia.
Stylianou said a system making use of Boxfish’s big data capabilities could be local or cloud-based.
The company said it has raised $7 million from Atlantic Bridge and Samsung, and that it is currently beta testing guides with recommendations with with two Tier 1 video programming distributors that Stylianou said the company is not yet at liberty to identify.