Is Personalization The Key To Video Search, Or Is Search The Key To Video Personalization?
Try as I might, I just can’t shake the video bug. It’s gotten so bad that I refuse to take the juicy GoogleClick bait for this column, and will instead continue down my video wormhole. There’s just so much going on in the video space and so much more to come. And, as usual, search lies at the intersection of today and tomorrow.
In my last column, I mulled over the difference between how people consume text-based content vs. video content. Inspired by TV Guide’s new Online Video Guide, I arrived at the conclusion that people generally search for text -- but discover video.
This week, it was a demo from ClipBlast that made me realize that, with discovery playing a lead role in video consumption, personalization becomes crucial. To draw and retain today’s short-attention-span eyeballs, content providers must give each consumer the opportunity to discover video that is likely to appeal specifically to him/her.
To be sure, personalization in the video space is certainly nothing new. It’s essentially the model cable TV was built on, with networks creating and clustering programming based on targeted audience demographic profiles.
The power of interactivity has taken personalization one step further, as evidenced by TiVo suggestions and Netflix recommendations. However, it’s easier to map to personal preference when dealing with TV shows or movies, because of their long-form content.
It’s much more difficult to build a personalization engine around the short video clips that are so popular online -- especially when dealing with UGC. For one, it can be very challenging to categorize something like a 3-minute clip of the Urban Ninja. And, second, people are much more likely to watch short-form video on a whim, as it’s not a big time commitment. But just because I watched a clip doesn’t mean it interested me.
This is why video providers must allow people to take control of their personalization. One way to do this is through user ratings. But a simple thumb’s up/thumb’s down ala TiVo is not enough. And even Netflix’ comprehensive ratings system -- which includes specifying favorite genres and movies -- doesn’t quite cut it.
The key is to overlay search activity. As John Battlelle posits, intent trumps content. Merely knowing that I watched a video tells you very little about me. Knowing whether or not I liked it starts to give you some insight. But knowing what I was looking for when I decided to watch it allows you to close the loop and truly understand my preferences.
Think of how much more powerful it is to know that my watching the Urban Ninja video followed a query for “funny action video.” Couple that with my 5-star rating and designation of the clip as a “favorite” and, the next time I search for “funny action videos,” you’ll have a much better idea of what I’m hoping to discover.
Enter ClipBlast. ClipBlast’s stated mission is “to organize and make the video web relevant, fast, and simple to navigate.” Sound familiar? To achieve this, ClipBlast has created a comprehensive video “search and navigation” tool. And by allowing users to set preferences, it combines past viewing history with self-selected interests to deliver a set of personalized results that update in real time as new video is published on the Web. Now the kicker -- in addition to “my categories” and “my providers” settings, ClipBlast includes “my searches,” so that actual queries are incorporated.
As with any emerging digital media technology, the key to continued innovation is monetization. And, when it comes to video personalization, another critical element is scale. ClipBlast is counting on the strong demand for point-of-query advertising opportunities to deliver sustainable revenue. ClipBlast allows advertisers to target specific categories, providers, or search queries.
And to achieve scale, ClipBlast crawls the Web and accepts video RSS feeds so it can amass large volumes and diversity of video content -- it even indexes Netflix videos. Why is this so important? In order to effectively create a personalization engine, users must be given the ability to choose from a wide array of video assets, including professional and user-generated, as well as short and long-form. Otherwise, there aren’t enough variables in play to drill down and make accurate predictions.
The size of the video index is just one side of the scale coin, though. User adoption is also critical. A large user base is not only important for learning and perfecting a personalization algorithm, but for generating advertiser interest. To that end, ClipBlast is distributing a widget that webmasters can embed on their sites to allow visitors to easily access its index. Today, ClipBlast averages only 10-15,000 page/video views per day, but it’s growing quickly.
This, of course, is where GoogleClick comes in (like a catfish around a vanilla-soaked night crawler, I just couldn’t resist taking the bait.) As the biggest online ad-serving company in the world, DoubleClick has tremendous scale. And its Motif rich media product has evolved through development and acquisition into a leading online video serving platform.
Marry the back-end analytics that DART provides to measure post-click interaction with the front-end query activity that Google captures, and all the pieces are in place to map intent to content. And given Google’s innovation in the personalization space (I may be a dork, but watching the sun rise and set on my customized Google homepage is pretty cool), it’s positioned well to capitalize on the emergence of online video, not to mention make its investment in YouTube finally pay off.
Now, if only Google can get those pesky networks to play nice and share their content, it will be primed to really blow this video thing out of the water. Otherwise, it may need to move to the ClipBlast/blinkx model and focus on simply indexing content instead of hosting it (which, ironically, is the model that Google was founded on and earned it the “fri” in the frienemy moniker). GoogTube reverting back a pure-play search engine? Wouldn’t that be quite the bait and switch.