Video Search Coming Of Age (Includes Video)

Here's one simple ingredient for monetizing video: Get lots of video traffic. Here's one simple ingredient for driving video traffic: Make the video great. Here's one simple ingredient for great video: Create content that matters. Here are three simple ingredients for creating content that matters: Make it entertaining; make it practical; make it easy to find. Blend rigorously, optimize video search engines, and wait for dough ($) to rise.Years ago, one of our search experts and I had a running debate. I insisted that the key to good traffic was great content. He argued I could post Java code in Braille, and as long as it was all properly optimized he could drive the traffic. The truth, we both knew, was somewhere in between (closer to my side of the argument, of course). But video is a different matter, partly because its creation carries heavier baggage, partly because it's not so easily indexed, partly because it lacks mass. Few people know how to make money because few people know how to drive traffic consistently. (And even if they did, traditional online advertisers are still unsure how to maximize the opportunity.)

Architecting For Viewership

There are plenty of options for driving video traffic (half-naked women), the first of which truly does require creating compelling video content. But what compels online is subjective and proven only with time and experimentation. Let's put aside placing television shows online, along with cats bathing monkeys or dancing babies on YouTube; the former is an issue of geography, the latter relies on the unpredictable viral winds and whims of consumers -- figure that one out, patent it, and make millions. No, I'm talking about something that helps your audience get its job done or keeps them informed, or just flat out entertains them, all in a palatable viewing environment.

We posted a video late last year of Yahoo's former CEO Jerry Yang practically melting down on stage at Web 2.0; it went bonkers. Some of this resulted from enabling others to embed the video, but most of it was because there are just some things people want to see.

My colleague Alex Wolfe, who creates and edits his own video content, sometimes in the confines of his home using a pint-sized camera with less-than-optimal lighting, gets lots of views because his content is hot and timely and people think he reminds them of Tony Kornheiser.

Sorry for this answer to the content question: Figure it out; I'm still wrestling with it my damn self.

But several months ago we started experimenting with some basic ideas. They aren't complicated or bone rattling, just basic Web stuff. Like many sites, especially those that post hundreds or thousands of videos each year, ours use a third-party video hosting solution (in our case, BrightCove). While all the video resides in that system, we pull a feed and use this feed to populate individual pages on our site dynamically. Each video gets its own page; each page gets its own metadata; all of this metadata is fed to the normal Web search engines. (Waiting for the dough to rise.)

One more thing: We don't just put the videos into a central video lounge in hopes people will stumble upon them; nor do we simply post the videos on our home pages. In addition, we tie each video into our site taxonomy, so that if you are anywhere on our sites, you'll have videos that are, contextually, in the same ballpark as the normal Web content you are viewing. This system is as imperfect as any top-level taxonomy, so the relevance only cuts so deep for now.

Next Up: Video Search Engines

The problem with Web search engines is that video can't be indexed like text, and that's where specialized video search engines come in. AOL's Truveo is the market leader, while Blinkx is the Valley's darling. These companies hold mostly similar world views, namely that when someone goes searching on the Web, they are looking for something specific, while someone searching for video is browsing to be entertained; browsing results in sporadic adventures into anything of interest, usually ranging from The Daily Show or Ellen clips to Christian Bale meltdowns.

The challenge, then, is how to optimize nonentertainment-based video for these systems. Both Blinkx and Truveo recommend submitting your content to their engines to be indexed. This isn't as straightforward as it might seem. We can, for instance, generate an MRSS (XML) feed for Blinkx fairly easily, but then our video would essentially be hosted from the Blinkx site, not ours.

While these engines do crawl the Web, it is often difficult to discern what is video and what isn't, especially if you're pulling in your content from a third-party hosting system. Both companies reminded me that the best thing to do is have great content, and that in fact their search engines try to measure and reward just that because it also is in their interest to be serving up the best content for users. One of the factors the Blinkx engine uses is video popularity. For news items, it factors in the temporal aspect.

Truveo recommends that you have a vertical expertise, that you include all of the right metadata in and around the video, put metadata in your feed, and in the metadata be sure to include what type of content it is (like a reality show). Also, include a compelling thumbnail; users click on compelling thumbnails.

Other things that matter in search rankings: the speed of your site (specifically the load time for your video), whether something is an entire show or just a clip. When looking at the metadata, the engines look at how you've categorized the video, its title, its tags, and who the director or artist is; they also looks at how the community responds to the video, and the quality of the video stream itself.

Blinkx also processes the video content, with an army of computers doing speech recognition and visual analysis, which, when tied to metadata, gives the engine more clues about the video and its proper context. Blinkx claims to have indexed and processed more than 30 million hours of content using large server farms in the United States and in Europe. When I asked Truveo about speech recognition, it said that the only real value of evaluating a stream is to validate metadata; otherwise it's not clear how else to use it and whether it's worth all of that extra effort (obviously this is much different when talking about things like video surveillance).

There are some other things you can do with Blinkx. For example, you can now use the company's premium play service, where you essentially buy a sponsored link on the search engine, and you only pay for a click. You also can use its Advanced Media Platform API, where you allow Blinkx to process your video, and it then offers the data (say the speech converted to text) back to you so Google can find it easier.

Truveo is the big monster in this, with more videos and a bigger user base, but Blinkx also is innovating. Blinkx CEO Surangi Chandratillake (for a great pronunciation of his name, watch the video above; you'll find yourself saying it all day) told us the company is focusing a great deal of time on the advertising problem, especially where user-generated content is driving viewership; these videos, he said, aren't generating revenue and they're costly because you have to host and serve them. For Blinkx, the solution is targeting. If you understand the video and the intent of users, your success rates will be greater. With the company's Ad Hoc product, you can pop up an ad when something relevant is mentioned (say if it is mentioned that Union Square in San Francisco is a great shopping destination, you could pop up an ad for a hotel in the area).

All of these suggestions and solutions are helpful and innovative; we are trying many of them ourselves. However, outside of Web surfers who are specifically looking for video entertainment, the future Web user will be looking for content in whatever form (text, audio, video) helps them do their jobs. These video search engines, or their capabilities, need to be blended into Web (Google or Yahoo or MSN) search.