When I grew up, there was handful amount of channels on television. It then became a few dozens but still a manageable amount of channels. I even remember that I actually knew the channel’s numbers by heart, so “Movies” was channel 18, and 93 was MTV and so forth. Wow, remembering channels’ numbers. To the extreme, it’s almost as if I won’t need to search the web because I remember URLs by heart.
Available content became huge. TV now has hundreds of channels, and the web has millions of videos available to watch at any given moment. There is a lot of available information, and it makes it more difficult to assume I know what I might like to watch next.
Discovery is a global fundamental issue. Exploding amounts of information means people can’t discover what they like as easily. Too much information = no information at all.
Advanced technological discovery tools are set out to solve that problem. The online video discovery market is estimated at $1.2B. The online video advertising market alone grew 100% from 2009 to 2010 to ~$1.6B.
The Nature of Web Video Browsing: Broken into 2 Segments:
1) Assisting Existing Viewers:
People that are already watching the first video, and might be interested in watching more during that session. This is similar to someone opening the TV on NY1 Channel. There is already a need to specific content, and now the question is what might help that person to watch more and stick around.
2 ) Converting Non-Viewers:
The 2nd type is totally different and actually much more common. This is people that are NOT watching video and are not consciously aware that there is a video that they may like to watch. Those people could be reading an article, watching an image gallery, etc. Converting those users to watch the first anchor video is considered a rather difficult technological problem, and it falls under the general video browsing experience as it is likely some video discovery engine will offer users that read articles few videos they might like. That’s where it most likely to being.
What Makes for a Good Video Recommendation?
As discovery engines are relatively new, with Amazon being the early ones to do a rather good job at it, and Youtube doing a sucky job as Jeremy Scott wrote, there is still an unsolved common question – what is a good recommendation engine?. If the article is about meat, should the video recommendations be about other meat videos? Makes sense right?
However, let me ask you — If you had known that the user that landed on that article is a vegetarian, or if studying million users landing on this article or topic you were to find out that all 90% of those readers actually wanted to watch movies kitchen tools and not more food – which is better? What would you chose.
- If you’re the editor of that site, I’m guessing you would pick “more food recommendations”.
- If you’re the community manager, you would go with what users love best, and you would go with the kitchen videos. This is because you want your users to have a good experience and come back happy.
- If you’re the ad-sales guy, you would chose the kitchen videos. You get commission on the amount of available inventory you can sell ads against. You need users to click, watch, stay and come back again.
- If you’re just a random guy, my guess is actually that you would judge a discover engine to create similar content. It just makes sense. Meat = more meat.
Who is right? What is a good video watching experience? Should people be served with more obvious recommendations or those that might actually interest them so they stick around?
As opposed to search engines where there is direct and clear intent as users actually type what they want, discovery engines have a bigger challenge of “guessing” what’s right without users ever telling them directly.
Not only discovery engines are required to guess what’s right, they need to go through some education process that “non similar” recommendations might be the right recommendations. Not easy as most of us “random guys” think otherwise.
What do You Think?
What is right thing to do then? Similar recommendations or the right recommendations? Feel free to comment below.