Content recommendation for years has been a startup graveyard. Sure, a few companies like Sphere/Surphace and StumbleUpon have been able to get out—but the number of folks who have tried to solve this is vast. I feel it’s sort of like the carnival basketball game—where they put a hoop like four feet from you and tell you that you can win a big prize if you get one in on three shots. The only thing you don’t realize is that the rim is bent and narrowed in a way that you can’t see and no one ever gets one in. It feels like someone should be able to do this right, but it just never happens.
People, on the other hand, are pretty great at recommending content. The best content recommendations I’ve ever seen that help me figure out what to pay attention to are ones driven by users—del.icio.us, Techmeme, Digg. In the case of Techmeme, half the job is done when you curate the list of which tech blogs you should be paying attention to. In a way, Techmeme isn’t really recommending stories as it is telling you which people you should be paying attention to.
And that’s actually a much easier task—finding people for me to pay attention to. There are some really simple algorhythms that people can produce to help me figure this out—and they’re simple and intuitive. Tell me who 5%-10% of my network follows that I don’t—so I don’t get recommendations to follow Fred Wilson or Robert Scoble, but I do get the entrepreneur in my network that I’m probably bound to run into and say, “How have we not met yet? We know 30 people in common.” How about just telling me all of the other entrepreneurs turned VCs—or all of the other people that teach entrepreneurship? Who else in the tech scene bikes? It’s not hard to match people based on this kind of criteria, but that really isn’t available anywhere in an easy format—one that helps me determine which attributes I want to follow people based on.
If I did solve for that problem—I believe it would be a pretty high quality stream to find me all of the articles that those other people are paying attention to. Solving for content recommendation on an article level is so incredibly difficult—there are so many factors that come into play. Do I want to read every article about the Mets, or only certain ones? Should 50% of my content stream be Mets articles, or should I be concerned about the earthquake in Chile, too? Computers aren’t very good at getting these kinds of judgment calls right—so why try?
I’d love to see a system that recommends a social graph to me—based on my personal and professional interests. That would get me so much further than current content recommendation systems. Get that right and you just have an interface and easy filtering/focus challenge which you would have had anyway—but at least we’d nail the universe of content we’re pulling from and that would make the content recommendation part easier.
And people want that—it’s why Mr. Tweet went viral and it’s what the team over at LiveIntent (FRC portfolio company) is working on. One of the reason why Twitter has onboarding and usage issues is because it’s hard to figure out who to follow, and without quality people to pay attention to, good luck getting someone to start tweeting or even understand what it is. It’s also a problem we had long term visions of tackling at Path 101—tell me, based on who I am, who I need to meet. It’s a basic issue that I think people should spend more time on than trying to figure out the single article I need to read today.