Site Meter

August 31, 2010

The Five Stages of Filtering, Relevance and Curation

Tonight's news of Gmail taking on information overload directly, using a combination of intelligent algorithms and your own feedback to build in box personalization is yet another hallmark move to taking on the increasing deluge of content approaching us from all directions - be it our e-mail, static Web pages, audio and video, or the many different social streams which we have subscribed to. There is no question that content creation and sharing is exploding and people are completely incapable of giving every single message and item their full attention. And many smart folks are looking to bring solutions to find the best and ignore the rest.

As I see it, there are five major ways companies and individuals take on the topic of relevance.

1. Editorial Filtering

Loose definition: I am the smartest person. I know best for you, and I deserve to decide for you what is the most important.

Example: The New York Times, Wall Street Journal, CNN and most mainstream media outlets today, who for years have been trusted arbiters to find the most important news and bring it to us in the way that they decide.

New media examples: The Drudge Report, which has grown from one man's curation and sorting to a full team, and Techmeme, which was once almost completely algorithm driven, and now is staffed around the clock by savvy editors who pluck the best of the tech Web.

Of course, it is easy for an individual to be a curator. I share a lot of content, manually, through @lgstream on Twitter, as well as on Google Buzz, FriendFeed and Facebook.

2. Global Popularity Filtering

Loose definition: The will of the people can be trusted, and they will decide what is most important, thanks to the most votes.

Examples: American Idol, Digg and Reddit. He with the most votes wins and gains a coveted front page slot.

New media example: Twitter Trending Topics display the most frequent topics and hashtags, not necessarily the most important. Also, you can see tools like Tweetmeme and FavStar which watch for number-driven popularity online.

This would also have included RSS shared items counters of the past, such as RSSmeme and ReadBurner.

3. Social Filtering

Loose definition: What your friends like, you will like. If it's important to them, it's important to you.

New Media Examples: Facebook recommended friends and pages, which display how many friends like them, Google Social Search, which pulls results from your friends content, and FriendFeed Best of Day, which shows the items from your friends that gained the most activity over a time period.

4. Explicit Personalization

Loose definition: You told us what you like or don't like, and since you know yourself best, you know what's important.

New media examples: Netflix's star ratings and TiVo's Thumbs Up or Thumbs Down feedback mechanisms, as well as Kosmix's MeeHive project.

5. Implicit Personalization

Loose definition: Just be yourself. Read what you want, do what you want, and the system will learn from you, continuously updating.

Examples: Amazon.com, my6sense.

There is a time and place for practically all types of filtering. Mona Nomura of Pixel Bits today talked about the serendipity algorithm and what it means to marketers looking to leverage machine learning. With Gmail's announcement, ReadWriteWeb's recent coverage of TrapIt and ChatterApp in an article on consumer information overload, and the high visibility of Facebook's News Feed, against the recent feed, the challenge has grown to a level where you no longer have to convince people there is a problem in filtering, but instead, you need to make a concrete decision as to how to approach that problem.

I believe that there is a role for trusted curators of news, people who have unique access or unique insight, who can get to news more quickly than anybody else, or dive into it more deeply. I believe that social similarities are a good hint at an individual's interests, but they cannot replace your own preferences - which go beyond your ability to fill out a form and try to tell the truth on what it is that you really like. The best systems, as Gmail is trying to do (with some help from your own feedback on whether they are getting it right), happen naturally and transparently in the background.

It's natural I would think this given my work with my6sense, but I have long believed in there being a perfect place for humans to act as curators and guides, while there is another perfect place for machines to provide, to the best of their ability, resources to aid your discovery. So when you are challenged with a mountain of information coming at you from any angle, think of the best way to get it handled. Should you turn to an editor, to the will of the people, to your friends, or to code? The options are all there, and more tools are coming to help you attack the noise - because there's little chance it will fade away any time soon - and a very strong chance it could get much worse very quickly.

Disclosures: I am vice president of marketing at my6sense. ChatterApp and TrapIt are assumed competition. In addition, Kosmix.com is a client of Paladin Advisors Group, where I am a co-founder. I was also previously an advisor to ReadBurner, since closed.