Some minor issues

I promised myself I’d blog some thoughts I have after attending the FOWA conference earlier this month in London. Things I noticed, trends, ideas, .. Here’s the first: the search for relevance.

Personally, I believe this is going one of the big topics the following years. Content creation is growing massively, an thus the amount of content rises fast and steady. But not all content is interesting, not for everyone. So how can we make sure the right kind of content hits your screen, and equally important: as less as possible crap.

The subject was touched in a number of sessions and subjects. How can Digg filter out these results that matter the most? How can Amazon optimize the recommendations. What about Friendfeed and Facebook? Personal facts of our friends and people we know are growing every minute. We’re not interested in all of that! But in some things we are. How do we filter this out?

The 1 billion dollar question alright. But there is an evolution going on, at least that’s what I learned from Digg and Friendfeed. Especially the latter is doing an incredibly good job to deal with it. Mechanisms to calculate interest include looking at the people who you’re most close to and see what they like, and your own rating behavior of course. That way dynamic clusters can be calculated. First you have to define a person’s opinion leaders (in an algorithm: people you tend to agree with, or people who’s contributions are valuable for you). Second you track their actions and translate them to the person it’s all about. Dynamic because your opinion leaders differ from mine.

Interesting is that we’re going back to the classical theories of influence. I know this has never been gone, and online as well these influential theories are applied. However people always assumed that some people are opinion leaders / influencers, and some not. I know I’m cutting corners here, but it’s generally true.
The next refinement movement was introducing the niche concept. This way reach wasn’t the biggest factor anymore. The expertise somebody has on a given subject was getting more important. And sure, expertise is measured as reach again, though not in absolute figures

Next phase: the individual influencer. As in real live indeed. One doesn’t have to be an expert on the subject, the fact that he/she is a friend of you makes you listen. So social roles and individual connections are back in the equation.

Now, the big challenge is to filter these out. If you can identify someone’s prime opinion leaders, in a dynamic way as well because depending on the subject you trust a different person, defining relevance just got easier. And of course there are grades, people who’s opinion matter the most, people who’s opinion matter somewhat and people who’s opinion doesn’t matter at all. In a way this is related to David Armano‘s Influencer 2.0, illustrated with ripples. Except in an other way it’s the complete opposite as well, because in that model it’s all about reach.

I think there’s one principle that’s a bit underrated to define (dynamic) opinion leaders, and that’s behavioral targeting. The way influencers are defined now is usually action based: you subscribe, vote, comment. However, there’s more than that. Reading something for example, or hanging out a certain amount of time. Following a link. Following someone followed by somebody else. I know it’s hard, there’s lot’s of data mining and not easy to program the algorithms, but in my believe this still is a treasure to be discovered.

One last thought (Kevin Rose, are you listening?) timezone does matter! Being active in Europe I know Digging something throughout the day is killing your chances to have it picked up. There’s just to little people (too little opinion leaders?) at that moment to translate your submission into an instant hit. So in the end, this is about reach after all. No reach, no effect. Taking this time difference into account may change something. On Digg this could be a section on the upcoming feed that states: Dugg at this time in Europe .. meaning x hours earlier. Filtering is one thing, making sure the information passes trough the filter is something else.