Dylan Hunter

Was an excellent post on O’Reilly Radar by Nat Torkington reminding us in the course of designing social suggestion-based sites / software to not forget one of the most critical reasons for in the first place - serendipity / new discovery. Sometimes this is best accomplished by “more like this,”‘es and/or “people who viewed this purchased..” kinds of things, (i.e. narrowing / refinement goal-oriented) and sometimes best accomplished by purposely broadening / adjusting the scope to introduce some new variance and seeing where things progress from there.

Many years ago there was a little search startup called Direct Hit here in MA, (who earned $500M on a cheaper version of what I wanted to do, but that’s another story) who really was the first entity to hang their hat on social suggestion-based search results, knowing full well that most folks only look at results 1-3, and almost no one beyond 10, so that those results that it initially presented as 1-3 would almost always tend to stay there, whether they were the best results or not - i.e. for anyone who’s ever used Excel, an infinite circular error.

As we all took this kind of click-through analysis / adjustment functions in to become one of the ranking criteria we used, (and / or at least used to internally evaluate how good a job we were doing in delivering relevant results - with all of the appropriate mechanisms for negating bounce - i.e. clicking through on a link, saying “this isn’t what I wanted” and hopping back to the search page, etc) the import of regularly introducing different results into the mix to make sure we were actually doing a much better job of delighting folks became more and more clear then, and still holds true today.

But enough of me babbling, Nat does a superb job of describing, so go read it from him!