[True for all posts on my blog, but bears repeating for this one - opinions are my opinion and aren't informed by anyone at Google or otherwise that actually works on this stuff :-)]
About 10 months ago, I started a blog post titled "How we f-ed it up: the Stream/Newsfeed" (screenshot below) so:
I jotted down a few bullet points, but realized in a couple of minutes that I was railing too much (in some cases at people that I knew) and not making any points that I wanted to read. So I gave up - there were other things to rail about.
I started the post again two months later after a discussion with co-workers that looked at me funny when I told them that streams/feeds were a bad idea for what they were working on, since they were going to die one day anyway.
I liked the title better this time (screenshot below).
But then I never finished the post. :-/
But the Facebook feed announcement made me revisit the draft. If you actually see this, I finished the post this time. :-D
I think the feed as a format to share information is now horribly broken - it wasn't always broken, and isn't broken for many uses today (e.g. it works well enough for Twitter), but it is incredibly sub-optimal for the majority of the use cases that its used for - particularly updates on social networks and content consumption services.
Think about what the first thing that any social service (or any content service) wants to do for their users
- to start with it needs to quickly tell you everything that you care about
- or might want to know more about and explore further
- its needs to do this knowing what you saw last (so that you don't consume the same stuff again)
- but in a way that you can easily find something that you just saw and wanted to see again.
- automated (i.e. code, not editors)
- personalized (what every person cares about as news - i.e. which friends' baby pictures do you pick, is it a big deal that a friend of a friend got engaged to a friend of another friend? did your cousin go on another vacation? - is very different from the algorithmic challenges of putting together just world news)
- structured (what are the right sections? - are they really photos and music? should they be family, work, friends? or something else?)
- and tries to sieve through and rank the incredible amount of data we have today (e.g. every song every one you know has ever listened to.)
- understands your state and moods (what did you see last? more importantly, what are you in the mood to see now?)
- Getting the right presentation format is hard: The consensus is that a "Pintrest-like UI" makes sense and there's tons of news-like services including Rebelmouse that seem to be right directionally, but its not clear that format will work as is for social information.
- All the right data isn't there yet: as much as we like to focus on the data overload problem, we also have a data incompleteness problem - e.g. yes there's a lot of baby pictures on Facebook and pictures of flowers on Google+, but I still miss the ones I should care about most.
- All the data there is, is still hard to convert into interesting news: i.e. we still have that data overload problem.