In part 1 I explained what we have right now: social networks and crowd processors. I then expanded on the future of social networks. Now let's talk about the future of crowd processors, which I called "social software" before. To be more usefully specific, let's give this software its own name:
- Social processor
- Software the creates value for its users by utilizing their social networks.
Social processors are combinations of the two existing types of social software, social networks and crowd processors. This solves two problems:
- Social networks don't do anything. They collect data about your social graph and then... nothing.*
- Crowd processors (see previous post) do a lot of processing, but their recommendations often throw up combinations that are strange or impersonal.
How crowd processors work
Crowd processors do a ton of processing on all their members to calculate recommendations of various types. They take two approaches:
- calculate global recommendations, e.g. the Amazon Best Sellers list: the first-pass approach, this quickly begins to fall down as membership grows, since you get a "lowest common denominator" effect: your food network begins to recommend McDonald's, your music network recommends you try Justin Timberlake, and Amazon recommends whatever Oprah is reading this week. This is an equivalent to the social paralysis problem in social networks: only the stuff that offends no-one can be recommended to everybody.
- calculate similarity recommendations, e.g. Amazon's "customers who bought X also bought Y": this is the current standard approach, and involve 4 steps:
- look at everything you've said you like, A
- find the people, B, who also like A
- find all the other things, C, that people in B like
- from C, subtract the things that everybody likes, to get distinctive results
What will social processors do?
The problem with both approaches taken by crowd processors is that they are an approximation to the real world. In the real world, you discover things you like from your friends, and the more of your friends who like something, the more likely you are to hear about it. Equally important, the closer you are to somebody -- the stronger your connection -- the more likely you are to be interested in their recommendations.
Therefore, social processors will use the data about your social connections -- gleaned from an existing social network, not a new one -- to calculate recommendations from your social circle, and only your social circle. A partial example of this is GoodRec, who can recommend things based only on your friends' recommendations. Although they currently require you to create a new friend network (or guess one inaccurately from your GMail address book), they could easily get it from, say, MySpace's Data Availability program (assuming your friends are on mySpace).
Why is this so much better?
Think recommendations only from people you already know sounds a little limiting? Far from it. This is how the world already works. Your taste in food is based on what people have fed you, or eaten around you. Your taste in clothes is based, even if only subconsciously, on what the people you interact with daily are wearing. The same is true for books, movies, music, even political ideology. The difference between this way and a crowd processor's way is no false positives. Have eclectic taste in friends? Then you'll get wacky recommendations. Are your friends adventurous musically? Then chances are you are too, and you'll get their new stuff. The fundamental point here is that you are like your friends. That's why they're your friends. And the humans work, the longer you know your friends and the closer you are to them, the more like them you become.
But if this is how the world works now, why bother with software at all? Because in the real world, communication of preferences and interests and consumption is ad-hoc and incomplete. You don't start every conversation with everybody you know by asking them for an exhaustive list of the TV, movies, music and books they're consuming and their opinions of each -- although each of these things are popular topics of conversation. You can get the network to do the work for you, and when that happens, new things that are popular will spread incredibly quickly.
This is why it's important that social processors not attempt to create their own networks to work with. The network it uses has to be complete and detailed, with nuances such as lengths of friendships** and frequency of interaction (do you exchange messages all the time? Then you're probably close). It's not just tiresome to do this over and over, it's a critical stumbling block. Social data is a key part of recommendations, and if you have crappy data you'll get crappy results. It is essential that a social processor use real, accurate, detailed data.
Okay smartypants, so what does it all mean?
So if this is where social software is going to go, how do we jump on the bandwagon and make money? If I knew, I'd be doing it already, I guess, but some general tactics that I think seem promising are:
- Build a social network: by far the riskiest approach, as this space is approaching saturation. Unless you've got a really, really good user interface and a much more detailed social graph that would make you simultaneously more attractive to users and developers, this is probably unlikely to work.
- Build on top of a social network: this is the next-best thing. For the love of god, if you've got social software that you're building right now, please do not demand that I "create my profile" and "invite my friends". I've done that a thousand times. Let me enter my Facebook or MySpace credentials and get that information from them.
- Again with the interoperability: even better, build on top of every social network. And since that involves a lot of implementation, I repeat my conclusion from the first half: the people who crack making social networks interoperable will make money hand over fist.
* MySpace attempts to solve this by being about music. Facebook attempts to solve it by introducing Facebook Apps, but it turns out they're mainly about wasting time, because nobody wants to run a business inside of Facebook.
** Ever wonder why Facebook asks you when you met somebody?