A bunch of thoughts have been buzzing around in my head recently about social software: what it is, where it's going, and what that means. I'm going to try and get those thoughts in order here. First, as is always useful, some definitions:

Social networks

These are characterized by one-to-one, personal, long-term connections. They are used primarily for communication and/or contact-management. All the usual culprits are here: Facebook, MySpace, LinkedIn, Twitter, Flickr; plus myriad others less familiar to my audience: Hi5, Ning, Mixi (Japan), Orkut (Brazil), etc..*

Crowd processors

These are many-to-one: they take data gathered from the behaviour of large groups of people and process it into something that is useful to you. One-to-one contacts may exist, but they are not the primary benefit of the service. This is where you find Last.fm, Pandora, Amazon, Del.icio.us, Yelp.

Social software

This is many-to-many, and is essentially a combination of the other two. It uses data from social networks to process data from those groups to derive value. It's my opinion that this is the holy grail: this is how real human beings work; their lives are a reflection of their community. It's also my opinion that not many of these exist yet, although various apps are getting there: Facebook and Twitter are approaching it from the social side, Last.fm and Amazon from the crowd processor angle.

There is a bunch to talk about here, so I'm splitting these up. First, let's tackle social networks. A lot of people are sceptical of socnets as the Next Big Thing, having seen the initial success of Friendster and its demise, followed by MySpace, followed by Facebook: if everybody keeps abandoning their socnet for the next new popular one, surely the whole industry is just a fad too?

Rapid growth, and limits to growth

A widely-held idea about socnets, fuelled by the fact that they often get so popular so quickly, is that there is going to be some sort of "winner": that one network will finally be good enough, have exactly the right mix of features and privacy, and it will grow until everyone is a member and abandons all competing networks. This is not going to happen. Social networks have a natural size limit determined by a number of competing forces:

1. Positive: virality: Metcalfe's law, more commonly known as the "network effect", says that as more people join a network, the number of possible connections between them increases exponentially. This tends to make a network more useful the more people who join. This is why social networks are intrinsically viral.
2. Positive: address book effect: It's nice to have a single big list of everybody you could ever want to contact. At a certain point you begin to add people reflexively. This is great for the size of the network, but bad for the usefulness -- see the next two point.
3. Negative: overload: Dunbar's number, the hypothetical limit to the number of people you can really "know" at one time, is somewhere around 150. If you have more than 100 "friends" on a social network, they are not really "friends". They're people you used to be in touch with, or met once but don't know very well, or are trying to hit on or do business with. Beyond 150 people, the value of knowing what those other people are up to tails off -- because you don't care.
4. Negative: social paralysis: the more people who can see what you're up to, the less you can actually do. Stuff that you'd talk about and pictures that you'd post for your friends to see are not what you want your parents to see or your work colleagues to know about. The key point is that nobody has just one personality, they have several, and social networks do not currently have a good way to manage this.

This is why people went from MySpace to Facebook to LinkedIn to insert-socnet-here: they aren't deleting those other accounts; they are instead striking different balances between these forces to suit themselves: using one network for their friends, maybe a different one for family, one to keep track of what people are doing, one as an address book. They're also trying to make walls between their different personalities and interests: one place they talk about their hobby, one place for relationships, one for their crazy S&M fetish group.

Partitioning is the next hard problem

People will have multiple online social networks because people have multiple social circles in real life. The sooner socnets themselves learn to accept this and work with it instead of against it, the happier everyone will be. The future of social networks is that everyone will have a few, and they will mainly be small. The "address book" network will be huge, but you probably don't want to be that network: it will be dull because of high levels of social paralysis, and also a victim of constant hacking and spam, because the biggest user database makes you the juiciest price.

Socnets are not unaware of the problem of identity and partitioning, of course. Yahoo! for a long time allowed users to create "aliases", separate usernames with different profile pictures and privacy levels, that they could use selectively across various Yahoo! properties (they proved cumbersome and confusing, and have been phased out in favour of just letting people have lots of different usernames, which is what they were doing anyway). Facebook and several others have Groups, which allow you to sort your friends into categories that you can contact and invite to events as a group; this solves only part of the partitioning problem -- the easy part.

There are a bunch of much harder problems still to be solved around partitioning. Feel free to base your startup around one or all of these:

• Selective data portability: want to use the same profile picture on two different socnets? It should be as easy as entering your password but not automatic -- you don't want your S&M profile photo turning up on Facebook (or do you...?).
• Friend discovery: on joining a new socnet, it should be possible to find your friends from other socnets. This is a thorny problem, because if the network in question is that S&M club, they might not want to be findable -- and that should be an option.
• Identity verification: is this person the Stephen Colbert, a Stephen Colbert, or fake Stephen Colbert? And how do you prove it?
• Alerts and messages: if you join twenty different socnets -- not unreasonable for a person with varied interests and an active social life -- you end up with 20 different inboxes in addition to your email box, each with their own UI and bugs. Ugh! A unified messaging platform is key.
• Handling Dunbar's number: oh no, too many connections! A sea of irrelevant status updates! An app that works out who you actually care about and filters by the strength of your social connections will do well: a basic example being Twitter's web-only vs SMS functionality, but look for something more subtle.

Let a thousand socnets bloom

So what are the key conclusions to be drawn here:

• No big winner, lots of also-rans. Not just "not one of these guys", no big winner, ever. It will be a classic long tail situation: a few big players, and then thousands of smaller players adding up to just as many users.
• To succeed, partition yourself. LinkedIn is a great example of how to do this: choose one type of social circle, and build tools specially for those people. MySpace arguably does something similar for music lovers, but I hate MySpace and hope somebody else trounces them (Last.fm are still too data-focussed).
• White labellers will do well. The best-known of these is Ning, but there are a ton of companies working to provide generic social networking software for you to create your own social network. These guys will do a lot of business, but the size of these networks will be very limited: past a certain level, a custom network like LinkedIn with specific features for that group will do better. (Which is why people will never really use Facebook for dating; even if they have smaller pools of users, dating sites have better tools)
• The Next Big Thing in social networking will interoperability. Generic social networks are pretty much already here on a regional basis; there's not much room for more. All five of the hard problems above are about gluing the networks together.

* I'm not going to talk about Friendster much. Friendster failed for technical reasons, so it's not really interesting.