Connection Types

Wednesday, February 23, 2011

Connection Types

I look at many variations of social networks at Spark Capital. There are a number of ways to organize social networks and some ways are more useful than others. You could organize by media type (ie Youtube is a Video network. Flickr is a Photo network), but I don’t see much value in this form of organization because it’s superficial and subject to simple change (Flickr now offers video). A more informative way to slice networks is by how connections (or edges, in graph-speak) in the network are defined. Here are some variations I see, with examples:

One-to-Many, single direction: Best example is Twitter. Twitter’s most popular use case is as a broadcast medium that allows one person to push messages to followers efficiently and without the burden of having to listen for responses or uphold a conversation with all the followers.

One-to-Many, bi-directional: This is a network in which one person or organization attempts to hold a conversation and be responsive to all edges in a network. Some examples are group chat products like Groupme. Some organizations like Zappos do attempt to use Twitter in this manner, but it’s difficult to keep up with the flow.

One-to-One, bi-directional: Each edge on the graph is a two-way communication between two individuals with an expectation to uphold the conversation. Some examples are private chat products such as AIM or Skype on the desktop and Kik or plain ol’ SMS on mobile devices.

One-to-One, single direction: Each edge on the graph is a one-way push between one individual to another with no expectation of response. This style of graph is rare inside social networks because it would make for a pretty lonely experience. Instead examples are stuff like transactional email between an organization and its members is a good example of this, such as a receipt from Square or a new follower notification email from a social service.

Many-to-Many, bi-directional: This is harder to define, but many people communicating with many other people is not easy to imagine. But, I do see this type of behavior between guilds in World of Warcraft, where groups are coordination with/against other groups, each of which is a collection of individuals.

Many-to-Many, single direction: I’m not sure how this would work… it’s a group of people all pushing messages to another group with no expectation of a response. Would love to hear examples in the comments.

So what?

Good question. Some social networks start in one category and try to evolve over time into other categories, often with mixed success because a different graph model is counterintuitive to users existing expectations.

For example, Facebook started as a one-to-many bidirectional graph: it was a way to broadcast your profile, photos, and conversations to your peers whom you explicitly permissioned as OK to access your data. But, as Facebook saw Twitter rise in popularity, Facebook tried to expand into the one-to-many single direction category by loosening default permissions of users’ data, thus reducing privacy. This shift from one graph category to another violated users’ prior understanding of the service’s privacy model and led to repeated controversies. Google Buzz made similar mistakes when the company assumed that the Email Contact graph would be a seamless way to bootstrap a social network, which was rife with issues.

When I’m looking at a new social network as a prospective investment, one of the first questions I dig into is what type of connections are being established inside the graph. It’s an organizing principle worth establishing early because it will determine many future product and business design decisions going forward.