Spinn3r

Rank Overview

Spinn3r Social Media Rank is a new feature we're making available to both our customers and the Internet community.

We rank social media by looking at friendship and link relationships within the social graph.

We limit our published rankings to the top 10k sources from each service. If you would like more results please feel free to contact us.

Social Media Integration

Our social media rank backend uses plugins to tie into various social media platforms. The same algorithm is used regardless of social network.

The following platforms are supported. Click a given logo to view rank information for that social media platform.

Authority

Sources (or nodes) are ranked by authority whereby the more friends or inbound links you have the higher your rank.

Our key differentiator is that we do not consider raw inbound link count to be an accurate representation of authority. This is highly vulnerable to spam and rank errors as users who attract a large number of links (either through black hat methods, link baiting, or viral marketing) can inflate their rankings (and harm other legitimate users).

We consider the quality of inbound links to be far more important. You can observe this in our results as the authority for a source is not a direct function of raw inbounds links. Some users can have high authority but very few (relative) inbound links.

Driven by Approved Sources

One trick we use to prevent spam in our results is that we start with a set of users called 'seeds' which then approve (or grant some authority) to other sources. In order to obtain rank you have to be at least partially connected to a seed.

We select our seeds by picking the top users of each site (and often the founders or other key members) and start computing rank forward their positions in the social graph.

This has the advantage that spammers and low ranking sources must obtain links from the current connected portion of the social graph in order to obtain rank.

Spam Prevention

This data can be valuable to Spinn3r customers as we can use this to remove spam from social networks by focusing on the top 1 million, or any arbitrary number of sources.

Without a solid ranking mechanims, it's possible for spammers to game the system by creating a large number of accounts or target valuable keywords.

Future Directions

The current version of our ranking is computed once and published manually. We want to get an idea of how valuable this is for both our customers and the Internet community.

Twitter

Right now we're only sampling Twitter as the depth of this social graph is quite large. We're estimating that we will require a significant amount of memory and 24 hours of compute time to compute a complete Twitter rank. This isn't much of a problem for our infrastructure (we are designed to work on datasets this large) but it will just take us some more time to have this coordinated.

Facebook

Computing the full rank over Facebook has similar computational challenges to Twitter and we would like to compute a full rank for Facebook users as well.

Strength of Relationships

All of the networks we are monitoring have friendship graphs. However, htis is only a piece of the puzzle.

You may have 100 friends but only ever interact with 10% of them. We are going to be extending this algorithm to factor in the strength of each friendship by observing how often you interact.

Custom Ranking

One thing we want to experiment with is to compute custom rankings for user based on verticals such as technology, politics, entertainment, etc.

In the current version of our social media rank is popular we will probably end up publishing rank for vertials as well.

Questions

If you have any other questions feel free to contact us.