Our Blog

Under Twitter’s hood


Posted by
2 September 2010
11:16
0 comments

The beauty of Twitter is its simplicity and atomicity. It’s a WYSIWYG service – very little of its complexity is hidden. Most of Twitter’s auxiliary functionality was not invented – it organically grew out of the user community. Take retweets, direct messages, replies and mentions, for example. All these started as text cues users left each other in messages, without any preconceived design from the Architectures of Twitter.

Therein lies its beauty.

This does, however, imply that there might not be a lot of information in each tweet. This is not entirely true.

Twitter API

Recently we have built a simple Twitter application that displays all available extra information. It serves us as a quick reference of what is available, how many people are using it, etc. Kind of like a brainstorming tool (Click on the orange triangles to expand extra info about users and tweets).

Some of the fields of extra information are quite interesting, and deserve a spot of lateral thinking:

user: followers_count, friends_count, status_count, list_count – how does this reflect user influence?
These four counters form an interesting picture about a user’s online activity. The really exciting part is the idea that a user’s online influence, or rank, could be determined from these, and possible a few other, parameters. The magic algorithm to using these parameters to calculate a user’s online success still eludes everyone who tries to do it – at least to a certain degree.

The most success so far have been achieved by the Twitter influence calculation service, Klout. They calculate their klout score using these, and more than 20 other parameters, in a secret mathematical soup recipe, to produce an influence metric.

in_reply_to_status_id and Twitter conversations
While it is possible to reply to tweets without setting this field, most clients do set this field when a user replies to a tweet. It makes it possible to trace entire conversation threads, and display it as such. The Twitter user experience thus becomes multi-dimensional.

Conversation metrics also play a large role in determining a user’s online influence – the more a user is retweeted, the higher their online influence is (that is the assumption – like I said, this is still largely untamed territory).

coordinates and place: geolocation
Geolocation tends to open a whole new wonderworld of usability and social gaming. The current leaders in this field, utilizing their own network, is Foursquare. having said that, both Twitter and Facebook have recently jumped on the places bandwagon, making it possible to build something like Foursquare entirely on the Twitter or Facebook platform.

While many twitter clients are starting to use Twitter’s geolocation fields to display a user’s location on a map, the real potential lies in the gameability and utility unlocked by geolocation.


It’s becoming clear that there is a lot more potential in Twitter’s data than its simplicity seems to imply. Add to this the possibility of adding metrics, like klout score, and the possible future of user defined, service specific parameters, and the amount of data per user or tweet might actually grow exponentially over time.

The crowd that we have recently come accross that shows the most promise in harvesting the potential of all this data, is Datasift. They have yet to launch their product though, so for the moment, the potential grows by the day.

Posted by

Leave a Comment

Your email is never shared. Required fields are marked *