Thursday, February 4, 2010


Some months ago I got interested in mining some Twitter data and played with their API and started following @TwitterAPI. Not much came out of it, except that I would get a few updates a week regularly about their API. Which seems to have progressed quite a bit while their website is still much the same. I decided to take a closer look.

It isn't hard to get data from Twitter. Kudos to them for that. Here's an interesting representation of the data I got about the different Twitter clients being used. The numbers are the percentage of times that client was used in a sample of public tweets taken over 6 days. Hover over the circles to see the data, or click on a row in the table on the left to see the corresponding circle in the chart.

(Feedreader users, you may have to click through to check out this graphic. All: I apologize for making you run a Java applet but this tool from IBM was perfect and I couldn't find anything in Javascript or Flash that had all this functionality)

48% of tweets are from the web. Which means the website. About half of that are from regular clients. From the looks of it, I guess about 8% are from the mobile. And the rest are from other web applications trying to get users to popularize their own content like tweeting about a video from YouTube or your scores from an online game, connecting your Tumblr (or blog/rss feed) to your Twitter account so updates get tweeted automatically, tweeting your Facebook or LinkedIn status automatically either through these sites or through an aggregator service like, etc.

So clearly, using Twitter for a viral effect is a popular use case - somewhat like the Facebook News Feed or FriendFeed. If you look closely at the various clients there, there are so many web apps are trying to get hooked into the users' Twitter accounts. And an API for that purpose makes a lot of sense.

Twitter lends itself naturally to mobile use, and I suspect Twitter has a higher proportion of its usage on the mobile than most other web apps. But the big question mark here is the Twitter website. So many people are still using the website. Why isn't Twitter doing something about that? Most Twitter clients offer much richer functionality than their site.

Firstly, by making the site a proper destination, Twitter will have a lot more data about user behaviour. Right now it is probably highly fragmented across the various clients. Secondly, being a destination is very useful when you want to experiment with features and see how users respond to them. As a semi-functional website, or a fully functional platform, they are removing themselves from the users, forgetting that a lot of the Twitter protocol (@ replies and RT's, for example) was invented by users. Thirdly, Twitter has given up control of content to clients. Which are not doing even a half-decent job. Sure, they have gotten better at spam filtering, but with millions of tweets a day, and hundreds of thousands of conversations, they can filter a lot more than just spam. Filter and collaboratively filter. And best way to bring this data to users is they were a destination. Lastly, there is a huge monetization opportunity in being the destination. And I don't mean ads. I am referring to Twitter as an incredible proven ability to enable conversations between businesses and their customers, regardless of whether the business is a 1-man show or a faceless corporation. Conversations are very underrated - in marketing, product design, in product support and this is where Twitter's real potential lies - in enabling multiple paradigm shifts, rather than being one.

Please go reclaim your website, Twitter.

No comments: