February 5, 2009

...Learn TDD with Codemanship

Twittizenship - A Metric-In-Progress

Just for jolly, I've been signed up to micro-blogging and social networking site Twitter for about three weeks now. Although the signal-to-noise ratio is quite poor (especially emenating from me), in just those few weeks some very interesting connections have been made and some very interesting outcomes achieved by "Tweeting" (the term Twitter uses for posting updates).

If you're not familiar with Twitter, it's pretty simple, really. There's the micro-blogging part, where you can post updates about what you're up to, or respond to other user's posts, or just post simple thoughts or links or wotnot. Because Twitter accepts posts by SMS, the length of a post is a maximum of 140 characters. So brevity is called for. (And, yes, it is a struggle, thanks for asking.)

Then there's the social networking part, where users can sign up to follow another user's updates, and connected users can send each other private direct messages. Thus a directed graph of followers and followed is constructed along which memes can propogate - which is the whole pay-off, I think.

Most users have a few dozen to a few hundred followers, and follow roughly as many. There's a kind of sense of obligation when someone follows you to check out their profile and - if they're not obviously a"marketing guru" or a "social media expert" (both euphamisms for "social network spammer") - you follow them back. Well, it's only polite, after all.

If someone you are following Tweets something of reusable interest, you can "Re-Tweet" (or "RT" for short) their post so that people following you (and not the original Tweeter) can see it. As with my famous forest fires, Re-Tweets may be Re-Tweeted, and as such can spread throughout the social network like wildfire when someone hits on a hot topic. Yesterday, Saving Bletchley Park was one of the most Re-Tweeted topics, thanks to a Re-Tweet by someone who we'll get to a bit later.

There are web sites that monitor Twitter traffic and report on various interesting trends, like what topics have been especially popular, or what posts have been re-Tweeted most that day, and so on. Thanks to it's simple API, Twitter data is pretty openly available. Unless they've specifically blocked a user (or all users) from seeing their timeline, any user's updates can be accessed, as well as lists of who they are following and who's following them.

This makes for some potentialy fascinating reading. For example, this list compiled for NOOP.NL of The Top 50 Twitterers To Follow For Software Development has given me hours of fun comparing the profiles and stats (and Tweeting styles) of the great and the good in our profession.

It's a commonly-held notion in Twitter culture, for instance, that Twitter is not a broadcast medium. It is considered the height of bad form to dish out update after update about yourself and mostly ignore the rest of the social network in which you are embedded. It's also considered a poor show if you are followed by lots and lots of people, but follow very few yourself. This means, essentially, that you are largely one-way traffic.

So, being a curious soul, I've devised a little metric - which I call Twittizenship - to help gauge just how good or bad a Twitter user has been at connecting with other users. Would you like to know how NOOP.NL's top 10 scored?

I've devised a formula thus:

TWITTIZENSHIP = (LOG(Users Followed)/LOG(Users Following)) x Number of Updates Addressing Other Users in Most Recent 25 Updates

In other words, the greater the proportion of people you follow to people who follow you, and the greater the number of times you reference other users in your updates, the greater your Twittizenship.

I've chosen to use Logs because there's a common sense issue here - as you get more and more followers, it becomes harder and harder to engage with them all, so I think I've made it fairer and more realistic by not keeping the ratio of following to followers linear.



Okay, so it's not an ideal metric. Just sampling the 25 most recent updates, for example, means that Twittizenship could fluctuate dramatically day to day. If I had the time, I would use the Twitter API to calculate the average per 25 Tweets across all of them.

And it fails to take into account the possibility of someone addressing the same tiny subset of Tweeters in multiple updates. Again, some jiggery-pokery with the Twitter API could calculate the reach as well as the frequency of their interactions with other Tweeters.

But the numbers I've come up with very roughly reflect my own qualitative impressions from examining these profiles. Kathy Sierra does seem to engage and interact noticeabley more than Jeff Atwood, for example.

Finally, for a bit of fun - as BBC elections pundit Peter Snow might say - let's do some interesting benchmarks that I hope will put these figures in some sort of context.

I have 131 followers and I'm following 133 other users. Out of the most recent 25 updates, 12 address other Tweeters. My Twittizenship is therefore 12.03. Let me assure you, I have not updated yet today or followed any new users, so I've not had a chance to game the numbers. Acclording to my metric, I'm a better Twittizen than the best in NOOP.NL's top 10. And, you know what, I reckon I jolly well may be, too.

Stephen Fry (I said we'd get to him) is an internationally famous writer, broadcaster and film director and has a massive Twitter following - 125,000 and growing very rapidly. His Twittizenship comes out at 8.84, but his stats are such an abberation that, like abnormally high IQ's (and I suspect he may be the owner of one of those, too) the figures are much harder to trust.

Jonathan Ross, another very famous UK Tweeter with a huge following, has a Twittizenship of 3.50

Okay, okay! I'm still working on it!






Posted 9 years, 4 months ago on February 5, 2009