Thursday, July 24, 2008

Twitter Meta Index revisited

You're probably aware of the consternation and shock engendered by my How to Calculate Your Twitter Meta Index article.

I don't claim to have perfected this micro-blogging measurement tool. My simple hope was that a conversation might be started that addressed this important issue.

To facilitate the ongoing discussion, and for those new to the topic of a Twitter Meta Index, I felt it best to feature the comments in their own post. I've never done this before, but the comments have brought to the surface of blogospheric consciousness some micro-blogging esoterica that's vital for both business and personal applications.

First, to complicate matters even more, Twitter Status blog mentions Following/Follower data problems recently. Oy vay! We need some psychoanalytic ethnomethodologist mathematicians to work this out with super-computers.


We’re still in the process of recovering from the missing follower/following problem that occurred earlier today. Over the next several hours, you may see inaccurate counts or timeline inconsistencies as the correct data is propagated to all parts of the system.

One thing to note: Even after this recovery is complete, your counts may appear lower than previously.

In almost all cases, this is not due to missing data. The counts we display on your profile page are not always up-to-date. For example, when we remove spammers from the system (which we’ve been doing a lot lately), the follower counts are not updated in real-time.

As we push out the changes to fix this afternoon’s problem, the counts will be updated to reflect the latest numbers.


Now, back to the Twitter Meta Index calculation controversy.

(1) comment from Ben

Ben Kunz said...
First, solid idea, I've used a similar mental yardstick to judge the potential value of people on Twitter.

I think the formula is off a bit. (You also have errors on Calacanis in following-followers and Scoble on updates/sq root, no matter for the sake of this argument.) The *real* challenge with the current formula is you can get to the same index number from very different types of Twitter behavior.

A spammer, for instance, who follows 10,100 people, has only 100 followers, and made 16,230 updates will achieve an index of 162.3 -- exactly the same index as Ochman who follows 390, has 1,374 followers, and made 5,092 updates.

Thus the formula has two logic errors:

1. A larger difference in followed - followers can offset a larger number of updates ... so people with very different behavior can get the same index number.

2. Someone with a large positive number in followed - followers, a likely spammer, should have an inverse, or *negative*, index compared to someone with a large negative number of followed - followers, a likely thought leader.

All of this is a lot to digest late on a Friday night, but the thought is solid. If you could address the issues above, you've got something people could use.Then, if you could write a software program to let people pick their settings and automate who they follow back, you'd have a cool service.


(2) comment from Ike

Ike said..

You've got a lot of the right pieces in there, but there are a couple of other factors that take the Meta Index out of one's own control.

First, leave in the negative values! When you take the square root of a negative number, the result includes the square root of -1, which is i. "i" stands for "imaginary," which fits the theme!

Second, there is no accounting for how someone gathers followers. I have a large contingent for a couple of reasons. While I engage in conversations with many people, I was also included on Guy Kawasaki's Alltop list, which means I get a LOT of followers from people who just signed up for Twitter and are looking for people to listen to.

Am I to be punished because a bunch of people blindly follow me and never engage?

Second, there's no room in your formula for those who voluntarily downsize. Even with some discretion about who and what I would follow, I got up to 893 people in my timeline. When you have to click back a couple of screens on the web just to find people you know, it's a problem.

I slimmed considerably, dropping more than 500 people from my list. Those I dropped were people who I could not recall having discussed anything. They seemed interesting at the time of the follow, but over time have never engaged or clicked. Many were no longer following me anymore.So, am I to be punished by your metric for unfollowing people? For making MY Twitter experience manageable? I'd hate for people to plug me into a formula and rate me as an a--hole because I'm being out-followed by a 4-1 ratio.

I know that you don't think that about me, Steven -- so any metric you derive needs to be tested against more varied sample sizes, and against known people.I know it gets harder when you delve beyond just the public info listed in a profile, but I believe the following factors play a role in who *I* might follow:1) Percentage of replies. How many of the most recent 100 Tweets are @'s? I say the most recent, because behavior can radically fluctuate. Someone might get more than 1000 updates in before really "getting it," and you don't want to hold past statistics against them.

Just like how in Fantasy Baseball I rarely make roster moves based on anything other than the last month's performance.2) Percentage of INCOMING @'s. This is harder, but essentially is a measure of how much you're willing to engage. If you've got 20 @'s in your most recent 100 Tweets, but you've been @'ed 200 times in that same time period, then you're being quite selective (and maybe standoffish.)3) Different metrics for scale.

When you're following 5000 people, your Twitter experience is very different that when you're following 50. AND, when half of the 5000 you're following are ALSO following more than 1000, then there is a LOT that flies by the wayside. "Influence" doesn't scale, it regresses.

Here's the example. Let's say I ask for advice about Sharepoint. When my timeline was smaller, I might have gotten 5 or 6 responses. Guess what? With nearly 1500 followers, I *still* get 5 or 6 responses. Since most of the people I am following *also* have more populous timelines, they are less likely to see my query. The more you have, the more inclined you are to rely on Replies.Essentially -- the Potential for Influence is factored on the Follower/Following ratio, but the makeup of those Followers and THEIR lists need to be calculated in. If you think about it as a signal-to-noise problem, then ask "how attuned are your followers?"


This is where the Twitter Search tools can come into play, helping derive some of these stats.Just a couple of thoughts... keep plugging away. I have a feeling that the metric that ends up matching your gut instinct will be the best -- AND it will include factors that can be measured across other socnets.


No comments: