Analyzing Pune’s top twitter users

Twitter
Image via Wikipedia

(Twitter has quickly become one of the most important new methods of communication, and Pune’s techies have taken to it quite enthusiastically. As its popularity grows and more and more people find out about its utility as a medium of communication, conversation, networking, as a source of news, or as a source of information about interesting hobbies or people, the number of people on twitter is growing. One of the questions most people have is – “Whom should I follow?” and the related question “Who are the top twitterers of Pune?” That is a difficult question to answer because everybody’s criteria are bound to be different – and existing “objective” mechanisms of measuring this are not really that good. Last week twitter released lists, and Dhananjay Nene argues that lists are a new way of measuring the “follow-worthiness” of twitter accounts. With this in mind, he analyzed who would be Pune’s top twitterers according to a number of different criteria. His he published some results of his investigation on his blog /home/dhananjay, and it is reproduced here with permission for the benefit of PuneTech readers.

This should be interesting to you for a number of different reasons. First, of course, this gives a list of the top twitterers in Pune. It is also an example of how a simple question can get quite complicated when you try to get computers to find the answer – and the approaches taken by different algorithms and their results are interesting to see. Finally, I think this is a sign of things to come – I’m convinced that twitter will be an integral part of the communications of the future, and twitter lists are an important way in which we will separate out the spammers and idiots from the useful content on twitter.

And, oh, by the way, are you following PuneTech on twitter? You should – there is info+links in the PuneTech twitter that will not be found on the PuneTech site. (And if you’re not on twitter at all, then please crawl out of your cave and get with the program.)

Anyway, here’s Dhananjay’s article.)

So twitter launched lists and many believe these will be a new mechanism for computing reputation instead of the current defacto followers. It is not a restricted knowledge that using follower counts as a measure of effectiveness of twitter is a extraordinarily error prone and brave exercise due to the obvious. Given the appearance of twitter lists, I was keen on figuring out if there is a way to reasonably measure effectiveness of a twitter id. This post details the exercise I went through. While there could be discussion around the exact semantics of such a computation and whether the results are consistent with everyone’s expectations, let me assert that I find the result sufficiently superior to anything else I’ve seen or I’ve been able to imagine so far. And that may stem from or despite the fact that two of my twitter handles (@dnene and @d7y) feature in this list.

As an input I took the top 50 handles from pune from twittergrader.com. Why top 50 ? Only part of the process was automated – the remaining required manual input. I did not want to spend too much time on doing data entry. This also gives you the twitter grader grade. I subsequently looked at the reputation of the handle in Klout, looked at the lists which included the handle and finally also looked at the twitter rank as expressed by yet another site twitter-friends.com. I computed rankings using each of these. I finally created a sum of all the ranks, and create a composite rank based on the sums. The interesting aspect of this computation was not just the end results but also some of the intermediate results.

So without further ado – here’s what I found

Ranked as per twitter grader

  1. shinils
  2. arthut
  3. indianguru
  4. sandeepjain
  5. tmalhar
  6. brajeshwar
  7. rohit_shah
  8. ghoseb
  9. rkartha
  10. prateekgupta
  11. ajinkyaforyou
  12. gauravsaha
  13. inkv
  14. aparanjape
  15. scepticgeek
  16. meetumeetu
  17. nishantmodak
  18. czaveri
  19. phpcamp
  20. ngkabra

The rank based on followers or twitter grader ranks was not well correlated with the other ranks. In my mind there is a sufficient rationale to question the effectiveness of both followers count or twitter grader as an ability to reach or influence or engage with others, even though twitter grader grade is slightly better than a folliower count. Thats why the other ranks turned out sufficiently differently ?

Ranked by Klout

  1. brajeshwar
  2. scepticgeek
  3. gauravsaha
  4. ichaitanya
  5. sahilk
  6. indianguru
  7. irohan
  8. rkartha
  9. phpcamp
  10. dnene
  11. ghoseb
  12. ngkabra
  13. prateekgupta
  14. d7y
  15. trakin
  16. aparanjape
  17. adityab
  18. punetech
  19. inkv
  20. nishantmodak

To my lay reading this had a stronger emphasis on people who engaged with others, were conversational and had a high update count as well.

Ranking by Twitter Lists

  1. sandygautam
  2. indianguru
  3. scepticgeek
  4. dnene
  5. brajeshwar
  6. phpcamp
  7. ghoseb
  8. adityab
  9. inisa
  10. rkartha
  11. aparanjape
  12. gauravsaha
  13. prateekgupta
  14. meetumeetu
  15. punetech
  16. ngkabra
  17. trakin
  18. freemanindia
  19. aaruc
  20. rush_me

To me this reflected not the spread of the following as much as the strength of the following. Notice how @sandygautam who very tightly focuses on psychology and is well respected twitterer in that area moves to the top (in a rather dominating way I might add)

Rank using Twitter Rank computed by Twitter-Friends

  1. scepticgeek
  2. ghoseb
  3. prateekgupta
  4. gauravsaha
  5. aaruc
  6. dnene
  7. rkartha
  8. adityab
  9. aparanjape
  10. sandygautam
  11. trakin
  12. d7y
  13. meetumeetu
  14. irohan
  15. aditto
  16. clickonf5
  17. rush_me
  18. sahilk
  19. punetech
  20. brajeshwar

This is an interesting metric and while I couldn’t help clearly identify what drove this, would be certainly willing to lend a ear if you want to come up with a suggested rationale.

So the final 20 pune power twitterers based on a composite of the 3 metrics, which in my perception is not terribly different than a list that I would come up with using my gut feel (though perhaps with different rankings) is … drumroll … drumroll …

Pune power twitterers

  1. scepticgeek
  2. gauravsaha
  3. ghoseb
  4. dnene
  5. rkartha
  6. brajeshwar
  7. prateekgupta
  8. indianguru
  9. adityab
  10. aparanjape
  11. sandygautam
  12. phpcamp
  13. trakin
  14. sahilk
  15. d7y
  16. irohan
  17. ngkabra
  18. punetech
  19. meetumeetu
  20. ichaitanya

Note : All the computations results are visible in the attached PDF. Also in a few case klout ratings or twitter friends rankings were not available. In such cases I have applied a klout rating of 0 and twitter friend ranking of 999999. Obviously it reduces the probability of such handles appearing in the overall rankings substantially – but there was no other reasonable option I could think of.

Disclaimer : At the end I am certain there can be a number of views on how such an exercise could be conducted. There might even be some complaints. Being aware of that, I list results of what I believe to be a “fair” exercise. Whether it is a “just” exercise is left to the reader. Also be aware that I have two of my twitter handles in the list above. You may choose to believe my assurance that I did not tweak the logic based on a first pass of results – the logic I decided to apply was not changed once the results were visible.

(Comments are closed on this article. Please comment at the original article.)

Reblog this post [with Zemanta]