Blog

Openinfluence

Openinfluence is an open-metric developed at Paradigmalabs and tries to define the relevance of each user in Twitter. It is open because you can see the formula and contribute to improve it. You can see the formula in the picture below:

As you can see, the formula has two main components “Popularity” and “Influence“. Popularity is related to static properties of your social network. It’s some kind of “potential influence”, the beforehand capability of getting your tweets spread.Influence is related to the propagation and repercussion of each of your tweets, the effective reach of your messages.

We have applied successfully this metric in several analysis, e.g.: during the Andalusian elections campaign or UX Spain Conference.

Currently we can represent this formula with the next plot:

We are involved in trying to improve this metric, because the two main parts have the same weight in the formula. However, is this metric more related to Influence? Is the formula below better?

We have tested Openinfluence with the next dataset. In the picture below, you can see the number of followers degree of each user in the sample (in logarithmic scale):

The correlation between Popularity and Influence (dataset) shows that the main stream of people has more or less the same Popularity and Influence. By means of the structure of this formula, some users have 0 of influence and n>0 popularity however they have not null relevance.

Suggest us your point of view !! We are expecting to improve it!!

About the authors

6 comentarios

  1. Rianne says:

    Very interesting post. I was just wondering, what do you mean by Retweets audience?

    • Roberto Maestre Martínez says:

      Hi Rianne,
      Retweets audience means the number of potential readers, e.g.: if one person tweet something and another three persons with (125, 300 and 400 followers) retweet it, the Retweets Audience will be (125+300+400)/3 = 275. We apply this formula to the whole set of retweets belonging to a user.

      Perhaps, we can improve it.
      Best regards.

      • Rianne says:

        Thanks for the reply. I’ve got some more questions. Why do you average the retweets audience? You should account for the overlap in followers of the persons that retweet, but I’m not sure that’s why your averaging it.
        The retweet audience depends on who is retweeting your tweet, but how do you calculate it for the complete set of retweets in one week? If one person retweets 2 of your tweets, do you count his retweet audience twice?

        Kind regards,
        Rianne

        • Roberto Maestre Martínez says:

          We are trying to keep a balance between functionality and accuracy.

          We are limited by twitter API constraints regarding number of queries per hour, and we would need a lot of queries just to calculate the overlap for each pair of users; worse than that, following actions are changing all the time, so we should repeat those queries continuosly to get this data in real time. Therefore, in order to limit the number of queries required, we calculate an approximation without overlaps.

          Besides the problem with duplicated followers, the more retweets you retrieve, the more queries you have to use, and even if you didn’t have the queries rate limitation, there’s another constraint in twitter API, you only get up to 1500 retweets. Therefore, in order to limit the number of queries and being able to get an aproximation for those accounts with huge amounts of retweets we retrieve a sample of the whole retweets, calculate a simple average of retweets audience per retweet and the rate of retweets per time unit (Number of retweets in the time frame between the first retweet in the sample and the last one) and we extrapolate the data to one week.

          Thank you for your suggests and comments,
          best regards.

          • Rianne says:

            Thanks for your reply. I am thinking about a similar measure, and indeed I also had some issues with the Twitter API rate limits. The retweet audience is the only component I didn’t use, that’s why I was curious how you calculate it. I would like to test and compare some of these influence measure but I am not sure how to get some kind of ground truth. Did you test your measure in any way?

          • Roberto Maestre Martínez says:

            We tested this metric in two kinds of sceneries: global and small events. We have chosen this two sceneries because in a global one, you manage a huge amount of information and the process is more static, the second one is more dynamic and you manage less and local information.

            A global event like Andalusian parliamentary election of 2012 we obtained the following results: José Antonio Griñan won the campaign, however he obtained the third position in the final ranking of Openinfluence (http://labs.paradigmatecnologico.com/eleccionesAndaluzas2012/), and in small event like UX Spain conference (http://paradigmatecnologico.com/uxspain/).

            Results shows that test this metric is really difficult, sometimes twitter is only a part of the global reality, therefore we try to identify key variables in a theoretical-logic formula to Influence.

Escribe un comentario