Introducing Insider score

The score indicates whether other members of a community would consider a given account ‘one-of-us’ or not.

This addition is an important complement to Attention scores. Together they offer a more nuanced view into each group. There are many profiles that receive lots of attention, but if you ask around you would find out that they are not considered part of the community. 

Take Donald Trump as an example (when he was still president). He would get a high Attention score in many communities, including those that hated his guts. They still paid attention to him and that is what the score indicates. As long as people pay attention to you you have an ability to influence them in one way or another – even if they do the opposite of what you say. However, this leaves out a big part of the picture. It does not tell you whether somebody belongs to that community or not. That is where the Insider score comes in. 

How Insider score works 

Your Insider score is going to be high if people who pay a lot of attention to other accounts with high scores pay attention to you as well. Of course, it is more nuanced than that, but this is the most intuitive explanation we could come up with. 

The scores are scaled from 0 to 1. The closer to 1, the more of an insider a given account is. We have updated visualizations on the website so that it is more intuitive. The size of the colored circle indicates Attention score and the position of the colored circle indicates Insider score. The more the two circles are aligned, the higher the Insider score is. 

It is still in Beta! 

Insider score is currently in Beta. We have been working on this release for close to a year. Hundreds of hours of research went into this. However, this work is not nearly done. 

For example, our algorithm gives a lower Insider score to large accounts, such as Elon Musk or Vitalik Buterin. Clearly, Musk should have a high Insider score in the Tesla community and Buterin in Ethereum’s. We are working on this issue. 

We have released the scores because we think that they got to the point where they can be useful. Another reason is that we need your feedback to improve them further. We are seeking your input particularly in two areas (but all feedback is welcome!) 

  • Where are the scores incorrect? I.e. which accounts, based on your insider knowledge of a given community, should have higher or lower scores and why? The more of this kind of input we have, the more precise experiments we can design. 

  • What could be a better way to present these scores? I.e. We know that scores expressed on a scale from 0-1 are not the easiest to read. Especially, for somebody who is not deeply familiar with a particular group. We are looking for better ways to present this information. If you have any suggestions – we are all ears! 

You can send your feedback to or DM us on Twitter.