There’s been a lot of talk lately about failures of popular influence measures in social media.Over the past year it’s a topic we’re thought a lot about here at ManageFlitter. One of our most requested features has been to include influence ranking in our filters to see who the most and least influential people you follow are. We continue to reflect on the best way to tackle this.
Up to this point we’ve resisted, because we have always thought that the existing approaches just don’t feel right. It’s common place to compare the influence scores of two people and find that they make no sense.
I believe that the explanation for this problem is straight forward.
Influence can’t be a ranked.
One of the great powers of social media is that within hours someone with only 10s of followers can become know around the world after a single tweet becomes retweeted an amplified through the entire Twitter network.
When you send out a Tweet, all kinds of factors change how many people it will reach, from the content of the tweet, through to the time of day, through to current events or even who’s randomly viewing their stream at the time.
Past influence is not a good indicator for future influence. Attempting to consolidate influence to a single 1-100 ranking is nonsensical. It would be like trying to rank the most “important” people in the world. Everyone would have their own personal understanding of what importance is and would disagree with your one global ranking. Even worse is that attempting to produce this ranking with algorithms removes a lot of sanity checks from the process. Ultimately a global influence rank is just an arbitrary popularity contest that is ripe to be gamed and has little meaning in the real world.
And yet… we can’t throw the idea of influence measures out of the door.
The data exposed in the followers number on Twitter or Facebook is not sufficient to understand an account’s real social medal impact. How many people in that follower count are really active? How many times is a tweet read? There’s a need for deeper analysis to give people the tools to inspect their account’s performance.
We believe that a credible influence measure should have the following features;
1.Expose understandable, meaningful metrics. Don’t hide behind complex algorithms. Don’t cram all the metrics together into a meaningless global ranking.
2. Be transparent about algorithms. Explain in plain English how results are calculated. Allow people to dig into their own account data and see how every individual data point influences a metric.
3. Understand people’s need for privacy. Allow a global opt-out. If you don’t want to have your metrics calculated, then we’ll never process data about your account. Even if your existing data was public, if you want to opt-out of our calculations, then you should be able to.
We would be interested in your thoughts. What do you think people really care about when they look at influence scores?