We released Fohr Card's Follower Health Tool a few months ago and for the first time we gave brands and influencers insight into the health of their following. We spent half a year building the tool and devising a way to differentiate between what was real and what what not. There were two ways we could have gone about trying to determine if someone's following was authentic.
1. Daily follower growth: Tools like Social Blade look at how many followers an influencer gains or loses in a day and from there people make their own decision on if this person is buying followers or not. We looked at this but didn't like it as a method because of the following issues.
- We wouldn't have historical information on a following, so if someone bought 50,000 followers then signed up for Fohr Card and didn't buy another follower we wouldn't be able to detect it.
- These spikes could be due to a press mention, giveaway or some other occurrence that would result in big spikes in follower growth. Mislabeling this growth as someone purchasing followers would be a mistake that could potentially ruin an influencer's career.
2. Follower Type: Because we have access to an influencer's followers through the Instagram API we were able to actually look at the profiles of every person who follows each of our influencers. From there we build a set of criteria to "score" those followers to try and understand how "healthy" they were in terms of authenticity. What we were trying to determine was is this a real person who is using Instagram in a normal, natural way and do we think this person followed the influencer because they wanted to see their content, or were they purchased or sent their through some service.
To determine this we look at a number of factors:
- Handle: What is the follower's handle? Is it Instafollowersforfree or jamesnord, does it look and feel like a real handle?
- Name: Do they have a full name in their profile or again is it gibberish or something that looks like a bot
- Bio: Do they have a bio? What does that bio say? Does it say 33 year old Mom, Granola enthusiast or does it say "Like and I'll Follow" or does it simply not exist?
- Follower to following ratio: A lot of the "lurkers" will have strange follower to following ratios. Normal people don't follow thousands of accounts, so if we see someone with 500 followers who follows 7,000 people we will give them a worse score than someone with 300 followers following 500 people.
- Number of photos: Instagram will not allow us to see the last time someone posted, but we can look at number of photos, so if an account has 5 photos but have 1,000 followers it raises a red flag, if they have no photos it raises an eyebrow. If they have hundreds or thousands of photos their scores will be better.
All of these factors will either give a follower a +1 or -2 or +3 etc dependent on the algorithm we have built and a follower will be given a final score and that score will put them in the bot (-8 to -3) lurker (-2 to 0) or active (1-8).
To determine which influencers were Verified Authentic we computed thousands of Influencer Health Scores and looked at the averages for each category (bots, lurkers, active) at a number of following levels:
We then took influencers who were a certain amount ABOVE that average and verified their accounts. It's important to note a few things:
- Because this is proprietary technology we cannot release the exact % for each category to get authenticity.
- Because this is based on math the % to reach authenticity changes with every new influencer who signs up. This means you if you are right on the line you may lose or gain an authenticity badge as the levels change.
- We are not saying that those influencers who have authenticity badges have never bought a single follower, that would be impossible to know. What we have done is come up with a system that we think does a good job at verifying the health of an influencer's following.
- This system will continue to improve as we learn more and get more data. We are consistently buying followers from a number of sources to see the types of accounts they send and we are tweaking the algorithm accordingly.