Credibility Score

Welcome to a world of transparency (and of no more not-so-viable “fraud detection hacks”).

Warning. Do not try this at home. The following feature was made possible by great data scientists, performed under the supervision of Yoda & required hundreds of millions of accounts.

Let us explain further...

Google “how to detect influencer fake followers / traffic / likes / comments” - and you’ll find a ton of platforms, tools, articles & third party services telling you how to untangle this mystery. What they’ll all say is basically;

Calculate the engagement vs number of followers (100,000 followers with 3,000 likes = good. 100,000 followers with 300 likes = bad.)
Look at the accounts following and / or interacting with the influencer (are they named qwerty123456 / jasdeed9ewfds etc)
Browse the comments (one-by-one) to see if there’s suspicious activity (i.e does 90% of the comments look like “Nice! Check out my profile!”).

These options are, for two reasons, no longer viable if you’re really keen on transparency.

One. You’ll faint from boredom going over all the accounts and comments.

Two. Buying followers & having third party services boost followers & engagement are getting seriously advanced. The account names looks somewhat “regular”, they have a biography on their profile, uploads pictures, have more followers than they themselves follow and so on.

Sample data

Certainly a real person behind the account. However, following the charts & our software, this account was analyzed along with thousands more, also interacting with the influencer we checked out.

This very account engaged with 724 other posts on the same day, as well as commenting on 227 of them.

Engaging with 15-20 posts per day is the average..


Analyzing tens of millions of accounts to track user patterns and connect the dots

By analyzing a big bundle of accounts, and their activity, generates a percentage of how much of the influencers total views & engagements that are derived from accounts with very suspicious activity.


From big data to market application

How does all of this translate into something that could help regulate the market and boost the confidence of marketing directors & agencies across the globe to further invest in influencer marketing?


For simple maths sake, let’s assume you’re paying $1,000 for a sponsored post to an influencer with 100,000 followers.

The account is not super-niche, have an average engagement rate (3%) and post high quality content on a regular basis.

Both parties agree that it’s a fair commission, as it results in a $10 CPM.


Using the Collabs software, we analyze the followers of the influencer and in specific, their activity patterns. Circa 9.8% of all followers gets a red flag. Those accounts all look rather ordinary on the surface, but their activities are quite the opposite, posting hundreds of comments & likes in very short time-frames.

It’s obvious that almost 10% of the influencers views & engagements are derived from bot-activity, something that then would logically bring the commission down to $900.


The reason is rather simple. A percentage of the views were never seen, liked or commented on by an actual human being.

Now, this does not necessarily mean that the influencer you just agreed to pay $1,000 tried to fool you. The bot-activity might have been something they actually didn’t know anything about, or even considered. Nonetheless, it should not be paid for as regular real interactions.

We add a layer of transparency to this problem and detect fake traffic.