Mavrck defines influencer fraud as any act that artificially inflates reach, engagement, clicks, or any other performance metrics relevant to influencer marketing.
Our fraud detection methodology is currently limited to Instagram, and analyzes the characteristics of an influencer’s account as well as the accounts that follow or engage with the influencer.
Our first pass algorithm uses statistical methods to identify influencers who have an abnormal follower count relative to their engagement levels. We also assess additional metrics to identify outlier ratios. We call this analysis our ‘shallow’ fraud detection process.
Additionally, we have established a ‘deeper’ machine learning algorithm that decides if a given instagram account is from a real person, used in an authentic manner, or if the account is operated by a ‘bot’ for the purposes of creating artificial followers, likes, and comments.
We then apply this algorithm to a statistically significant sample of accounts drawn from an influencer’s followers as well as accounts that like or comment on their posts.
Based on this analysis we then flag the influencer as being High, Medium, or Low risk of having purchased followers or engagements. The thresholds for these categories are being continuously refined as we collect more data, but are meant to indicate risk rather than certainty.
A good way to consider these risk levels is as such:
- Low: We detected either no fake accounts in our sample, or the volume was such that it could be consider coincidental. Fake accounts have been known to follow and engage influencers randomly to boost their credibility.
- Medium: We detected a significant amount of fake accounts within our sample, and it is likely that some fraudulent behavior is occurring.
- High: Our sample contains a very high percentage of fake accounts, and it is very likely that this influencer is paying for followers or likes.
As of June 18th 2018, our Fraud Analysis capability is available to a limited number of customers. We will be rolling out the functionality to a broader set of customers over the summer.