Brand Safety By Anthony Svirskis

Why Fake Influencers Don't Stand a Chance

At TRIBE, we’re often asked by brands…  

“How do you ensure influencers have real followers and engagement?”

Many think because anyone can download the TRIBE app we might be more vulnerable. Quite the opposite. Anyone can download the app, but only influencers with the highest quality of audience can submit to a brand.

Read here for an insiders view in to influencer fraud with PerformanceIN

Therefore, we thought it might be valuable breaking down exactly what happens the moment an influencer connects their social account to TRIBE. Obviously we can’t reveal everything, because that wouldn’t be very secure, would it?

THE MOMENT YOU CONNECT

TRIBE scans an influencer’s social account and compares key metrics against normalised data from more than 10,000 influencer accounts. Some of these include:

Sudden spikes in followers
Historical jumps of 1,000, 5,000 or 10,000 followers is obvious.

Unusual follower or following activity
We detect 4 times a day how many accounts an influencer follows and unfollows. The first tell-tale sign for using follower tools and bots.

Follow-vs-Following-Ratio

Time and posting activity since account was opened
Should the account have too few posts for a large following, or sporadically post at odd times, it’s flagged.

Engagement velocity
After posting, likes slowly build then jump by 1,000 for example - another easy detection.

Ratios of comments to likes
We know in great detail how an organic ratio of ‘comments to likes’ should appear. Anything outside of this is immediately identified.

Comment-to-Like-Ratio

Ratios of likes to views
If more people have liked or commented on a video than have actually viewed it, it’s clearly fake engagement.

Likes-Than-Views

Repetitive engagement
We detect repetitive interaction from the same groups of people and scan for bot behaviour from those who think they’re fooling the masses.
Fake-Comments-EngagementAnd more
Most we can’t reveal.

Identifying these abnormalities is easier than you think. One of the exercises we performed early on, was commissioning hackers to use every trick in the book to build fake profiles.

Their findings were handed to our data scientists who built the earliest iteration of an algorithm we call our AVS scan. It stands for Audience Vital Signs and continues to evolve to include dozens of scans. 

The majority of influencers in TRIBE proceed through these checks with flying colours and can use the app immediately. However, we apply Amber and Red flags to any accounts that register results outside of what we’d expect from an authentic influencer account.

AMBER FLAG

Should an influencer account be pushing the boundaries of normal follower and engagement activity, an amber flag is automatically assigned.

In this state, the influencer has app usage however submitted content is held from brands until the review process is complete. Amber flagged accounts enter a manual review queue which goes to one of our TRIBE experts to review and assess within 24-48 hours of connection.

Often there are valid reasons why an authentic influencer raises an amber flag (e.g. spike in followers due to TV appearance or PR) so if these are established we remove the flag and any pending submissions are released to the brand.

RED FLAG

Access denied. When our data shows an influencer account sits well outside the norms of usual activity, then a red flag is placed on the profile and the app is unusable.

Red flags enter a manual checking queue like Amber flags, but more for the health of our algorithm to ensure we’re continually analysing data effectively. The influencer is notified of the block and offered advice on influencer expectations.

AVS-Flag-Account-App

RECURRENT SCANNING

The AVS not only scans an account upon entry, but then 4 times every day across our entire database in the rare event an influencer account enters clean, but records unusual activity after entry.

Every single submission from an influencer to a brand, passes through an online image search which picks up unoriginal or doctored content and rejects it before it reaches a brand’s inbox.

The AVS scan is a sophisticated system that can identify fraudulent behaviour in ways a human cannot.

Equally, our expert staff can then apply a critical lens our algorithm cannot.

Platform and industry integrity has always been paramount to TRIBE and we were proud to have our Data Scientist share his insights within a keynote to a roomful of science marketers at MSIX in 2016.

Nick-Barrington-Data-Scientist-TRIBE

In mid-2017, we released pre-purchase analytics and influencer ratings within our platform so brands could get a critical view into influencer professionalism, content quality and genuine audience influence. These ratings now provide brands and agencies with a tangible window into influencer integrity, should a brand ever be in doubt.Influencer-Rating-Tool-TRIBE-App

HOW EFFECTIVE IS THE AVS?

More than we expected. It turns out the websites that sell fake followers and engagement are embarrassingly unsophisticated. Plus the major social platforms are simply shutting them down.

Every day on average we observe 170 downloads of the TRIBE app, of which only 3 flags are raised. 2 are Red which we block, and the other is Amber, with 1 in 3 Ambers getting cleared upon manual review. 

These numbers indicate that fake influencers are not nearly as big an issue as first thought. You’d agree most fraudulent behaviour is incredibly easy to spot with the naked eye.

Add in a sophisticated scan and for the thousands of brands and agencies using TRIBE, it should be of no concern whatsoever. 

Get-started-for-free