Bursting of the social-media bubble

February 13, 2018

(This article was originally published at Big Data, Plainly Spoken (aka Numbers Rule Your World), and syndicated at StatsBlogs.)

Nytimes_followerfactoryBitcoin, stock market - the bubbles are bursting left and right. The next bubble getting ready to burst is the social media bubble.

If you are following this blog, it will come as no surprise that a lot of "big data" are fake, and many data analysts, sadly, are spending our time crunching fake numbers. Fake data come in all shapes and sizes.

There is an ongoing battle of quantity and quality in the data business. There are lots of "easy" data to collect but easy data are almost always bad data; good data, by contrast, are typically costly or cumbersome to obtain. "Big data" is a reflection of abundance - it is the elevation of quantity over quality.

Take social media. It is conceptually a wonderful thing for researchers because all of a sudden, people are sharing their data, and exposing their social networks. Facebook, Twitter, Snapchat, Instagram, etc. collect mountains of data on their users.

What are they collecting? Followers, views, plays, clicks, messages, images, comments, likes, etc. These are all counts of "interactions." Because these items are easy to compute (there is no statistical thinking involved at all), they quickly become the standard metrics for social media. Then, these social media companies discover that businesses have to make money, and given that few users are willing to pay anything to use their services, they all eventually turn into advertising companies. They need something to sell to advertisers.

At first, advertisers were deterred because the new social media do not have proper performance metrics. There is no Nielsen rating. Eventually, Facebook solved this problem by selling them the metrics mentioned above. They argue that those metrics are better than Nielsen's because their dataset is larger. Nielsen data are collected by recruiting panelists who allow their viewing to be tracked - and for some of their products, the panelists also annotate the data with qualitative information.

Lots of advertising dollars have shifted to social media. A relatively recent phenomenon is the "booming economy of online influence," as the New York Times calls it. Marketers like to spend money to hire "influencers" to push their products. Influencers in the old world are celebrities, movies stars, star athletes, etc. In the social media world, an influencer is anyone who has amassed over X number of followers. So there are those who maintain Instagram or Twitter accounts just so as to earn sponsorship money from advertisers who thirst after their collection of followers.

But eventually, advertisers are waking up to the fact that the "big data" contain lots of little scams. They purchased quantity at the expense of quality.

For example, someone may have 43,458 followers - but what kind of followers are they? Loyal fans? Casual acquaintances? Purchased fake followers? Bots? Turns out: All of the above.


The recent New York TImes expose on the company called Devumi is required reading for those interested in following this story.

Devumi "sells Twitter followers and retweets to celebrities, businesses and anyone who wants to appear more popular or exert influence online."

How it works: Devumi maintains over 3.5 million "fake" bot-operated accounts. Customers pay Devumi for certain target number of followers. Those customers see their follower count go up as Devumi directs its bots to follow their accounts. And it's not just Twitter followers. Devumi customers buy YouTube views, SoundCloud plays, Linkedin endorsements, etc.

The bots come in different varieties. "High-quality" bots use stolen identities based on real people: you just pirate the data from a real account and make a few changes. "Low-quality" bots make clearly fraudulent accounts. Customers are typically served by high-quality bots initially until the period of review is over and then when they are not looking, their accounts are stuffed with automatically generated fake followers.

The social media companies downplay the extent of the "follower factory". But anyone can search for "how to buy instagram followers" and get lots of interesting links back.

The Times came up with a list of celebrities and famous and not-so-famous people who are Devumi customers. Their explanations sound like Olympics dopers confronted with their steroids. "Everyone does it", "I only tasted it once (did not swallow)", "I thought that was legit", "someone else did it without letting me know", and the likes. That section is very delicious. It's as if everyone is doing it. The Times even discloses that it owns an "influencer agency," and, you guessed it, their own agency vigorously checks the credentials of all the contractors it hires, just like, hmm, the other agencies who also strenuously deny that they break any rules of ethics.

Then, the article turns to the founder of Devumi. We learn that he fakes the details on his Linkedin profile. We also learn that Devumi doesn't even program the bots - there is a market to buy them in bulk.





Please comment on the article here: Big Data, Plainly Spoken (aka Numbers Rule Your World)

Tags: , , , , , , ,