Uncovering How Bots Boost - And Distort? - Engagement

Awareness does not equal action. Results are still being skewed by marketers selling big numbers to their customers. JL

Sapna Maheshwari reports in the New York Times:

16.4% of the followers on Instagram’s top 20 accounts were fraudulent. Many influencers, particularly on Facebook, reach only 2% of their audience.“Though brands are looking for engagement, the actual compensation influencers are getting is still based on the follower number.”Marketers are increasingly skeptical of audience numbers that determine how much money social media stars can command from advertisers. “Most people are not seeing posts even if they’re real people, but many are not real."

In a sunny office in the SoHo neighborhood of Manhattan, Mike Schmidt spends his time ferreting out fake Instagram accounts.

Some are obvious, like the one that had never posted a photo and lacked a profile picture yet followed about 7,500 accounts — the maximum allowed by the social media site. Others are trickier. Mr. Schmidt had to scroll down a little on an account with the name @ailebnoblk before the same stock image of a car showed up three times in a row, a clue that there was no real person behind the profile.

Image

The users who uploaded the same stock image of a car did not respond to messages on Instagram.

“The amount of bot activity that’s happening on these platforms is pretty insane,” Mr. Schmidt said. “Just the amount of new accounts and times these folks are liking and commenting with spam and positive comments and happy-face emojis.”

Dovetale, a four-person software company Mr. Schmidt co-founded in 2016, has devised a range of tactics to identify large numbers of fake accounts that follow popular Instagram personalities. It then packages that information for marketers, who are increasingly skeptical of the audience numbers that often determine how much money social media stars can command from advertisers.

Marketers are flocking to businesses like Dovetale, prompted by revelations like those in a recent investigation by The New York Times that detailed the booming industry of people buying fake followers and fraudulent engagement on Twitter and other social media sites. Some of these fake accounts, in an attempt to seem legitimate, use personal information from real people without their knowledge. That has provoked concern among brands and their agencies, which often rely on metrics like the number of followers an account has when hiring people on YouTube and Instagram to promote their products. These social media stars can often fetch thousands of dollars for one post promoting a product.

We knew this kind of day of reckoning would come,” said Erick Schwab, the co-founder of Sylo, which vets influencers for fraud and aims to assign a numerical score to their content akin to a Nielsen rating. “We’ve gotten tons of brands, agencies, vendors emailing us, who we’ve been having conversations with for a while, but now they’re sort of like, this is being demanded.”

Krishna Subramanian, a founder of Captiv8, which connects brands with influencers, has seen a surge in requests for fraud detection from agencies. “Everyone is definitely scrambling because they don’t want to be held responsible,” he said.

The interest in such firms reflects how easy it is to fake popularity on platforms like Instagram, where bots seem to run unchecked even on accounts where people have not paid for them. While many advertisers have become aware of this, and tried to place more emphasis on content quality or favorable comments, follower numbers still tend to loom large.

“Even though brands are looking for engagement more, the actual pay and compensation that influencers are getting is still based on the follower number,” said Alivia Latimer, a photographer with about 102,000 Instagram followers. Ms. Latimer, who has worked with brands including Lush Cosmetics and Hollister, said that she charged about $1,200 for a branded post. She added that she knew people with two million followers who charge $40,000 per post.

That means new kinds of detective work are needed for brands that still want the endorsements of the young and hip online. Dovetale said it uses more than 50 metrics to analyze the Instagram followers of popular accounts, including the language in the bios, the rate at which they hit “like” and “follow,” and their country of origin. (An influencer with a high number of followers from Turkey, Brazil and China, for instance, can raise red flags for Dovetale, which has frequently seen fake followers come from those countries.)

The clues can be complex. Dovetale flagged one account that claimed to be someone named Meg Cragle because it was part of a group of profiles that had made one or two unrelated posts and contained similarly worded bios of exactly 99 characters that ended with ellipses. The discovery was reinforced by a Google search for phrases in the account’s bio like “award-winning bacon fanatic,” which matched the terms in a now-deleted Twitter bio generator online.

Dovetale acknowledged that its methods were not foolproof, but they are valuable enough in a murky landscape that the agency 360i said it was now unlikely to hire influencers for campaigns if Dovetale’s database said more than 2 to 3 percent of their followers were bots. Dovetale said that, on average, 16.4 percent of the followers on Instagram’s top 20 accounts were fraudulent.

Sylo, which requires influencers to share access to their public and private post statistics, said it had rejected 77 percent of influencers who have tried to register on its platform after their accounts showed issues like abnormal spikes in engagement on posts or a large number of generic, emoji-laden comments that bots are known for.

“In the absence of direct pressure on the platforms, it’s a way for advertisers to take more control of their own spend and not be at the mercy of the platforms themselves,” said Jeff Semones, head of social media at MediaCom, which has recommended Sylo to clients. While he said Instagram had removed many accounts that were flagrantly violating its terms and conditions, “a lot of one-off activity goes unchecked.”

A spokeswoman for Instagram said that the platform’s “internal estimates show that spam accounts make up a small fraction” of Instagram’s 800 million monthly users.

Some believe that the new awareness around bots highlights the misguided expectations that marketers have for how many people they can reach through influencers.

Changes to algorithms on Facebook and Instagram have significantly reduced the number of people who will see a person’s posts without paid promotion. And unless advertisers are paying Instagram for the information, they typically have to rely on screenshots from influencers for data on how many people saw a post. Influencers like Ms. Latimer said that even then, not all brands request those screenshots.

Tyler Stark, director of marketing at Traeger Grills, said that many influencers, particularly on Facebook, reach only 2 percent of their audience. That has made smaller influencers more appealing and put a focus on engagement, he said, with the thought that a post with a large number of likes and comments will end up in the feeds of more people.

Still, Bob Gilbreath, chief executive of Ahalogy, a marketing technology company in Cincinnati, said that he recently heard a major retailer recommend that brands work only with influencers who have at least 200,000 followers.

“Most brands would say that a follower is someone who’s definitely going to see the post,” Mr. Gilbreath said. “Not only are most people not seeing the posts even if they’re real people, but many of them are not real people.”

Soapbox Influencer Marketing, a company that connects brands with influencers, features a case study on its website for a new hummus from Bush’s Beans. The site claims that the campaign, which featured 519 influencer posts, resulted in a whopping 891 million impressions — nearly triple the American population.

When asked about the figure, Beth Stephens, the company’s president, said it was “incorrectly listed on our website and needs to be corrected.” She added that Bush’s Beans also worked with “a number of other influence and social service providers,” making it difficult for Soapbox to directly assess the figure. Bush’s Beans didn’t return requests for comment.

Others tout the “total reach” of fashion and food influencers on their websites. In other types of media and Instagram’s own business analytics, “reach” refers to the number of different people who actually saw an ad. But in influencer marketing, it often refers to an aggregation of followers across platforms and can even include monthly unique visitors to blogs.

Mr. Gilbreath is still a proponent of the work that comes out of social media mavens, but has instead focused on hiring creative “micro” influencers with 10,000 or more followers. The firm pays to promote their recipes or tips and tracks subsequent visits to their websites, he said, adding that the data is otherwise too unreliable.

Corey Martin, head of influencer marketing at 360i, said that the agency was increasingly paying Instagram to promote posts from influencers, which allows the agency to “track and trace consumer behavior in a way we previously hadn’t.”

Some hope for more action from Instagram, especially since it plans to restrict access to some of its data later this year, which may impede the work of bot-hunters like Dovetale.

“It will be unfair until Instagram really just cleans out the bots and the lurkers,” said Caitlin Sakdalan, who runs the account @BeFatBeHappy. “There’s so many different programs that people use and algorithms and we can all say this and all say that, but until Instagram itself kind of does it for the community, I think it will be an unfair playing field.”