Numbers can help us understand what’s happening. But they also can misrepresent the facts or confuse us about what they mean. Most people, though, tend to believe numbers as valid and objective truths, without questioning what they really represent.

Last week, I focused on the circulation figures for the Star-Advertiser. Oahu Publications has been marketing it as the 12th largest newspaper in the country. Yet its subscription numbers, which establish that ranking, include both digital and print customers; and they include the circulation of many “branded” publications that I doubt anyone outside the industry-oriented tallying offices would consider to be “the Star-Advertiser.”

The overall circulation number, though, is really important to Oahu Publications, because it is used to establish how much the newspaper charges for ads. So, as is true with all for-profit media companies, the more that number can be inflated, the more money Oahu Publications can make.

As circulation numbers become more cryptic and convoluted, though, a new coin of the ad-buying realm has been rising: The social-media audience. This audience, by comparison, seems easy to count and make concrete.

A media company just creates a clearly branded Twitter account, such as @StarAdvertiser, for example, and it starts amassing “followers.” When I checked last week, that Star-Advertiser Twitter account had 47,864 followers (compared to its listed “circulation” number of 285,680 and its “total readership” number advertised as 537,669).

The specificity in those numbers seems sure-footed and meaningful. At that size for Twitter, for example, you almost can imagine filling Aloha stadium. For an advertiser, that might seem like a good-sized audience to try to reach. So one might quickly assume that these are real and valuable customers; and even more of them would be better, right?

But I will argue that social media follower totals, and the like, are highly inflated and susceptible to unethical manipulation. More followers, from that perspective, doesn’t necessarily mean more value (or even more real people), as I will show in a comparison of local media sources, after a bit of additional context.

Just What Is A Follower?

Data analysis has been important in journalism for decades, but innovations in digital media have led to the rise of the Quants, data analysts who can parse statistics and algorithms with ease. They offer a welcome underpinning of in-depth numerical analysis to journalism. But they also carry a great potential for oversimplification, and for skewing findings that often are deeply embedded in translating human behavior into numbers.

The Nate Silver-led startup FiveThirtyEight has proven a positive role model for this type of reporting. Recently, it tried to determine which politicians had the most fake followers on Twitter. The loser: Donald Trump.

Numbers are rhetorical constructs, not objective facts. Choosing what to count and what not to count makes a numerical representation, such as a number of fake followers, a subjective, rhetorical label, not an objective truth. So what I’m lobbying for here is for you to look past the numerical labels and think about what they actually mean.

Numbers are rhetorical constructs, not objective facts.

What might seem simple (determining whether a follower is fake) becomes quite complicated (because of rigging of the systems), especially on a scale of millions of followers, which pragmatically requires machine intervention (a computer to do the work).

We can see clearly how many followers Trump has, in the raw sense of how many other accounts are connected to his Twitter account (roughly 8.5 million). But we understand far less clearly what those connections indicate, how they should be valued, or even, conceptually, what a Twitter connection means. Knowing these things should inform our understanding of these numbers.

I’m focusing on Twitter, because journalists have adopted that system en masse, but other social media systems and agents within those systems can be analyzed in similar ways, whether the social network calls them friends, comrades or followers.

The Fantasy Of Social Media Audiences 

The first aspect to understand about a follower is that there is essentially no cost and no maintenance to be or to have one. With so little friction in this system, fake-follower farms have formed throughout the world, in which workers create countless fake accounts used to boost the appearance of a large audience. Web tools have been created to identify fake followers, but that isn’t an easy algorithm to design.  

Each social media site has its own culture and quirks. But these ruses all work in roughly the same manner, focused on broad appearances. Instagram conducted a major crackdown in 2014 that wiped millions of fake followers from its system. In that process, hip-hop artist Mase’s followers dropped from 1.6 million to 100,000 (he deleted his account afterward, apparently from embarrassment) and Justin Bieber lost 3.5 million “fans” overnight.

Hip-hop artist Mase’s Instagram followers dropped from 1.6 million to 100,000. He deleted his account, apparently from embarrassment.

Despite Twitter’s own explicit rejection of them, the social-media company acknowledged in the filing for its 2013 public offering that about 5 percent of the audience was fake. You usually can spot them by their “egg” avatars, low number of followers and few tweets (or repetitive spam tweets).

As Carl Bialik noted in his piece about the Trump fakes, most accounts unwittingly attract these types of followers through bots designed to latch onto your channel and spread their spam. But a lot of people actually buy them, too, very cheaply (as in 10,000 followers for $40), from companies such as BuyCheapFollowersFast.com, Sozialy.com and SocialSpringer.com.

To see how Hawaii’s media companies with 20,000 or more Twitter followers were competing in this social space, I decided to collect data from their primary accounts. Then I ran them through two “fake follower” web tools (TwitterAudit and StatusPeople). Both are available to the general public (so you can run these tests yourself; other tools also are available, for fees).

Here is what I found:

As you can see from those comparisons, different algorithms view and characterize these accounts differently. The process of comparison is transparent — as in you, too, can plug account names into the web tool and see the results. Yet only the company that made the tool knows exactly the secret sauce of the algorithm, so we should take these results as indicators of what might be happening, rather than as incontrovertible evidence of conduct.

From that perspective, what I see here is that if you combine fake accounts with inactive accounts, you generally can halve the follower number, or more, to get a ballpark sense of real people connected to these accounts. From there, you can see that some companies have enormous numbers of suspicious accounts, such as Hawaii News Now, which has roughly 50,000 fakes, or half of its Twitter audience, and, adding together fake and inactive accounts, 79 percent classified as worthless.

A team of Italian and French scholars who studied this issue in depth – Stefano Cresci, Roberto Di Pietro, Marinella Petrocchi, Angelo Spognardi and Maurizio Tesconi – revealed some of the strategies for behind-the-scenes evaluations of followers and how you, too, might spot a fake. For example, a real person probably has a name on the account that’s not gibberish and a unique image; that account also probably lists a home-city location, has a biography inscribed on it, belongs to a list, uses punctuation and hashtags, has more than 30 followers and more than 50 tweets, etc.

A quick look at the most-recent followers for the low performers on our charts show many suspicious accounts, indicated by an army of “egg” people (those using the default egg icon) following many others but with few followers in return. Theoretically, these could be real people, using Twitter in an unusual way. Or they could be a bunch of fakes. The latter is more probable.

Some of the bigger questions I’m trying to raise here, though, are: What do we really want out of our media audiences? Do we want an informed citizenry engaged in issues of community concern, given agency and power through social media to react and respond in meaningful ways to local, regional and national news? Do we want gigantic raw numbers, used to sell more ads for more profit?

Are 100 great followers, who thoughtfully enrich community discussions, worth as much (or more) than the 10,000 fake followers one can buy for $40?

In other words, when you encounter these types of audience numbers in the media, I recommend you thoroughly consider at least these key ideas: Who is counting? What exactly are they counting? And for what specific purpose are they counting?