Understanding Twitter’s Recent ‘Followers’ Changes, How They’ve Been Misrepresented And Why Ariana Grande And Demi Lovato’s Twitter Ranks Rose While Britney Spears And P!nk’s Fell

On July 12th and 13th Twitter ceased including locked accounts in the follower counts of the accounts they follow. Locked accounts – which are different from spam and bot accounts – were already not counted in Twitter’s total user metrics, but until this week they still inflated individual follower counts. The change was most noticeable in the profiles with the most followers. The followers count of Katy Perry for instance, the most followed account on the platform, decreased by 2.8 million in one day; runner-up Justin Bieber was down 2.7 million followers. Those numbers by themselves however are meaningless, for reasons explained below

Twitter’s “legal, policy and trust & safety lead” Vijaya Gadde explained the changes in a July 11th post on blog.twitter.com titled “Confidence in follower counts.” “Follower counts are a visible feature, and we want everyone to have confidence that the numbers are meaningful and accurate,” she writes. She goes on to explain that “locked accounts” will be removed this week “across profiles globally,” how accounts get “locked,” how these accounts are “different from spam or bots” and that most twitter accounts will see a drop in their follower count (the more followers an account has, the more it may lose). Referring to the accounts that will no longer be counted in follower counts she says “in most cases, these accounts were created by real people but we cannot confirm that the original person who opened the account still has control and access to it.” Despite Gadde’s clear explanation, the changes are commonly misunderstood as a “crackdown on fake followers,” in large part due to misleading, ambiguous or downright inaccurate coverage by numerous influential news sites, many of which conflate it with Twitter’s separate effort in recent months to increase account suspensions.

Gadde’s claim that the bulk of profiles no longer counting among “followers” as the new policy sets in are not “spam or bots” and were mostly “created by real people” was substantiated by The Future Heart’s independent analysis of the changes. According to Gadde’s post, accounts can be locked if twitter detects “sudden changes in account behavior” or if “email and password combinations from other services posted online” duplicate those of a twitter account. The latter is assumed a potential security risk, the former is assumed a potential hack – though, as many twitter users can attest, it could also be nothing more than an error by twitter’s surveillance programs. For most active twitter users this is easily rectified. Twitter requires the accounts to change their passwords, then unlocks the account when they do. Inactive users however may never even know their twitter accounts are locked.

It’s these inactive accounts that were
at the center of this week’s changes
– not spam bots, not fake followers –
but legitimate albeit inactive accounts that
for one reason or another were locked by twitter.

The longer an account is dormant, the higher the odds it will be locked. Some are accounts that the owner lost access to (ie forgot their email/password combination); some were created by people who have since died, some are accounts that required the owner to provide their phone number to regain access and the owner refused; some were legitimately hacked, others are simply accounts of people who have lost interest in twitter and stopped using it years ago. More specific details have not been revealed, though independent analysis suggests that this week Twitter mostly focused on accounts that have not been active since 2014 or before to no longer count among profiles’ followers. One such account is the original twitter account of psychedelic rock band Stardeath and White Dwarfs. As you can see by visiting its profile, there is nothing “suspicious” about it: no sudden onslaught of tweets, follows or replies; no misleading links; nothing that suggests a large number of accounts blocked them, no spamming a hashtag – nothing. (Perhaps the lock happened somewhere in the process of the band changing their profile picture?) Whatever the basis, in August 2014 – for reasons never explained to the band – Stardeath could no longer access their twitter account. It’s stayed unaccessed for the past four years as the band abandoned it and created a new profile in its place:

Stardeath’s original twitter profile and millions of similarly locked accounts demonstrate how the changes that actually took place this week – misleadingly refereed to in the media as a “purge” – are generally misunderstood and misrepresented:

  • Millions of locked accounts like Stardeath’s that factored in this so-called “purge” remain as they did before this week, as well as they were before they were locked. These accounts were not deleted, removed or altered in any way. If they are public accounts they can still be viewed like any other account. They are still as they were, they simply no longer count among other account’s followers.
  • Active users can still engage (retweet, like, reply, tag, follow, unfollow, etc) with the locked accounts, such as Stardeath’s (though of course the locked account can’t reply):

  • The following and follower counts of the locked account are the same as they were at the time of the account being locked – plus/minus any accounts that followed/unfollowed since – and the following and follower counts are still listed on the locked account’s profile. See Stardeath’s for example.
  • All of the accounts that the locked accounts were following before they were locked are still listed in the “following” tab of their page (as can be seen from looking at Stardeath’s page for example). Although the locked account’s “following” count remains unchanged, it is no longer included in the “followers” count of the accounts they follow.
    For example, if Stardeath followed Katy Perry before their account was locked, Katy would still appear on their page under “following” and still be counted in their following count, but Stardeath would no longer appear on Katy’s page under “followers” or be counted in her followers count. (And because accounts – like Katy Perry’s – that had the most overall followers were also most likely to have the most followers with locked accounts- like Stardeath – these high-profile accounts generally “lost” the most followers).
  • Since the profiles that factored into the changes were already inactive, the reach of individual accounts that “lost” followers was in essence unaffected (irregardless of how much lower their follower count may be). Yes, Katy Perry for instance, had her follower count lowered 2.8 million in one day – but since those 2.8 million were inactive accounts the actual reach of her account has not changed.
  • Since Twitter was already not counting “locked accounts that have not reset their password in more than one month” in daily active users or monthly active user, Twitter’s site-wide metrics are also unaffected by this week’s changes.
  • The tweets of locked accounts do not appear Twitter searches, but tweets from unlocked accounts tagging them do. See this search of Stardeath for example.
  • This week’s changes did not target spam bots, fake followers or other nefarious actors on twitter; nor was it part of an effort to weed out “fake news.” It was about updating the reliability of follower counts as an indication of current reach.

Separate from this week’s changes, Twitter concurrently continued to target spam bots, fake followers etc as well, at an a rate that has increased in recent months according to a July 6th Washington Post article. “The rate of account suspensions, which Twitter confirmed to The Post, has more than doubled since October, when the company revealed under congressional pressure how Russia used fake accounts to interfere in the U.S. presidential election,” the report reads. “Twitter suspended more than 70 million accounts in May and June, and the pace has continued in July, according to the data.”

The suspensions detailed in the Washington’s Post’s report last week are a separate effort not to be conflated with the changes that took place this week. Since the implementation of the new policy immediately altered follower counts this week – by millions for top accounts – its impact was apparent. Conversely Twitter’s suspensions are not as many at once and therefore are less obvious. As a daily effort spanning years, the nature of individually banning accounts makes them harder to spot, even at the recently increased rate of over one million per day. Since the accounts that lose the most followers due to suspensions (of bots, fake followers, spam, or otherwise scamming accounts, etc) are also the accounts that gain the most followers daily, the loss of the banned accounts in their follower count is undetectable. An account like Barack Obama’s can be expected to have over 20 of its followers banned every day – a number that is obscured by the 20,000 new followers it can be expected to gain daily. The loss of banned accounts might however be seen in his following count. Obama’s account apparently used an auto-follow tool during the 2008 election and early in his presidency to poke others into following him back and raise his status on the platform. In the process he unwittingly followed countless bots, fake follower accounts, porn, etc. Obama’s account followed well over three-quarters of a million profiles at its peak throughout the spring of 2009. (As of the moment an archive snapshot of his twitter page was taken on June 17, 2009 by Wayback Machine he followed 774,943 accounts; unfortunately no earlier snapshot exist for that June or anytime in May 2009 so the exact high point can not be confirmed but the following count presumably peaked sometime in the month prior, as it has steadily fallen ever since). Now nine years later Obama’s following count continues to decrease – with a daily average decline of 23 for the past month – but currently still follows over 620K. In the past week, the drop in Obama’s following averaged 26 per day, with as little as 21 (on July 9th) and as much as 31 (on July 12th). Is Obama (or somebody on his staff) logging on twitter everyday to manually unfollow 23 accounts? Not likely – especially considering there’s little other activity on the account. (He hasn’t “liked” a tweet since February; in the last month he tweeted just one time and retweeted his wife once, raising his total “Tweets & replies” count by 2 in the same timeframe his following count dropped by 741). The obvious explanation is that these accounts are not being “unfollowed”, rather they are being suspended (and consequently lower the following counts of all that followed them, such as Obama).

See also: Yoko Ono – whose twitter account currently follows over 924k others. Like Obama, Yoko’s account was one of the most aggressive followers on twitter. For several years early in Yoko’s twitter account’s history it would follow-back every single account that followed it. As a strategy to build a following it worked. (Yoko currently has 4.8 million twitter followers – far more than her late husband’s 353k, surviving Beatle Paul’s 3.89 million and Ringo’s 1.86 million, or George’s 1.12 million). But it also meant untold accounts that had no interest in Yoko’s tweets followed her just because they wanted an easy follow-back. At its peak in July 2013 Yoko followed over 1 million accounts. She hasn’t employed this technique since, instead – as with Obama’s account – Yoko slowly but steadily decreases her following count apparently just by those accounts being removed from the platform. Everyday for the past month for instance Yoko’s following count has decreased by an average of 52.

From Yoko Ono we can glean that at Twitter’s
current rate of suspensions, if you follow a million accounts
about 50 of them are likely to be banned everyday.

Again, these suspended accounts (whose removals are most easily observed through the daily declines of the following counts of accounts like Obama and Ono’s) are different than the locked accounts that all at once stopped being included in follower counts this past week. This week’s change wasn’t a matter of weeding out spam, abuse or “fake news,” but rather a one of improving the reliability of followers count as a proxy for reach. For advertisers a follower count is only as useful as the degree to which it reflects the current reach of the account. That an account may have been widely followed at some point in the past does advertisers no use if those following accounts are not still active. From this perspective, an account with 1 million active followers is the same as an account with two million followers of which half are inactive. The July 12th changes updated follower counts to more accurately reflect the current reach of accounts not by deleting inactive accounts, but by removing them from the follower count of the accounts they followed, while the inactive accounts themselves remain unaffected.

Rank – Owner (Linked To Twitter Account) – Followers As Of July 14th

1. Katy Perry 107 million (down from 110 million on July 11th)
2. Justin Bieber 104 million (down from 107 million on July 11th)
3. Barack Obama 101 million (down from 104 million on July 11th)
4. Rihanna 87 million (down from 89 million on July 11th)
5. Taylor Swift 83 million (down from 86 million on July 11th)
6. Lady Gaga 76 million (down from 79 million on July 11th)
7. Ellen DeGeneres 76 million (down from 78 million on July 11th)
8. Cristiano Ronaldo 73 million (down from 74 million on July 11th)
9. YouTube 70 million (down from 72 million on July 11th)
10. Justin Timberlake 64 million (down from 66 million on July 11th)

As stated above, since the profiles that factored into the changes were already inactive, the reach of individual accounts did not change. Despite Katy Perry’s account for instance having 2.8 million less followers on July 13th than July 11th, the actual reach of the account was the same. What changed is the appearance of reach, not the reach itself – or more specifically, the degree to which accounts appear to have reach relative to others. Thus, contrary to the many headlines that suggested otherwise, it was not news that Katy Perry lost more followers than any other account except for twitter’s itself. That was the obvious, expected outcome of the policy change, reported as click-bait. Katy or other individual account’s follower count decrease as a raw number is a meaningless statistic in itself. So is the percentage of an individual account’s decrease by itself (since the rate of drop differs by total follower count). What is relevant is how much an account’s followers changed relative to other accounts that had a similar amount of followers. Among the the top ten most followed accounts for instance – shown above – the relative rank of each remained the same before and after the July 12th changes. Below the top ten however the fall of @twitter from the 11th most followed account before the changes to the 16th after is connected to the rank rises of Kim Kardashian West (from 12 to 11), Ariana Grande (from 14 to 12), Demi Lovato (from 15 to 13) and Selena Gomez (from 16 to 15); Ariana and Demi’s rank rise is also connected to the fall of Britney Spears (from 13 to 14). The fall of P!nk (from 44th to 46th) is connected to the rise of Liam Payne (45 to 44), Harry Styles (46 to 45).

The raw numbers don’t reflect the impact of the changes, how they differ among similarly ranked accounts does. @Twitter@britneyspears and @Pink were all shown to have had a larger percentage of inactive accounts among their followers than profiles that had slightly less followers than them before the changes. Since the reach of all of these accounts – whether they rose or fell in rank – is the same before and after, the accounts that have gained followers at a higher rate in recent years (like Ariana Grande and Demi Lovato) benefit by now having their reach more accurately represented by their follow count, while those that gained at a higher rate many years ago (Britney Spears and P!nk) no longer have their current reach misrepresented by a follow count inflated by inactive accounts that followed them long ago.

That Ariana Grande and Demi Lovato are ranked higher
after this week’s changes, while Britney Spears and P!nk are lower
could be predicted by when each had their biggest hits.

What’s relevant is how each account is now repositioned relative to similarly ranked accounts, which have their followings better represent their reach in 2018, and which lose a false impression of a larger reach than they actually have (distorted by years-old follows from now inactive accounts). The official twitter account dropped the most ranked positions of any account among the top 50 most followed twitter accounts: down five spots to the 16th position (now with 55 million) from the 11th most followed account (with 63 million on July 11th). Lil Wayne dropped two spots (from 39th most followed to 41st), as did P!nk (from 44th to 46th). Several other accounts dropped one spot in the top 50 since the policy changes: Britney Spears (dropped from 13 to 14), Oprah Winfrey (25 to 26), CNN (29 to 30), Niall Horan (30 to 31), Instagram (32 to 33), Kevin Hart (37 to 38), Wiz Khalifa (41 to 42), and Alicia Keys (48 to 49). The account that rose the most spots on the top 50 was Salman Khan (three spots, from 42 to 39). Other accounts rose two spots – Ariana Grande (from 14 to 12), Demi Lovato (from 15 to 13),  Neymar Jr (from 31 to 29) – or one spot: Kim Kardashian West (from 12 to 11), Selena Gomez (from 16 to 15), The New York Times (26 to 25), BBC Breaking News (33 to 32), Amitabh Bachchan (38 to 37), Liam Payne (45 to 44), Harry Styles (46 to 45), Real Madrid C.F. (49 to 49).

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s