Twitter’s user growth isn’t as strong as it seems.
Investors have decided, for better or worse, that the best measure of Twitter’s potential is how many users it adds every three months. The stock jumped more than 20% after the company reported it had 271 million monthly active users (MAUs) at the end of June, an increase of 6.3% from the previous quarter and more than most analysts were expecting.
The report offered a different narrative from that of skeptics who thought user growth had stalled. But, really, it depends on how you define “user.” Here are some examples of what passes for an active user on Twitter.
On one hand, it makes perfect sense to include these accounts as active users: The rogue news outlet and the cat with more than a million followers are an important part of what makes Twitter so great. You enjoy following Quartz, don’t you? And many of these accounts have real-life humans behind them. Is a brand any less of a Twitter user?
Well, in some crucial respects, yes. If MAUs are used to judge Twitter’s growth—how broadly the company has spread to people across the globe—then accounts representing brands and other non-humans seem beside the point. And they are nearly irrelevant to advertisers, who are generally interested in reaching people, not other brands. Most importantly, including brand accounts among MAUs leads to lots of double-counting of actual people. (I actively tweet from four accounts myself.)
It’s impossible to say what portion of Twitter’s active users aren’t human, and in fairness to Twitter, estimating that would be very difficult. But a glance at most Twitter feeds suggests it’s a large portion, indeed. That makes it fallacious to compare Twitter’s MAUs to Facebook, which only allows human accounts, though some animals slip in. On Facebook, brands must use “pages,” which aren’t considered separate users.
A related issue is automated accounts. Buried in Twitter’s quarterly earnings reports, the company discloses what percentage of its MAUs “used applications that have the capability to automatically contact our servers for regular updates.” Translation: how many active accounts access Twitter outside of the official website and mobile apps, using Twitter’s API.
That includes lots of accounts that read or send tweets with apps Twitter doesn’t own. Those are legitimate accounts, probably more active than a typical user. But this category also includes accounts that are totally automated. These accounts can be very useful—sending alerts of earthquakes, new articles, the top of the hour, or anonymous Wikipedia edits—but they can hardly be considered human. They are bots.
When Twitter was preparing to go public last year, 7% of its active users used the API. The company also said (p. 49) it expected that percentage to ”decline over time, particularly as usage of our mobile applications increases.”
But actually, the portion of Twitter’s MAUs using the API has doubled, to 14%. That segment of users is growing much faster than MAUs that don’t use Twitter’s API. Those now represent 37.9 million active accounts, up 148% in the past year.
If you remove these API accounts from Twitter’s results, which isn’t entirely fair, last quarter’s MAU growth was only 3.9% over the previous quarter, instead of 6.3%.
Again, using Twitter’s API doesn’t necessarily make an account a bot. And bots are totally different from spam, which Twitter estimates to represent under 5% of its accounts. But the explosion in this category of user should raise eyebrows, particularly after Twitter predicted the opposite would happen.
It stands to reason that the enormous growth in these types of accounts isn’t coming from humans integrating their accounts with other services, especially as Twitter has clamped down on third-party apps using its API. For what it’s worth, RBC analyst Mark Mahaney disagrees with me. In a note after Twitter’s quarterly earnings this week, he wrote:
Our interpretation is that this implies that a growing number of users are using apps to refresh/access with Twitter content. These Users are making more of an effort to engage with Twitter. That’s actually a good thing.
BTIG analyst Rich Greenfield interpreted the data less charitably (registration required). Even if these accounts are humans, he wrote, “Off-Twitter activity is important to understand as those MAUs are not currently being monetized.” That is, advertising sold by Twitter, which is growing strongly, doesn’t reach accounts using the API—for instance, on Flipboard or Tweetbot. That’s the chief reason Twitter started restricting usage of its API in the first place.
Ultimately, it’s not possible to estimate how many of Twitter’s monthly active users are actually human. But it’s fewer than 271 million, and there’s reason to suspect that recent growth has included lots of accounts that don’t have a pulse.
One lesson here is that MAUs aren’t a good way to measure Twitter. The company has begun arguing that it should be judged by how many people encounter tweets on the rest of the web and even television, and it may start reporting data along those lines. It also has an existing metric—timeline views—that serves as a better measure of actual use of Twitter by humans. That figure is up 15% over the past year, compared to 24% growth in MAUs.
Even better would be if Twitter disclosed how much time people spend with the service. Facebook recently disclosed that its average user in the US spends 40 minutes a day there. That’s a mind-blowing statistic against which all other companies that sell advertising should be measured.
* * *
The photo above, from Twitter’s new headquarters, is by Scott Beale/Laughing Squid and used under a Creative Commons license.