A Philippine Eagle Owl (Bubo Philippensis) is seen inside the Ninoy Aquino Parks and Wildlife Rescue Center in Quezon City, Metro Manila November 6, 2009. The center, which serves as a repository and rehabilitation facility for confiscated, donated or abandoned wildlife with the objective to release endemic and indigenous animals back to their habitat, also serves as a venue for public education and both a training and research facility for future veterinarians and biologists.REUTERS/John JavellanaTweets in decline.
Twitter  $16.93
TWTRChange-0.97%Change-5.40
Disclaimer
The number of tweets per day created by Twitter's users has fallen by more than half since a peak in August 2014, according to a sampling of data from Twitter's API. (An API — application programming interface — is the portal through which other apps access Twitter so their software can function together.) The data was given to Business Insider by an app developer who has tracked Twitter users since 2013.
Tweets per day reached a peak in August 2014 of 661 million, our source says. That 30-day sampling period included the World Cup final. In January 2016, there were only 303 million tweets per day, on average, during the 30-day period.
TwitterBusiness Insider
"This data is not correct," a spokesperson for Twitter told Business Insider. The company, which has a policy of not commenting on third-party data, declined to elaborate.
Whether it is accurate or not, the numbers could be a problem for Twitter. This is the data that app developers see when they take samples from Twitter's datasets to gauge activity on the platform. If it is accurate, it suggests that people are using the platform less frequently. If it isnot accurate, it suggests that the activity developers see on the platform is a misleading guide as to what is actually happening, which isn't helpful for developers.
The app already suffers from a growing population of "users" who don't actually do anything on it. And growth in "monthly active users" has all but stalled, at 320 million, according to Twitter.
Data samples taken by app developers have proved to be misleading in the past. In July 2014, a sample of API data leaked to Business Insider suggested Twitter may have seen a brief decline in MAUs, but that didn't happen. MAU growth did become very soft that year, and on Twitter's own metrics the population of users who are logged-in to the platform but who don't actually do anything increased.
The disconnect between the numbers Twitter discloses to the SEC and the numbers developers see via the API occurs because developers cannot see the data Twitter uses to calculate its MAUs. Developers can only see active users and accounts.
In addition, Twitter stopped reporting its estimate of users who consisted of "third-party applications that may have automatically contacted our servers for regular updates without any discernable additional user-initiated action." Twitter has not updated that estimate since December 2014, making it more difficult to estimate what portion of Twitter's user-base isn't being operated by a human.
Our source suspects that what while Twitter has retained a core of older professional users, like media professionals and politicians, it has lost a lot of younger users who tweeted a lot at their friends. Younger users create two or three times the number of tweets that older users do, our source believes. Those young people went to rival apps like Snapchat and Instagram, our source thinks. Both those apps have overtaken Twitter in monthly active users in the last couple of years.
Another possible explanation is that Twitter has become better at rooting out spam accounts that tweet on autopilot. (But that would imply that a huge volume of activity on Twitter a year ago was merely spam.) 
Broadly, the data shows the number of active users on Twitter has been flat for two years. Twitter has publicly reported similar trends in its statistics, but the company's definition of "monthly active users" counts users whose only activity on the app is performed by another appthat happens to ping Twitter for an update, without any human intervention.
Also, the data shows that Twitter's longtime problem of users who sign up for the service and then abandon it has not gone away. The total universe of Twitter accounts is up to 1.6 billion, our source's data show. Of that total, the vast majority fall silent, don't tweet, operate in "listen only" mode, or are spam bots, our source believes.
Our source took a sample of 100,000 Twitter accounts to create each data point. The data describes the total number of accounts in existence, the number of accounts that are active, the number of new accounts created, and the number of accounts that actually tweet.
While Twitter claims it has up to 320 million monthly active users, our source's data shows that active users — people actually tweeting or creating new accounts in a 30-day period — hover around the 130 million mark. That discrepancy is likely explained in part by the fact that Twitter counts as "active" any user who is logged in during the period, whereas our source counts only users who are logged in and perform an action.
Here is the same data as above, including the total universe of existing accounts:
twitterBusiness Insider
Our source believes the majority of abandoned accounts are probably spambots. There is a still a thriving ecosystem of companies that let you buy followers. Twitter takes actions to stamp those accounts out, but it does not catch them all.
Disclosure: The author owns Twitter stock.