Twitter Users’ Behaviour Found to be Highly Predictable: Study

Twitter users – including company accounts – are nearly as predictable as clockwork, a new study suggests.

Research from the Imperial College in the UK analyzed more than 160,000 tweets and created a system that predicts the nature of an account (a person, a brand, or a bot) with 83% accuracy. The system can also accurately predict when accounts can tweet.

The latter component of the study also revealed an interesting disconnect between brands marketing to Twitter users and the users themselves. According to the research, personal accounts experience peak activity in the afternoons and evenings, while brands typically tweet only during business hours, particularly mornings.

“We have measured and shown that Twitter-using individuals have a distinct and characteristic structure in their tweeting behaviour, characterised by the tails of their inter-tweet time distribution and their rather more unstructured hourly tweet probability,” the study reads.

Since its creation in 2006, Twitter has become an increasingly popular medium enabling over 500 million active users who produce 65 million tweets per day.