DataWeek 2013 Conference & Festival

Some Tweets receive tremendous engagement from followers while others languish in their timelines. They also arrive in a variety of contexts: day-to-day occurrences, public events like elections and sports games, and user-specific events like album releases and TV show airings. These factors complicate the question: how do I write a Tweet to have the maximum impact?

Previous studies have focused on an atomic approach, treating all Tweets identically. In this talk, we will discuss how we use Amazon Mechanical Turk to implement a multi-level study that can reveal more insight.

By combining hard and soft data, we have identified key subjective metrics that describe Tweets from multiple contexts and within many social circles. We will show that the most important piece to the puzzle is knowing your audience, and we will describe how best practices morph among Twitter users to suit the expectations and desires of their followers.