This analysis of Elon Musk’s tweets is a fun exploration of the Tesla CEO’s social media habits. Among the takeaways is that the majority of his tweets are replies to fellow Twitter users, and that he will tweet to anyone regardless of follower count. We also feature some of Musk’s most interesting and indicative tweets for good measure.
What makes this project innovative?
This project makes use of a hot new format — the social media-inspired "tap story" — and uses it to tell a data-driven story. Combined with subtle animation, this keeps the reader engaged while maintaining a sleek, minimalist design.
What was the impact of your project? How did you measure it?
The story was considered a success internally, with almost 100,000 page views. Our analytics tell us that many of those readers found the story incredibly engaging: the average time on page was 1:35, and 26% of visitors to the page read to the final slide. It was also featured in The Verge’s daily newsletter "The Interface" (https://www.theverge.com/2018/8/1/17637540/facebook-influence-campaign-russia-liberal-activists-resisters).
Source and methodology
We collected historical tweets from Elon Musk and other big-tech CEOs using Twitter's API. For Musk’s tweets posted from June 2017 to July 2018, we read every tweet to classify it in various categories. There were two main classifications: (a) company — whether the tweet was related to one of the many companies Musk runs and if yes, which specific company. This was partly automated: we listed keywords related to specific companies and classified when the word was present in the tweet; followed by manual categorisation for the remaining set. (b) whether his replies were in response to a critic or otherwise. This was done manually—the context of the tweet was necessary to understand whether he was responding to a critic or not.
For the initial analysis, we used Python/Pandas and Google Sheets. For the visual presentation, this project uses a custom fork of WSJ’s proprietary "tap tool" format. The charts were drawn using D3. We wrote a custom D3 component for the grid animation that made generating these animated charts fast and flexible. The tweets were not embedded the typical way, but in fact recreated using HTML, in order to improve page performance and make sure that the contents are preserved.
Scott Austin, Samarth Bansal, Elliot Bentley, Jieqian Zhang